Brex vs Navan

Accounting sync

1. finance leadTransactions sync to QuickBooks, NetSuite, or Xero with GL account, class, and department mappings I control

   weight 3 · round drawn
   Brex—none0/10

   The evidence pack shows Brex has a one-click accounting sync feature, but it is documented only as syncing to Puzzle (brex-docs-17, 18, 40), with no mention of QuickBooks, NetSuite, or Xero, nor any documentation of finance-lead-controlled GL account, class, or department mapping for such syncs. Cost-center/department fields exist in the user API filters (brex-docs-19, 33) but these aren't tied to any accounting-sync mapping capability.

   • [claimed-docs] “Prepared expense data sync automatically to eliminate manual set up or matching between Brex and Puzzle.”
   • [claimed-docs] “Get global card and a treasury product that earns yield†, plus modern accounting from Puzzle.”
   • [claimed-docs] “Set up your accounting system in one click.”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”

   proof ↗⚑ flag

   Navan—none0/10

   The evidence pack contains no mention of QuickBooks, NetSuite, Xero, or any accounting-system sync, nor GL account/class/department mapping controls — it only covers Navan's API/MCP for expense querying, travel booking, and rewards. This is a plausible and expected axis for a T&E platform, but nothing in the evidence demonstrates it.

   ⚑ flag

Close

1. finance leadSpend is coded as it happens — merchant, category, memo, receipt — so month-end close is a review, not an archaeology dig

   weight 2 · round to Brex
   Brex✓fullclaimed8/10

   Brex documents real-time spend coding at point of transaction: receipt auto-matching to expenses (brex-docs-7, brex-docs-27), AI agents auto-categorizing/approving and writing memos (brex-docs-23, brex-docs-16, brex-docs-41), policy/budget enforcement applied per transaction (brex-docs-8, brex-docs-15, brex-docs-36), and explicit close-in-real-time messaging (brex-docs-37) plus one-click accounting sync (brex-docs-17, brex-docs-40). This directly maps to merchant/category/memo/receipt capture happening automatically rather than at month-end. Missing for 10: independent/hands-on verification that AI-drafted memos and auto-categorization are accurate in practice, and no third-party case study confirming reduced close time.

   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “Upload receipt and match automatically.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Automatically block out-of-policy purchases, with preset rules and limits.”
   • [claimed-docs] “Manage global spend and close the books in real time — with unmatched automation of manual expense work.”
   • [claimed-docs] “Prepared expense data sync automatically to eliminate manual set up or matching between Brex and Puzzle.”
   • [claimed-docs] “Set up your accounting system in one click.”

   proof ↗⚑ flag

   Navan~partialclaimed7/10

   Navan's card-transaction API, receipt scanning, custom-field management, and policy-violation flagging (navan-docs-4, -21, -22, -23, -24) show spend arriving already tagged by merchant/category/policy status, and MCP queries let a finance lead pull flagged/over-limit items by department in natural language (navan-docs-5, -21) instead of manually reconciling. Missing for 10: explicit evidence of automatic memo generation and confirmation that coding happens in real time at swipe rather than batch/after-the-fact, plus independent (non-vendor) validation of close-time accuracy.

   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Analyze spend data — query expenses, identify trends, and break down costs by department, vendor, or category.”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”
   • [claimed-docs] “Receipt scanning for on-the-go expenses”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “Card management — see which cards are active or inactive, how many cards a user has, and other card details.”

   proof ↗⚑ flag

Multi entity

1. finance leadRun multiple legal entities and currencies in one account with consolidated reporting and per-entity books

   weight 2 · round to Brex
   Brex~partialclaimed3/10

   Evidence shows Brex's API supports a `legal_entity_id` field for filtering users, implying some multi-entity data modeling exists, and general mentions of 'global spend'/'more currencies' support, but there is no documentation of consolidated cross-entity reporting or dedicated per-entity books for finance close workflows. missing for 10: consolidated reporting across entities, per-entity books/ledgers, multi-currency accounting close workflow, any first-party or independent confirmation of multi-entity account structure.

   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “Manage global spend and close the books in real time — with unmatched automation of manual expense work.”
   • [claimed-docs] “Create a virtual card instantly to pay by card and earn rewards. Send ACH, checks, and wires from any bank account or a Brex business accoun…”
   • [claimed-docs] “Get global card and a treasury product that earns yield†, plus modern accounting from Puzzle.”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence pack covers Navan's MCP/API for expense and travel data, card management, and pricing/rewards, but contains no mention of multi-entity or multi-currency account structures, consolidated reporting, or per-entity books — core requirements for finance-lead accounting-close workflows.

   ⚑ flag

Agent access

1. ai-native userPoint an agent at llms.txt or agent-oriented docs

   weight 2 · round drawn
   Brex✓fullprobed9/10

   Brex hosts a live llms.txt at developer.brex.com/llms.txt (confirmed via HTTP 200 probe) that provides a structured table of contents for AI agents, and it also documents an official MCP server for agentic access to Brex data, showing clear agent-oriented documentation design. Missing for 10: independent third-party confirmation that agents actually consume llms.txt successfully in practice.

   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.brex.com/llms.txt # Brex Developer API > APIs to build on top of the Brex financial OS. ## T…”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [probe] “official MCP server documented at https://developer.brex.com/docs/mcp”

   proof ↗⚑ flag

   Navan✓fullprobed9/10

   A live llms.txt file is confirmed at developer.navan.com/llms.txt (HTTP 200) describing MCP-based agent access, and the developer portal has dedicated agent-oriented docs (/mcp/, /api/) with example natural-language queries. Missing for 10: independent third-party confirmation that agents successfully consume this llms.txt in practice.

   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.navan.com/llms.txt # Navan Developer Portal > Connect AI assistants to Navan travel and expen…”
   • [claimed-docs] “Connect AI assistants — Claude, Cursor, ChatGPT, Codex, OpenClaw — to Navan data using the Model Context Protocol.”
   • [claimed-docs] “The Navan MCP lets you connect Navan to your favorite AI products, so you can query Navan spend and travel data using natural language.”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”

   proof ↗⚑ flag

2. ai-native userRun the product headlessly / in CI for automation

   weight 2 · round to Brex
   Brex~partialclaimed6/10

   Brex exposes a full REST API with token-based auth (no interactive login required), webhook subscriptions, and Postman/Zapier/Workato/Pipedream automation examples (bulk card creation, scheduled exports), all of which support headless, non-interactive use suitable for scripted/CI-style automation. However there is no explicit mention of CI/CD pipelines, SDKs, or headless test/deploy tooling — missing for 10: explicit CI/CD integration docs, official SDKs for scripting, and confirmation of non-interactive token refresh/rotation suitable for unattended pipelines.

   • [claimed-docs] “Sign in to dashboard.brex.com as an account admin or card admin... Click Create Token. Create a name for your token... Choose what level of …”
   • [claimed-docs] “generate a user token from your Brex dashboard and pass it along in your API call headers”
   • [claimed-docs] “Register a URL to receive the webhooks using the [register webhook] endpoint.”
   • [claimed-docs] “Register an endpoint to start receiving selected webhook events”
   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”
   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Monthly Export of Brex Card Transactions to CSV in Google Drive”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”

   proof ↗⚑ flag

   Navan~partialclaimed5/10

   Navan's API documents an OAuth 2.0 client-credentials flow (machine-to-machine auth with no interactive login), which is the kind of mechanism needed to call Navan programmatically from a CI/automation pipeline, and its REST/webhook endpoints could be scripted headlessly. However, there is no explicit CI documentation, CLI, or example of running Navan (or its MCP server) unattended in a pipeline — missing for 10: explicit CI/automation guide, headless MCP server deployment instructions, and any hands-on evidence of it being run in CI.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”

   proof ↗⚑ flag

3. ai-native userPlug MCP servers into this product so it can use their tools

   weight 3 · round drawn
   Brex—none0/10

   Evidence only shows Brex publishing its own MCP server so external AI assistants can call Brex's tools (docs-3, docs-24, probe-3) — the reverse of this story, which asks whether Brex can consume/plug in external MCP servers to use their tools. No evidence indicates Brex itself acts as an MCP client or lets users register third-party MCP servers for its AI agents to use.

   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [probe] “official MCP server documented at https://developer.brex.com/docs/mcp”

   proof ↗⚑ flag

   Navan—none0/10

   All evidence describes Navan exposing its own MCP server so external AI assistants (Claude, Cursor, ChatGPT) can query Navan's travel/expense data — this is Navan acting as the MCP server/provider, not as a client consuming external MCP servers' tools. There is no evidence Navan (or its Ava assistant) can plug in and use tools from third-party MCP servers. Missing for 10: any documentation of Navan's AI features connecting to external MCP servers, any tool/plugin ecosystem for consuming outside MCP tools, and any hands-on confirmation of such client-side behavior.

   • [claimed-docs] “Connect AI assistants — Claude, Cursor, ChatGPT, Codex, OpenClaw — to Navan data using the Model Context Protocol.”
   • [claimed-docs] “The Navan MCP lets you connect Navan to your favorite AI products, so you can query Navan spend and travel data using natural language.”
   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.navan.com/llms.txt # Navan Developer Portal > Connect AI assistants to Navan travel and expen…”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”

   proof ↗⚑ flag

4. ai-native userConnect an agent via an official MCP server

   weight 3 · round drawn
   Brex✓fullprobed8/10

   Brex is not itself an AI agent but a financial platform, so publishing an official MCP server is a fair ecosystem capability, and Brex documents a first-party MCP server letting AI assistants manage expenses, users, cards, transactions, and bills. Missing for 10: independent/hands-on confirmation of setup and reliability beyond vendor docs.

   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [probe] “official MCP server documented at https://developer.brex.com/docs/mcp”

   proof ↗⚑ flag

   Navan✓fullprobed8/10

   Navan publishes an official MCP server (developer.navan.com/mcp/) explicitly for connecting AI assistants like Claude, Cursor, ChatGPT to Navan spend/travel data, with documented natural-language query examples and OAuth-secured API endpoints backing it. Missing for 10: independent/hands-on third-party corroboration of the MCP server working in practice beyond vendor docs.

   • [claimed-docs] “Connect AI assistants — Claude, Cursor, ChatGPT, Codex, OpenClaw — to Navan data using the Model Context Protocol.”
   • [claimed-docs] “The Navan MCP lets you connect Navan to your favorite AI products, so you can query Navan spend and travel data using natural language.”
   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.navan.com/llms.txt # Navan Developer Portal > Connect AI assistants to Navan travel and expen…”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “What's our company policy on rideshare, and which transactions from last month violate it?”

   proof ↗⚑ flag

5. ai-native userUse an official CLI

   weight 2 · round drawn
   Brex—none0/10

   Evidence pack shows Brex's developer platform (APIs, Postman collection, MCP server, Zapier/Pipedream/Workato integrations) but no mention of an official CLI tool anywhere in docs or changelog.

   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”
   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Monthly Export of Brex Card Transactions to CSV in Google Drive”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.brex.com/llms.txt # Brex Developer API > APIs to build on top of the Brex financial OS. ## T…”

   proof ↗⚑ flag

   Navan—none0/10

   The axis applies to this product kind (peer products hold positive or none verdicts on this story), so lack of evidence for an applicable capability is "none", never "na". (na/none harmonized at arena bring-up — see pipeline/scripts/na-harmonize.ts.)

   ⚑ flag

6. ai-native userDrive the product through a documented public API

   weight 3 · round drawn
   Brex✓fullprobed8/10

   Brex has extensive documented public REST APIs (transactions, budgets, expenses, payments, onboarding, webhooks) with authentication via dashboard-generated tokens, Postman collections, and a llms.txt developer index, giving AI-native users a clear path to programmatically drive the product. Missing for 10: a live discoverable OpenAPI spec (probe found 404s on standard OpenAPI paths) and independent/hands-on third-party corroboration of API robustness beyond vendor docs.

   • [claimed-docs] “Sign in to dashboard.brex.com as an account admin or card admin... Click Create Token. Create a name for your token... Choose what level of …”
   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “generate a user token from your Brex dashboard and pass it along in your API call headers”
   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”
   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.brex.com/llms.txt # Brex Developer API > APIs to build on top of the Brex financial OS. ## T…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.brex.com/openapi.json, https://developer.brex.com/swagger.json, https://developer.…”

   proof ↗⚑ flag

   Navan✓fullprobed8/10

   Navan provides a documented public API (developer.navan.com) with OAuth 2.0 client-credentials auth, retrieval/update endpoints across expense, card, payroll, custom fields, and webhooks, plus a first-party MCP server for natural-language AI access — corroborated by a live llms.txt probe. Missing for 10: independent third-party developer accounts of building against the API and clearer rate-limit/versioning documentation.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”
   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.navan.com/llms.txt # Navan Developer Portal > Connect AI assistants to Navan travel and expen…”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”
   • [claimed-docs] “Connect AI assistants — Claude, Cursor, ChatGPT, Codex, OpenClaw — to Navan data using the Model Context Protocol.”

   proof ↗⚑ flag

7. ai-native userIssue scoped/least-privilege API credentials for an agent

   weight 2 · round to Brex
   Brex✓fullprobed7/10

   Brex's developer docs show admins can create named API tokens and choose specific data-access scopes at creation time (brex-docs-2, brex-docs-26), which is a direct least-privilege credentialing mechanism that would apply to any API consumer, including an AI agent calling the API or via the documented MCP server (brex-docs-3, brex-probe-3). Missing for 10: an explicit enumerated list of available scopes, agent-specific credential guidance, per-agent revocation/audit trail details, and independent corroboration of scope enforcement in practice.

   • [claimed-docs] “Sign in to dashboard.brex.com as an account admin or card admin... Click Create Token. Create a name for your token... Choose what level of …”
   • [claimed-docs] “generate a user token from your Brex dashboard and pass it along in your API call headers”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [probe] “official MCP server documented at https://developer.brex.com/docs/mcp”

   proof ↗⚑ flag

   Navan~partialprobed3/10

   Navan's developer docs mention an OAuth 2.0 client-credentials flow for API access (navan-docs-2), which implies some credential-based authentication, but there is no explicit documentation of scoped permissions, granular access levels, or least-privilege controls specifically for AI agents connecting via MCP or API. missing for 10: explicit scope/permission definitions, least-privilege role configuration, agent-specific credential restrictions, and any documentation or independent verification of access-control granularity.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Connect AI assistants — Claude, Cursor, ChatGPT, Codex, OpenClaw — to Navan data using the Model Context Protocol.”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”

   proof ↗⚑ flag

8. ai-native userBuild against official SDKs

   weight 2 · round to Navan
   Brex—none0/10

   Brex documents REST APIs (OpenAPI specs), authentication, webhooks, a Postman workspace, and third-party integrations (Pipedream, Zapier, Workato), but nowhere in the evidence pack is there mention of official client-library SDKs (e.g., Node.js, Python, Java packages) that AI-native developers could build against.

   ⚑ flag

   Navan~partialprobed5/10

   Navan documents an official REST API (OAuth2 client-credentials, expense/card/travel endpoints, webhooks) and an MCP server for AI assistants, which developers can build against, but there is no evidence of language-specific official SDKs (e.g., Python/JS client libraries) — only raw API/MCP documentation. missing for 10: explicit SDK packages/libraries, code samples in multiple languages, independent developer corroboration of SDK usage.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”
   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.navan.com/llms.txt # Navan Developer Portal > Connect AI assistants to Navan travel and expen…”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”

   proof ↗⚑ flag

9. ai-native userSubscribe to events via webhooks

   weight 2 · round to Brex
   Brex✓fullclaimed8/10

   Brex documents a dedicated webhooks API and guide for registering endpoints and receiving selected events, including examples like referral activation, indicating genuine subscribe-to-event support. missing for 10: no independent/hands-on corroboration of webhook delivery reliability, and no explicit AI-native/agent-triggered subscription workflow shown.

   • [claimed-docs] “Register a URL to receive the webhooks using the [register webhook] endpoint.”
   • [claimed-docs] “you can receive a webhook when a referral you made has activated”
   • [claimed-docs] “Register an endpoint to start receiving selected webhook events”

   proof ↗⚑ flag

   Navan—none0/10

   Webhooks are mentioned only once in passing as part of the API surface (navan-docs-2), with no documentation of subscription mechanics, event types, payload formats, or delivery guarantees; nothing shows an AI-native user can actually subscribe to events via webhooks.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

Agentic features

1. ai-native userGet AI-generated insights and suggestions from my data inside the product

   weight 2 · round to Navan
   Brex✓fullclaimed7/10

   Brex AI features (audit agent, review agent, policy-based auto-approval, PO/payment matching suggestions) are documented as generating insights and suggestions directly from account/expense data. Evidence is first-party marketing/docs only, with no independent hands-on corroboration of these AI insight features in practice. missing for 10: independent/hands-on verification of AI insight quality, UI examples showing generated suggestions in context.

   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Brex AI will capture every itemized detail, suggest the matching PO, and draft a payment.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”

   proof ↗⚑ flag

   Navan✓fullclaimed8/10

   Navan's Ava AI and MCP integration provide AI-generated insights such as analyzing travel/spend data, summarizing by category, comparing to policy, predicting future spend, and flagging policy violations, all directly from Navan data. missing for 10: independent/hands-on verification of insight quality beyond vendor blog and docs, and details on how proactive/unprompted these suggestions are versus query-driven.

   • [claimed-docs] “Analyze travel spend and craft suggestions to save”
   • [claimed-docs] “Summarize travel spend by month and by category”
   • [claimed-docs] “Analyze travel spend and compare it to company policy”
   • [claimed-docs] “Predict future travel spend habits with ease”
   • [claimed-docs] “Analyze spend data — query expenses, identify trends, and break down costs by department, vendor, or category.”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “What's our company policy on rideshare, and which transactions from last month violate it?”

   proof ↗⚑ flag

2. ai-native userSet up automations that run autonomously in the background

   weight 2 · round to Brex
   Brex✓fullclaimed7/10

   Brex documents autonomous background agents (audit agent monitoring expenses, review agent auto-approving low-risk items and escalating exceptions) plus webhook events and third-party automation platforms (Zapier, Workato, Pipedream) that can run unattended once configured, and an MCP server for AI assistant integration. This covers the core 'runs autonomously in background' story via policy-driven agents and event-driven webhooks/automations. missing for 10: independent/hands-on verification that these agents truly run unattended without human triggers, and technical detail on scheduling/trigger architecture beyond marketing copy

   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “Register a URL to receive the webhooks using the [register webhook] endpoint.”
   • [claimed-docs] “Register an endpoint to start receiving selected webhook events”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence shows Navan's AI features (MCP, Ava) are query/chat-based tools that answer questions or analyze spend on request, and Navan Edge explicitly states it 'will always ask for your explicit confirmation before booking, changing, or canceling any trip,' which is the opposite of autonomous background automation. No workflow, trigger, or scheduling mechanism for unattended background automations is documented anywhere in the pack.

   • [claimed-docs] “Navan Edge will always ask for your explicit confirmation before booking, changing, or canceling any trip.”
   • [claimed-docs] “No. Navan Edge will always ask for your explicit confirmation before booking, changing, or canceling any trip.”
   • [claimed-docs] “The Navan MCP lets you connect Navan to your favorite AI products, so you can query Navan spend and travel data using natural language.”
   • [claimed-docs] “Analyze spend data — query expenses, identify trends, and break down costs by department, vendor, or category.”
   • [claimed-docs] “Unlimited policy and approval workflows”

   proof ↗⚑ flag

3. ai-native userDelegate tasks to a built-in AI assistant inside the product

   weight 3 · round drawn
   Brex✓fullclaimed7/10

   Brex documents built-in AI agents (audit agent, review agent, Brex AI) that autonomously categorize expenses, auto-approve low-risk spend, write memos, fetch receipts, and file reimbursements, allowing users to delegate financial tasks directly within the product. This is distinct from the MCP server (which is for external AI clients) and represents genuine in-product agentic delegation. Missing for 10: independent/hands-on verification of these AI agent features actually working, and more detail on how users configure/invoke the assistant.

   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Brex AI will capture every itemized detail, suggest the matching PO, and draft a payment.”

   proof ↗⚑ flag

   Navan✓fullclaimed7/10

   Navan documents built-in AI assistants (Ava for analyzing/summarizing/predicting travel spend, and Navan Edge for booking/changing/canceling trips with confirmation) that users can delegate expense and travel tasks to. Missing for 10: independent/hands-on verification of these assistants in action and detail on broader task types beyond travel/expense.

   • [claimed-docs] “Analyze travel spend and craft suggestions to save”
   • [claimed-docs] “Summarize travel spend by month and by category”
   • [claimed-docs] “Analyze travel spend and compare it to company policy”
   • [claimed-docs] “Predict future travel spend habits with ease”
   • [claimed-docs] “Navan Edge provides 24/7 access to human travel agents via chat or phone at +1-877-536-3343 in English.”
   • [claimed-docs] “Navan Edge will always ask for your explicit confirmation before booking, changing, or canceling any trip.”
   • [claimed-docs] “Navan Edge personalizes results based on: Your preferences shared during chat, Your booking history, Loyalty programs connected to your acco…”

   proof ↗⚑ flag

4. ai-native userOperate the product with natural-language commands

   weight 2 · round to Navan
   Brex~partialprobed7/10

   Brex documents a first-party MCP server explicitly built so 'AI assistants interact directly with your Brex account... through natural language' for expenses, cards, banking, and bills, plus AI agents (audit/review) that act on policies automatically. This directly supports natural-language operation for AI-native users, though it's vendor-documented only. Missing for 10: independent/hands-on evidence of the MCP server's natural-language commands working reliably, and detail on command coverage limits.

   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [probe] “official MCP server documented at https://developer.brex.com/docs/mcp”

   proof ↗⚑ flag

   Navan✓fullprobed7/10

   Navan's MCP server and Ava assistant let users query spend, travel, policy, and card data with natural-language commands (e.g. flagged expenses, rideshare policy violations), and Navan Edge/Ava support conversational trip booking with confirmation steps. missing for 10: independent/hands-on verification of natural-language accuracy and breadth beyond vendor docs, and no evidence of NL commands for arbitrary in-app actions beyond travel/expense domains.

   • [claimed-docs] “Connect AI assistants — Claude, Cursor, ChatGPT, Codex, OpenClaw — to Navan data using the Model Context Protocol.”
   • [claimed-docs] “The Navan MCP lets you connect Navan to your favorite AI products, so you can query Navan spend and travel data using natural language.”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “What's our company policy on rideshare, and which transactions from last month violate it?”
   • [claimed-docs] “Analyze spend data — query expenses, identify trends, and break down costs by department, vendor, or category.”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “Analyze travel spend and craft suggestions to save”
   • [claimed-docs] “Navan Edge will always ask for your explicit confirmation before booking, changing, or canceling any trip.”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”

   proof ↗⚑ flag

Api quality

1. ai-native userExplore an interactive API reference with runnable examples

   weight 2 · round to Brex
   Brex~partialprobed4/10

   Brex provides detailed API reference docs (openapi/*.md pages) and a Postman workspace with collections for all APIs, which allows some hands-on exploration, but there's no evidence of an interactive in-browser API reference with embedded runnable examples (e.g., a Swagger/OpenAPI explorer); a probe for standard openapi.json/swagger endpoints returned 404s. Missing for 10: an embedded interactive console/try-it feature, confirmation of runnable code snippets directly in docs, and independent corroboration of hands-on use.

   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”
   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.brex.com/openapi.json, https://developer.brex.com/swagger.json, https://developer.…”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence shows a developer portal describing API endpoints (OAuth, expense/card endpoints, webhooks) and an MCP server for natural-language queries, but nothing indicates an interactive API reference with runnable/try-it-now code examples (e.g., Swagger/OpenAPI explorer or live sandbox).

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”

   proof ↗⚑ flag

2. ai-native userDownload a machine-readable API spec (OpenAPI or equivalent)

   weight 2 · round to Brex
   Brex~partialprobed4/10

   Brex publishes API reference docs under an /openapi/ path pattern (transactions_api.md, budgets_api.md, expenses_api.md, etc.) and offers a Postman collection covering all Brex APIs, both of which imply structured, machine-readable spec data. However, a direct probe for standard OpenAPI/Swagger JSON files (openapi.json, swagger.json, etc.) returned 404s, meaning there is no confirmed single downloadable OpenAPI spec file. Missing for 10: a discoverable raw OpenAPI/Swagger JSON or YAML file, and confirmation that the Postman collection is exported in OpenAPI format.

   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.brex.com/openapi.json, https://developer.brex.com/swagger.json, https://developer.…”

   proof ↗⚑ flag

   Navan—none0/10

   Navan's developer docs describe REST API endpoints, OAuth 2.0 flows, and an MCP server, but no evidence pack item mentions a downloadable OpenAPI/Swagger spec or any machine-readable API definition file. missing for 10: explicit OpenAPI/Swagger spec file, API reference generated from a spec, or a documented download link.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”

   proof ↗⚑ flag

3. ai-native userTest against a sandbox environment without touching production data

   weight 1 · round drawn
   Brex—none0/10

   No evidence of a sandbox/test environment separate from production data; the docs pack covers authentication, webhooks, MCP server, and API endpoints but never mentions a sandbox mode or test credentials. Missing for 10: any mention of a sandbox environment, test API keys, or documentation distinguishing test vs. production data.

   ⚑ flag

   Navan—none0/10

   No evidence pack item mentions a sandbox, staging, or test environment for Navan's API, MCP server, or app; all endpoints and integrations described operate on live/production expense, travel, and card data. Missing for 10: any mention of a sandbox/test mode, staging API keys, or demo data environment separate from production.

   ⚑ flag

4. ai-native userRely on versioned APIs with a documented deprecation policy

   weight 2 · round drawn
   Brex—none0/10

   The evidence shows a changelog tracking feature additions (e.g., new filter params, policy fields) but nothing documenting API versioning scheme or a deprecation policy for breaking changes. No mention of version numbers, sunset timelines, or backward-compatibility guarantees appears anywhere in the pack.

   • [claimed-docs] “Added `policy` to booking objects, exposing whether a booking was out of policy and which travel policy rules it violated.”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”

   proof ↗⚑ flag

   Navan—none0/10

   The evidence pack documents Navan's API/MCP endpoints and OAuth flow but contains no mention of API versioning scheme or a documented deprecation policy for consumers to rely on.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”

   proof ↗⚑ flag

Ai coding

1. ai-native userAI codes and audits expenses for me — reading receipts, suggesting categories and memos, and catching duplicates or fraud

   weight 2 · round to Brex
   Brex✓fullclaimed8/10

   Brex AI explicitly offers audit and review agents that read receipts, categorize/flag violations by risk level, auto-approve low-risk expenses, and write memos/fetch receipts/file reimbursements automatically, matching the story closely. missing for 10: independent/hands-on verification of accuracy (duplicate/fraud catching specifically) and no third-party corroboration beyond vendor marketing copy.

   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”

   proof ↗⚑ flag

   Navan~partialprobed6/10

   Navan's MCP/API and Ava AI assistant let AI query flagged expenses, analyze spend, and check policy violations (navan-docs-5, navan-docs-6, navan-docs-22), plus native receipt scanning (navan-docs-24) — covering audit/flagging aspects. However there's no explicit evidence of AI auto-suggesting categories/memos or specifically detecting duplicate/fraudulent transactions. Missing for 10: automated category/memo suggestion, explicit duplicate-detection or fraud-catching functionality, and independent verification of these AI audit claims.

   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “What's our company policy on rideshare, and which transactions from last month violate it?”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “Receipt scanning for on-the-go expenses”
   • [claimed-docs] “Analyze spend data — query expenses, identify trends, and break down costs by department, vendor, or category.”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”

   proof ↗⚑ flag

Ai policy

1. ai-native userAn AI assistant enforces policy conversationally — chasing missing receipts, explaining declines, and answering "can I expense this?" before the spend happens

   weight 2 · round to Brex
   Brex~partialclaimed7/10

   Brex explicitly markets AI agents that follow up on out-of-compliance spend, fetch receipts, and offer policy help ('AI assistants do employees' expenses... fetching receipts, offering policy help' and 'review agent... follows up with anyone out of compliance'), plus an MCP server letting assistants query expenses/policy conversationally. However, there's no concrete evidence of the assistant proactively answering 'can I expense this?' before spend occurs or explaining specific decline reasons conversationally. Missing for 10: documented pre-spend conversational policy checks, explicit decline-explanation flows, and independent/hands-on confirmation of these AI behaviors in practice.

   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”

   proof ↗⚑ flag

   Navan~partialclaimed5/10

   The Navan MCP explicitly supports conversational policy Q&A — asking about policy, approval flows, and why a transaction was flagged/declined (navan-docs-22, navan-docs-6) — and Ava can compare travel spend to policy after the fact (navan-docs-19). However there is no evidence of proactive 'chasing missing receipts' workflows or a pre-spend 'can I expense this?' check before the transaction occurs; the examples shown are retrospective queries on past transactions/spend. Missing for 10: proactive missing-receipt nudges/reminders, and evidence of a pre-purchase 'is this expensable' conversational check.

   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “What's our company policy on rideshare, and which transactions from last month violate it?”
   • [claimed-docs] “Analyze travel spend and compare it to company policy”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

1. ai-native userPerform bulk operations across many items at once

   weight 2 · round to Brex
   Brex✓fullclaimed8/10

   Brex documents explicit bulk operations via automation platforms (bulk create virtual cards via Pipedream, bulk create users/set limits via Workato) and supports filtering/listing at scale via API query params, plus bulk-friendly endpoints for users, cards, vendors, and expenses. This is corroborated by first-party developer docs across multiple integration guides, not just a single mention. Missing for 10: no native in-product bulk-action UI evidence or independent hands-on confirmation of bulk API throughput/limits.

   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”

   proof ↗⚑ flag

   Navan~partialclaimed4/10

   The API docs mention batch-managing custom field option values and polling async job status, implying some bulk-processing capability, but there is no evidence of general bulk operations (e.g., bulk-approving multiple expenses, bulk-updating many transactions) across the platform for an AI-native user. missing for 10: evidence of bulk approve/reject workflows, bulk transaction updates via MCP, and independent confirmation that bulk API calls work at scale.

   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

2. ai-native userDefine rules that trigger actions automatically on events

   weight 3 · round to Brex
   Brex~partialclaimed7/10

   Brex offers webhooks that fire on events (brex-docs-1, brex-docs-25, brex-docs-35) plus built-in policy rules that auto-trigger actions like blocking out-of-policy spend, auto-approving low-risk expenses, and escalating exceptions (brex-docs-8, brex-docs-9, brex-docs-16, brex-docs-36, brex-docs-41). Third-party integrations (Zapier, Workato, Pipedream) extend this into custom event-triggered automations (brex-docs-30, brex-docs-31, brex-docs-32). Missing for 10: a native no-code rule-builder UI for arbitrary custom triggers/actions beyond preset expense/travel/approval policies, and independent hands-on confirmation that webhook-to-action automations work reliably in practice.

   • [claimed-docs] “Register a URL to receive the webhooks using the [register webhook] endpoint.”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Create limits for stipends, travel, procurement, and more. Auto-enforced controls empower employees to spend with confidence.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Automatically block out-of-policy purchases, with preset rules and limits.”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Monthly Export of Brex Card Transactions to CSV in Google Drive”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [claimed-docs] “Register an endpoint to start receiving selected webhook events”

   proof ↗⚑ flag

   Navan~partialclaimed4/10

   Navan documents configurable 'policy and approval workflows' and webhook endpoints, which imply some rule-based triggering of actions (e.g., flags, approvals) on expense events, but there is no explicit documentation of a user-facing rule/automation builder for defining custom event-triggered actions. Missing for 10: dedicated automation/rules engine documentation, examples of custom trigger-action definitions, and independent verification of automation depth beyond basic policy enforcement.

   • [claimed-docs] “Unlimited policy and approval workflows”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

3. ai-native userSchedule recurring jobs or workflows

   weight 2 · round to Brex
   Brex~partialclaimed3/10

   Brex offers webhooks for event-driven automation and third-party integration examples (Zapier, Workato, Pipedream) that can implement recurring workflows like a 'Monthly Export of Brex Card Transactions to CSV', but there is no native Brex scheduling/cron API for recurring jobs — the scheduling logic lives entirely in external tools. Missing for 10: a first-party recurring-job/schedule API or documented native cron-like trigger, and any hands-on evidence that AI-native users actually configure recurring automations directly through Brex rather than third-party platforms.

   • [claimed-docs] “Monthly Export of Brex Card Transactions to CSV in Google Drive”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Register a URL to receive the webhooks using the [register webhook] endpoint.”
   • [claimed-docs] “Register an endpoint to start receiving selected webhook events”

   proof ↗⚑ flag

   Navan—none0/10

   Navan's evidence covers MCP-based querying, expense/travel management, webhooks, and async job polling, but there is no mention of scheduling recurring jobs, cron-like automation, or persistent workflow triggers for AI-native users. Absence of evidence for this applicable automation-depth capability yields none.

   ⚑ flag

4. ai-native userVersion, review, and roll back my automations

   weight 1 · round drawn
   Brex—none0/10

   No evidence of any versioning, review workflow, or rollback mechanism for automations/policies/agents — the docs describe automation features (approvals, expense policies, AI agents, MCP) but nothing about version history, change review, or reverting configurations.

   ⚑ flag

   Navan—none0/10

   No evidence of any versioning, review, or rollback mechanism for automations/workflows in Navan or its MCP/API offerings; the pack covers expense/travel management, MCP querying, and API endpoints but nothing about automation version control or rollback.

   ⚑ flag

Bill pay

1. finance leadRun accounts payable in the same platform — capture invoices, route approvals, and pay vendors by ACH, check, or wire

   weight 2 · round to Brex
   Brex✓fullclaimed8/10

   Brex Bill Pay covers invoice capture (AI extracts itemized detail and matches PO), configurable multi-level approval routing by amount/vendor/role, vendor onboarding via secure payment-detail links, and payment via virtual card, ACH, checks, or wires from a Brex or external bank account — all in one platform, backed by both marketing docs and API endpoints (vendor creation, payments API). missing for 10: independent/hands-on verification of AP workflow reliability (only vendor docs, no third-party case study or user report specific to bill pay), and detail on check-payment mechanics/timing.

   • [claimed-docs] “Tailor multi-level approval flows based on amount, vendor, employee role, and more — and Brex will route for you automatically to ensure the…”
   • [claimed-docs] “Enter your vendor's email address and Brex will send a secure link where they can upload their payment details.”
   • [claimed-docs] “Brex AI will capture every itemized detail, suggest the matching PO, and draft a payment.”
   • [claimed-docs] “Create a virtual card instantly to pay by card and earn rewards. Send ACH, checks, and wires from any bank account or a Brex business accoun…”
   • [claimed-docs] “This endpoint creates a new vendor.”
   • [claimed-docs] “Eliminate data entry, automate approvals, and spin up virtual cards in one click to maximize your rewards.”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence covers Navan's travel and expense management (card transactions, reimbursements, receipts, policy approvals) but contains no mention of invoice capture, vendor bill management, or paying vendors via ACH, check, or wire — the core AP capabilities the story requires.

   ⚑ flag

Card controls

1. finance leadRestrict cards by category, merchant, and amount — and the platform auto-locks or declines out-of-policy spend at the point of sale

   weight 3 · round to Brex
   Brex✓fullclaimed7/10

   Brex's product docs explicitly describe auto-blocking out-of-policy purchases with preset rules and limits, auto-enforced spend controls, and a Budgets/Spend Limits API for programmatic control, matching the finance-lead story of restricting and declining spend by policy. Missing for 10: explicit documentation of category/merchant-level rule granularity and independent/hands-on confirmation that point-of-sale declines work as described in practice.

   • [claimed-docs] “Automatically block out-of-policy purchases, with preset rules and limits.”
   • [claimed-docs] “Create limits for stipends, travel, procurement, and more. Auto-enforced controls empower employees to spend with confidence.”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence shows general card management (active/inactive status), policy workflows, and MCP queries about flagged/violating transactions, but nothing documents configurable restrictions by category, merchant, or amount, nor automatic point-of-sale locking/declining of out-of-policy spend. missing for 10: explicit documentation of category/merchant/amount-based card controls, missing for 10: evidence of real-time auto-lock or point-of-sale decline enforcement.

   • [claimed-docs] “Card management — see which cards are active or inactive, how many cards a user has, and other card details.”
   • [claimed-docs] “Connect existing corporate or business credit cards”
   • [claimed-docs] “Unlimited policy and approval workflows”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”

   proof ↗⚑ flag

Card issuance

1. founderIssue physical and virtual corporate cards to the team in minutes, each with its own spend limit

   weight 3 · round to Brex
   Brex✓fullclaimed8/10

   Brex documents physical and virtual card issuance on Mastercard with instant/one-click virtual card creation, per-card/employee spend limits and budgets, mobile app management, and bulk issuance via Workato/Pipedream integrations for teams. This directly matches the founder story of quickly issuing cards with individual limits. Missing for 10: independent hands-on timing verification ("minutes") and no first-party case study explicitly confirming founder-level self-service issuance speed.

   • [claimed-docs] “Physical and virtual cards issued on the leading worldwide network, Mastercard, with acceptance in 210+ countries and territories.”
   • [claimed-docs] “Get a card limit that keeps pace with your business based on financial factors such as revenue or dollars raised.”
   • [claimed-docs] “Manage your card from your phone. Pay with Apple Pay, cancel your card, and view your spend all in one app.”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “Create limits for stipends, travel, procurement, and more. Auto-enforced controls empower employees to spend with confidence.”
   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [claimed-docs] “Eliminate data entry, automate approvals, and spin up virtual cards in one click to maximize your rewards.”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence only shows querying/viewing existing card status via MCP (navan-docs-7) and connecting existing cards (navan-docs-9), but nothing about issuing new physical or virtual cards, setting per-card spend limits, or speed of issuance ('minutes'). Missing for 10: any documentation of card issuance workflow, virtual card creation, spend-limit configuration, or time-to-issue claims.

   • [claimed-docs] “Card management — see which cards are active or inactive, how many cards a user has, and other card details.”
   • [claimed-docs] “Connect existing corporate or business credit cards”

   proof ↗⚑ flag

Virtual cards

1. finance leadCreate single-vendor virtual cards for SaaS subscriptions and procurement so one compromised vendor never exposes shared credit

   weight 2 · round to Brex
   Brex✓fullclaimed7/10

   Brex documents virtual card creation (docs-14, docs-38), bulk virtual-card issuance for employees/vendors (docs-30, docs-32), and procurement-specific spend limits (docs-9), which together support finance leads creating vendor-specific virtual cards to isolate compromise risk. Missing for 10: explicit documentation of a 'lock card to single vendor' control or vendor-merchant-locking mechanism, and independent/hands-on evidence confirming this containment behavior in practice.

   • [claimed-docs] “Create a virtual card instantly to pay by card and earn rewards. Send ACH, checks, and wires from any bank account or a Brex business accoun…”
   • [claimed-docs] “Eliminate data entry, automate approvals, and spin up virtual cards in one click to maximize your rewards.”
   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [claimed-docs] “Create limits for stipends, travel, procurement, and more. Auto-enforced controls empower employees to spend with confidence.”
   • [claimed-docs] “Physical and virtual cards issued on the leading worldwide network, Mastercard, with acceptance in 210+ countries and territories.”

   proof ↗⚑ flag

   Navan—none0/10

   The evidence pack covers Navan's MCP/API integrations, travel booking, rewards, and pricing highlights, but contains no mention of virtual card creation, single-vendor card controls, or SaaS/procurement card issuance. Card-related evidence only covers connecting existing corporate cards and viewing/managing card transaction data, not creating new single-vendor virtual cards.

   • [claimed-docs] “Card management — see which cards are active or inactive, how many cards a user has, and other card details.”
   • [claimed-docs] “Connect existing corporate or business credit cards”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”

   proof ↗⚑ flag

Agent access

1. ai-native userAn agent can pull uncoded transactions via API or MCP, propose categorizations and policy flags, and push clean coding back for review

   weight 3 · round to Brex
   Brex~partialprobed6/10

   Brex offers a Transactions API to pull transaction data, an official MCP server that lets AI assistants 'manage expenses' and 'review banking transactions,' and AI features (audit/review agents) that categorize spend and flag policy violations (policy field on bookings, out-of-policy detection). However, the evidence never shows a specific 'uncoded transactions' filter/endpoint or an explicit write-back/update-category API call that pushes AI-proposed coding for human review — it only shows general expense/transaction read access and rule-based auto-approval. Missing for 10: explicit uncoded-transaction query capability, explicit category/coding update endpoint, and documented human-review workflow tying AI-proposed codes back to approval.

   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “Added `policy` to booking objects, exposing whether a booking was out of policy and which travel policy rules it violated.”
   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [probe] “official MCP server documented at https://developer.brex.com/docs/mcp”

   proof ↗⚑ flag

   Navan~partialclaimed5/10

   The API exposes retrieval and update endpoints for card/Connect-card transactions and custom-field management (navan-docs-2, navan-docs-4, navan-docs-23), and MCP lets agents query uncoded/flagged transactions and policy violations in natural language (navan-docs-5, navan-docs-6, navan-docs-21, navan-docs-22). However, MCP is documented primarily as a query/analysis interface, not a write-back tool, and there is no explicit documentation of an agent proposing categorizations/policy flags and pushing clean coding back for human review as a workflow. Missing for 10: explicit MCP write/update support for coding transactions, a documented propose-then-review workflow, and independent confirmation of this end-to-end loop.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “What's our company policy on rideshare, and which transactions from last month violate it?”
   • [claimed-docs] “Analyze spend data — query expenses, identify trends, and break down costs by department, vendor, or category.”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”

   proof ↗⚑ flag

Api access

1. developerPull transactions, expenses, and receipts through a documented REST API with OAuth and scoped tokens

   weight 3 · round to Navan
   Brex~partialprobed7/10

   Brex has a documented REST API covering transactions, expenses, receipts (upload/match), budgets, and webhooks, with token-based authentication using scopes (brex-docs-2, brex-docs-4, brex-docs-7, brex-docs-27, brex-docs-19). However, the auth mechanism is described as dashboard-generated 'user tokens' with scopes rather than a full OAuth flow, and an OpenAPI spec could not be located via standard endpoints (brex-probe-2). missing for 10: explicit OAuth 2.0 flow documentation (vs. static token generation), publicly discoverable OpenAPI/swagger spec, and independent developer corroboration of the API working as documented.

   • [claimed-docs] “Sign in to dashboard.brex.com as an account admin or card admin... Click Create Token. Create a name for your token... Choose what level of …”
   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “Upload receipt and match automatically.”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “generate a user token from your Brex dashboard and pass it along in your API call headers”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.brex.com/openapi.json, https://developer.brex.com/swagger.json, https://developer.…”

   proof ↗⚑ flag

   Navan✓fullclaimed8/10

   Navan's developer portal documents a REST API with OAuth 2.0 client-credentials flow, and endpoints covering transactions, expenses, receipts, custom fields, and webhooks. Missing for 10: explicit mention of scoped/permissioned tokens (vs just client-credentials) and independent third-party corroboration of the API's documented behavior beyond vendor docs.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”
   • [claimed-docs] “Receipt scanning for on-the-go expenses”

   proof ↗⚑ flag

Card api

1. developerIssue and manage cards programmatically — create a card with a limit, lock it, update controls — via the public API

   weight 2 · round to Brex
   Brex~partialclaimed4/10

   Evidence shows Brex exposes a broad API surface with a Postman workspace for 'all Brex APIs' and third-party integration recipes (Pipedream, Workato) that issue virtual cards and set spend limits/monthly limits, plus a budgets/spend-limits API. However, there is no first-party documentation of a dedicated Cards API (create/lock/update card controls) in the evidence pack — no cards_api.md or explicit lock/update-controls endpoints are cited, only inferred usage via third-party automation platforms. Missing for 10: a documented Cards API reference showing create-with-limit, lock/freeze, and update-controls endpoints, and any independent hands-on confirmation of programmatic card lifecycle management.

   • [claimed-docs] “we’ll show you how to bulk create virtual cards for your employees”
   • [claimed-docs] “Workato...to enable various use cases including bulk creating users, setting monthly limits for users, issuing virtual cards, and sending ve…”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “Create limits for stipends, travel, procurement, and more. Auto-enforced controls empower employees to spend with confidence.”
   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”

   proof ↗⚑ flag

   Navan—none0/10

   The API docs mention retrieving/updating card transactions and MCP tools let you view card status (active/inactive, count), but there is no evidence of creating a new card, setting a spending limit, locking a card, or updating card controls programmatically via the public API — missing for 10: card creation endpoint, limit-setting endpoint, lock/freeze endpoint, control-update endpoint documentation.

   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Card management — see which cards are active or inactive, how many cards a user has, and other card details.”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

1. ai-native userDo everything through the API that I can do in the UI

   weight 2 · round drawn
   Brex~partialprobed5/10

   Brex exposes a broad public API (transactions, budgets, expenses, users, webhooks) and an official MCP server for AI-assistant access, covering many core UI workflows, but there is no evidence of full parity — e.g. no documented API for bill pay approval routing, travel policy management, or card issuance settings that mirror the UI feature set. missing for 10: explicit parity statement, bill-pay/travel/card-management API endpoints, independent confirmation that all UI actions are API-reachable.

   • [claimed-docs] “Sign in to dashboard.brex.com as an account admin or card admin... Click Create Token. Create a name for your token... Choose what level of …”
   • [claimed-docs] “The Brex MCP server lets AI assistants interact directly with your Brex account: manage expenses, look up users, check cards, review banking…”
   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “Tailor multi-level approval flows based on amount, vendor, employee role, and more — and Brex will route for you automatically to ensure the…”
   • [claimed-docs] “Create a virtual card instantly to pay by card and earn rewards. Send ACH, checks, and wires from any bank account or a Brex business accoun…”
   • [probe] “PROBE llms.txt: HTTP 200 at https://developer.brex.com/llms.txt # Brex Developer API > APIs to build on top of the Brex financial OS. ## T…”
   • [probe] “PROBE openapi: all candidate paths 404 (https://developer.brex.com/openapi.json, https://developer.brex.com/swagger.json, https://developer.…”

   proof ↗⚑ flag

   Navan~partialprobed5/10

   Navan's API/MCP covers expense retrieval/update, transactions, custom fields, receipts, webhooks, and read-style spend/policy queries, but there is no evidence of API parity for travel booking, rewards, or other core UI workflows (e.g., booking trips, managing 'price to beat', Navan Edge concierge features) — those remain UI/chat-only per docs. missing for 10: API endpoints for travel booking/itinerary management, evidence of full UI-to-API parity beyond expense/finance domain, independent confirmation of completeness.

   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”
   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Discover company fields, batch-manage option values, and poll async job status.”
   • [claimed-docs] “Global travel inventory and exclusive rates”
   • [claimed-docs] “Navan Edge provides 24/7 access to human travel agents via chat or phone at +1-877-536-3343 in English.”
   • [probe] “official MCP server documented at https://developer.navan.com/mcp/”

   proof ↗⚑ flag

2. ai-native userExport all of my data in open formats and leave

   weight 3 · round to Brex
   Brex~partialclaimed4/10

   Brex exposes APIs for transactions, expenses, users, budgets, etc. and a documented Zapier example for CSV export of card transactions, which lets a user pull data out in open formats; but there is no first-party documentation of a comprehensive 'export all your data' feature, bulk archive export, or account-closure data-portability process. missing for 10: a dedicated full-account data export tool, documentation of exporting all data types (not just transactions) in one open format, and any offboarding/portability guidance for leaving the platform.

   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “Monthly Export of Brex Card Transactions to CSV in Google Drive”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “go to the Brex developer workspace which has collections for all of the Brex APIs”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence shows API/MCP access for querying and updating expense/travel data, but nothing about bulk data export in open/portable formats or any documented account-closure/data-portability process for leaving the platform. Missing for 10: bulk export functionality, open-format data dumps, documented data portability/exit process.

   ⚑ flag

Approvals

1. finance leadBuild multi-step approval chains — manager, budget owner, finance — with delegation and escalation when approvers sit on requests

   weight 2 · round to Brex
   Brex~partialclaimed5/10

   Brex documents multi-level, rule-based approval routing (by amount, vendor, employee role) and an AI review agent that auto-approves low-risk items and escalates exceptions, which covers the core approval-chain and escalation concept. However there is no explicit documentation of delegation (e.g., proxy approver when someone is OOO) or a named manager→budget-owner→finance sequential chain structure. Missing for 10: explicit delegation/out-of-office reassignment mechanism, documented sequential role hierarchy (manager, budget owner, finance) rather than generic 'role-based routing', and independent/hands-on confirmation of escalation behavior.

   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Tailor multi-level approval flows based on amount, vendor, employee role, and more — and Brex will route for you automatically to ensure the…”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “Assign top-line budgets, and track spend instantly, ensuring every dollar is allocated where it has the biggest impact.”

   proof ↗⚑ flag

   Navan~partialclaimed2/10

   Navan mentions 'Unlimited policy and approval workflows' as a pricing feature and MCP queries reference 'who needs to approve' a transaction, confirming approval workflows exist, but there is no evidence describing configurable multi-step chains (manager → budget owner → finance), delegation, or escalation logic when approvers sit idle. missing for 10: documentation of multi-step approval chain configuration, delegation rules, escalation timers/triggers, and named approver roles.

   • [claimed-docs] “Unlimited policy and approval workflows”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”

   proof ↗⚑ flag

Audit trail

1. finance leadEvery expense carries a full audit trail — edits, approvals, policy checks — that survives an external audit

   weight 2 · round to Brex
   Brex~partialclaimed6/10

   Brex documents policy-violation flags on bookings, multi-level approval routing, budget/spend limits, and a dedicated 'audit agent' that monitors and categorizes policy violations — all elements that would feed an audit trail. However, there is no explicit documentation of an immutable, exportable audit log capturing edit history, approval timestamps, and policy-check results in a form built for external audit review. missing for 10: explicit immutable/exportable audit-log documentation, evidence of edit-history tracking per expense, independent confirmation the trail survives external audit scrutiny.

   • [claimed-docs] “Added `policy` to booking objects, exposing whether a booking was out of policy and which travel policy rules it violated.”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Tailor multi-level approval flows based on amount, vendor, employee role, and more — and Brex will route for you automatically to ensure the…”
   • [claimed-docs] “Your audit agent monitors company expenses according to your internal processes and categorizes potential violations by risk level.”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Automatically block out-of-policy purchases, with preset rules and limits.”

   proof ↗⚑ flag

   Navan~partialclaimed4/10

   Navan documents policy/approval workflows and the ability to query why a transaction was flagged or who must approve it, which implies some traceability, but there is no explicit documentation of an immutable audit log, edit history, or audit-survivability guarantees. Missing for 10: explicit audit-trail/version-history documentation, evidence of edit logging, and any mention of external-audit compliance or SOC/ISO controls.

   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “Unlimited policy and approval workflows”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

Policy engine

1. finance leadCodify our expense policy — limits by category, role, and context — so in-policy expenses auto-approve and only exceptions reach a human

   weight 3 · round to Brex
   Brex✓fullclaimed8/10

   Brex documents policy codification by category, role, vendor, and context (limits, spend rules, multi-level approval routing by amount/vendor/role) and explicit auto-approval of in-policy items with exceptions escalated to humans via its review/audit AI agents. Budgets API and policy-violation flags further support programmatic enforcement. Missing for 10: independent/hands-on verification of auto-approval accuracy and no detailed docs on granular category-level rule configuration UI.

   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Create limits for stipends, travel, procurement, and more. Auto-enforced controls empower employees to spend with confidence.”
   • [claimed-docs] “Tailor multi-level approval flows based on amount, vendor, employee role, and more — and Brex will route for you automatically to ensure the…”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”
   • [claimed-docs] “Automatically block out-of-policy purchases, with preset rules and limits.”
   • [claimed-docs] “Stop waiting for people to categorize, review, or approve spend. Let AI agents do it instantly, based on your policies and with total transp…”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “Added `policy` to booking objects, exposing whether a booking was out of policy and which travel policy rules it violated.”

   proof ↗⚑ flag

   Navan~partialclaimed5/10

   Navan advertises "unlimited policy and approval workflows" and its MCP/AI layer can explain which policy applies, why a transaction was flagged, and who must approve it, implying a rules engine with per-transaction policy limits (e.g., rideshare, hotel price caps). However, the evidence never explicitly describes configuring limits by category/role/context or confirms that in-policy spend auto-approves while only exceptions route to a human. Missing for 10: explicit documentation of policy-builder granularity (category/role/context), explicit auto-approval behavior for in-policy expenses, and independent/hands-on confirmation of the approval-routing logic.

   • [claimed-docs] “Unlimited policy and approval workflows”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “What's our company policy on rideshare, and which transactions from last month violate it?”
   • [claimed-docs] “Navan shows employees a "price to beat" for every hotel search — a smart target based on location, date, and demand that's under their poli…”
   • [claimed-docs] “Navan shows employees a "price to beat" for every hotel search — a smart target based on location, date, and demand that's under their polic…”

   proof ↗⚑ flag

1. ai-native userChoose where my data is stored (region/residency)

   weight 2 · round drawn
   Brex—none0/10

   No evidence pack items mention data residency, regional storage options, or any control over where data is stored; Brex evidence covers financial product features, APIs, and MCP integration only.

   ⚑ flag

   Navan—none0/10

   No evidence anywhere in the pack about data residency, region selection, or data storage location controls for Navan's data or its AI/MCP features; the pack only covers MCP capabilities, API endpoints, pricing features, and travel booking experience.

   ⚑ flag

2. ai-native userPrevent my data from being used to train AI models

   weight 3 · round drawn
   Brex—none0/10

   No evidence pack item addresses AI training data usage, opt-out settings, or a data-privacy policy regarding model training; the evidence only covers API/product features and unrelated community complaints about account closures.

   ⚑ flag

   Navan—none0/10

   No evidence in the pack addresses AI training data usage, opt-out controls, or data privacy policy regarding model training; all evidence covers MCP integrations, expense/travel features, and unrelated community complaints. Missing for 10: any explicit statement on AI/model training data usage, opt-out settings, or data retention/privacy commitments.

   ⚑ flag

3. ai-native userControl data retention and deletion

   weight 2 · round drawn
   Brex—none0/10

   No evidence pack item discusses data retention, deletion policies, or AI data usage/training controls specific to AI-native workflows; the API, MCP, and product docs focus on functionality (expenses, cards, bill pay) rather than privacy/data lifecycle controls. Missing for 10: any documentation on data retention periods, deletion requests, opt-out of AI training, or data residency/handling policies.

   ⚑ flag

   Navan—none0/10

   No evidence in the pack addresses data retention policies, deletion controls, or user/admin ability to manage how long AI/assistant data is stored or how to delete it — the pack covers MCP querying, API surfaces, pricing features, and travel booking, but nothing on retention/deletion controls.

   ⚑ flag

4. ai-native userOpt out of telemetry and usage tracking

   weight 2 · round drawn
   Brex—none0/10

   No evidence in the pack mentions telemetry, usage tracking, or an opt-out/privacy control mechanism for AI-native users; the docs focus on APIs, webhooks, MCP server, and financial products with no privacy-posture disclosures.

   ⚑ flag

   Navan—none0/10

   No evidence in the pack addresses telemetry/usage-tracking opt-out settings for Navan or its AI/MCP features; all evidence covers MCP data access, expense/travel features, and pricing.

   ⚑ flag

Mileage perdiem

1. employeeLog mileage with map-based distance and claim per-diem rates without building the expense by hand

   weight 2 · round drawn
   Brex—none0/10

   No evidence in the pack mentions mileage tracking, GPS/map-based distance calculation, or per-diem rate claims; Brex's expense documentation focuses on card transactions, receipt upload/matching, and policy enforcement but never describes mileage or per-diem automation. This is a fair capability to expect from an expense management product, so absence of evidence means none rather than na.

   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Upload receipt and match automatically.”
   • [claimed-docs] “Manage global spend and close the books in real time — with unmatched automation of manual expense work.”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence covers Navan's MCP/API integrations, travel booking rewards, and generic 'receipt scanning' and 'manage expenses' pricing bullets, but nothing describes mileage tracking, map-based distance calculation, or automated per-diem rate application. Missing for 10: mileage logging with GPS/map distance capture, per-diem rate lookup/auto-application, any auto-build-expense-from-mileage workflow.

   ⚑ flag

Receipt capture

1. employeeSnap a photo or forward an email and the platform OCRs the receipt and files it against the right transaction

   weight 3 · round to Brex
   Brex~partialclaimed6/10

   Brex supports receipt upload with automatic matching to expenses via API ("Upload receipt and match automatically", pre-signed S3 URL matching to existing expenses) and AI agents that 'fetch receipts' automatically, but there's no explicit evidence of OCR technology, photo-snap mobile capture flow, or email-forwarding-to-receipt-filing as described in the story. missing for 10: explicit OCR mention, mobile photo-capture UX details, email-forwarding ingestion mechanism, independent/hands-on confirmation of accuracy.

   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “Upload receipt and match automatically.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”

   proof ↗⚑ flag

   Navan~partialclaimed5/10

   Docs mention 'Receipt scanning for on-the-go expenses' as a pricing-page feature bullet and API docs mention a 'receipts' surface, implying photo capture and filing against transactions, but there's no detail on OCR accuracy, email-forwarding capture, or automatic matching logic. Missing for 10: technical documentation of the OCR/email-forwarding pipeline, evidence of automatic transaction matching, and independent/hands-on confirmation of accuracy.

   • [claimed-docs] “Receipt scanning for on-the-go expenses”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

Receipt matching

1. employeeReceipts match to card transactions automatically — with e-receipts pulled from integrations — and I get nudged only when one is genuinely missing

   weight 2 · round to Brex
   Brex~partialclaimed6/10

   Brex's API/docs confirm automatic receipt-to-transaction matching ("Upload receipt and match automatically" and pre-signed S3 upload that tries to match existing expenses), and its AI feature page claims agents can 'fetch receipts' and file expenses automatically, plus a review agent that 'follows up with anyone out of compliance.' However, there's no explicit evidence of e-receipts being pulled automatically from third-party integrations (e.g., travel/vendor e-receipt feeds) nor documentation describing smart nudging logic that only triggers when a receipt is genuinely missing rather than generic compliance follow-ups. missing for 10: explicit e-receipt integration sourcing, documented logic for suppressing false-positive missing-receipt nudges, independent user confirmation of matching accuracy.

   • [claimed-docs] “Upload receipt and match automatically.”
   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Your review agent auto-approves low-risk expenses and requests, escalates exceptions, and follows up with anyone out of compliance.”

   proof ↗⚑ flag

   Navan—none0/10

   Evidence only mentions generic 'receipt scanning for on-the-go expenses' and receipts as an API surface, but nothing describes automatic matching of receipts to card transactions, e-receipt pull from integrations, or smart nudging only when genuinely missing.

   • [claimed-docs] “Receipt scanning for on-the-go expenses”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

Domestic

1. employeeSubmit an out-of-pocket expense and get reimbursed by direct deposit in days, tracked from submission to payout

   weight 3 · round to Brex
   Brex~partialclaimed5/10

   Brex's docs confirm out-of-pocket expense submission (receipt upload/matching in brex-docs-7/27) and AI-driven automatic reimbursement filing (brex-docs-23), implying an end-to-end expense workflow, but there is no explicit documentation of direct-deposit payout mechanics, reimbursement timing ('days'), or a submission-to-payout tracking status view. missing for 10: explicit direct-deposit reimbursement flow, payout SLA/timing details, and a visible tracking/status dashboard for reimbursement progress.

   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “AI assistants do employees' expenses for them, writing memos, fetching receipts, offering policy help, even filing reimbursements automatica…”
   • [claimed-docs] “Upload receipt and match automatically.”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”

   proof ↗⚑ flag

   Navan~partialclaimed3/10

   Navan pricing page mentions it can 'manage expenses and issue reimbursements' and includes receipt scanning, but there is no evidence describing direct-deposit payout, a specific reimbursement timeline (days), or end-to-end tracking from submission to payout. missing for 10: direct deposit mechanism, reimbursement speed/SLA, submission-to-payout tracking/status visibility.

   • [claimed-docs] “Manage expenses and issue reimbursements”
   • [claimed-docs] “Receipt scanning for on-the-go expenses”
   • [claimed-docs] “Unlimited policy and approval workflows”

   proof ↗⚑ flag

International

1. finance leadReimburse employees abroad in their local currency without running a separate international payments process

   weight 2 · round drawn
   Brex—none0/10

   Evidence covers global card acceptance (210+ countries), bill pay wires 'in more currencies' for vendor payments, and general 'global spend' messaging, but nothing specifically addresses reimbursing employees abroad in their local currency or eliminating a separate international payments process for reimbursements.

   ⚑ flag

   Navan—none0/10

   Evidence only shows generic reimbursement management (docs-8) with no mention of local-currency payouts, multi-currency support, or eliminating a separate international payment process for reimbursing employees abroad.

   • [claimed-docs] “Manage expenses and issue reimbursements”

   proof ↗⚑ flag

Budgets

1. finance leadAllocate budgets to teams and projects and have card limits and approvals actually enforce them

   weight 2 · round to Brex
   Brex✓fullclaimed8/10

   Brex documents budgets/spend limits API and dashboard features for assigning top-line budgets, auto-enforced spend limits, policy-based blocking of out-of-policy purchases, and multi-level approval routing, directly matching the finance-lead's need to allocate budgets and have limits/approvals enforce them. missing for 10: independent/hands-on evidence confirming enforcement works reliably in practice (community evidence only covers account closures, not budget/approval enforcement), and no case study showing team/project-level allocation granularity in real use.

   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “Set rules for how employees can spend, what they need to document, and who needs to approve. Brex AI applies them to every transaction.”
   • [claimed-docs] “Create limits for stipends, travel, procurement, and more. Auto-enforced controls empower employees to spend with confidence.”
   • [claimed-docs] “Assign top-line budgets, and track spend instantly, ensuring every dollar is allocated where it has the biggest impact.”
   • [claimed-docs] “Tailor multi-level approval flows based on amount, vendor, employee role, and more — and Brex will route for you automatically to ensure the…”
   • [claimed-docs] “Automatically block out-of-policy purchases, with preset rules and limits.”

   proof ↗⚑ flag

   Navan~partialclaimed4/10

   Navan documents general 'unlimited policy and approval workflows' and card management/policy-limit features, and MCP queries can surface policy violations and approval status, implying some enforcement infrastructure exists. However, there is no direct evidence of budget allocation to specific teams/projects or of card limits being tied to those budgets and automatically enforced. Missing for 10: explicit budget-allocation-to-team/project feature docs, evidence that card limits are dynamically tied to allocated budgets, and any hands-on confirmation that approvals actually block over-budget spend.

   • [claimed-docs] “Unlimited policy and approval workflows”
   • [claimed-docs] “Card management — see which cards are active or inactive, how many cards a user has, and other card details.”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”

   proof ↗⚑ flag

Dashboards

1. founderSee company spend in real time — by team, category, and merchant — the day it happens, not when statements land

   weight 3 · round drawn
   Brex✓fullclaimed7/10

   Brex provides real-time spend tracking via the mobile app ("view your spend all in one app"), budgets API tracked instantly by team/cost-center, transactions API for granular transaction-level data, and webhooks for live event notifications, plus marketing claims of managing spend and closing books in real time. Category/department/merchant segmentation is supported via cost_center_id/department_id filters and the transactions API, though merchant-level breakdown isn't explicitly documented as a dashboard feature. Missing for 10: explicit merchant-level dashboard breakdown documentation, independent/hands-on verification that spend updates instantly rather than with delay, and no clear evidence of a dedicated real-time analytics dashboard beyond mobile app spend view.

   • [claimed-docs] “The transactions API lets you view your transactions, accounts, and statements.”
   • [claimed-docs] “The budgets API lets you manage your Brex Budgets and Spend Limits.”
   • [claimed-docs] “Assign top-line budgets, and track spend instantly, ensuring every dollar is allocated where it has the biggest impact.”
   • [claimed-docs] “Manage your card from your phone. Pay with Apple Pay, cancel your card, and view your spend all in one app.”
   • [claimed-docs] “Manage global spend and close the books in real time — with unmatched automation of manual expense work.”
   • [claimed-docs] “Added filter query parameters to List users: `name`, `email[]`, `status[]`, `manager_id[]`, `department_id[]`, `location_id[]`, `title_id[]`…”
   • [claimed-docs] “Register an endpoint to start receiving selected webhook events”

   proof ↗⚑ flag

   Navan✓fullclaimed7/10

   Navan's docs show spend can be queried by department, vendor/merchant, and category via MCP/API, and card transactions (including Navan-issued cards) are retrievable directly rather than waiting for statements, supporting near real-time visibility for founders. Missing for 10: explicit same-day latency claims for card feeds, a dedicated real-time dashboard description, and independent/hands-on corroboration of update speed.

   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Analyze spend data — query expenses, identify trends, and break down costs by department, vendor, or category.”
   • [claimed-docs] “Show me every flagged expense over $500 from Q1 that hasn't been approved yet, grouped by department.”
   • [claimed-docs] “Ask questions about policies, approval flows, and flag/decline reasons — understand why a transaction was flagged, who needs to approve it, …”
   • [claimed-docs] “OAuth 2.0 client-credentials flow, retrieval and update endpoints across the expense surface, custom-field management, receipts, and webhook…”

   proof ↗⚑ flag

Booking

1. employeeBook flights and hotels in-platform with travel policy applied at booking time, not expensed and argued about later

   weight 2 · round drawn
   Brex✓fullcommunity8/10

   Brex Travel is a first-party product explicitly combining booking with 'global inventory and auto-enforced expense policies on one platform' (brex-docs-39), and the API changelog shows booking objects carry a `policy` field flagging out-of-policy violations at booking time (brex-docs-6), confirming policy is applied during booking rather than after. Community evidence corroborates actual usage of Brex Travel for booking ('Brex Travel was luxurious') (brex-comm-1). Missing for 10: independent review of the booking flow's policy enforcement UX and more detail on flight/hotel inventory breadth.

   • [claimed-docs] “Simplify travel with global inventory and auto-enforced expense policies on one platform.”
   • [claimed-docs] “Added `policy` to booking objects, exposing whether a booking was out of policy and which travel policy rules it violated.”
   • [community] “I really liked the dynamic credit card generator, and Brex Travel was luxurious...”

   proof ↗⚑ flag

   Navan✓fullclaimed8/10

   Navan's core booking product shows in-platform flight/hotel booking with policy limits enforced at search time ('price to beat' under policy, docs-12/27), self-serve booking with 24/7 support (docs-11), human travel agents that must confirm before booking (docs-14-16), and unlimited configurable policy/approval workflows applied to travel (docs-25) plus connected corporate cards for direct billing (docs-9) rather than post-hoc expensing. Missing for 10: independent/hands-on confirmation that out-of-policy bookings are actually blocked or flagged in real time (only vendor marketing/docs cited), and no detail on how policy violations are handled during the booking flow itself.

   • [claimed-docs] “Global travel inventory and exclusive rates”
   • [claimed-docs] “Self-serve changes and 24/7 travel support agents”
   • [claimed-docs] “Navan shows employees a "price to beat" for every hotel search — a smart target based on location, date, and demand that's under their poli…”
   • [claimed-docs] “Navan Edge provides 24/7 access to human travel agents via chat or phone at +1-877-536-3343 in English.”
   • [claimed-docs] “Navan Edge will always ask for your explicit confirmation before booking, changing, or canceling any trip.”
   • [claimed-docs] “Unlimited policy and approval workflows”
   • [claimed-docs] “Connect existing corporate or business credit cards”
   • [claimed-docs] “Navan shows employees a "price to beat" for every hotel search — a smart target based on location, date, and demand that's under their polic…”

   proof ↗⚑ flag

Trip expenses

1. employeeA trip's expenses — bookings, cards, receipts — collect themselves into one itinerary-linked report

   weight 1 · round to Navan
   Brex~partialclaimed5/10

   Brex Travel and expense docs show pieces of this workflow — bookings carry policy metadata (brex-docs-6), corporate cards integrate with the platform (brex-docs-20/21), and receipts auto-match to expenses (brex-docs-7/brex-docs-27) — and the Travel product page claims to "simplify travel with global inventory and auto-enforced expense policies on one platform" (brex-docs-39). However, no evidence explicitly describes a trip/itinerary object that automatically aggregates bookings, card transactions, and receipts into a single itinerary-linked expense report for the employee. Missing for 10: explicit itinerary/trip-report generation feature, documentation of automatic booking-to-card-to-receipt linkage within one report, and any hands-on/independent confirmation of this consolidated view.

   • [claimed-docs] “Simplify travel with global inventory and auto-enforced expense policies on one platform.”
   • [claimed-docs] “Added `policy` to booking objects, exposing whether a booking was out of policy and which travel policy rules it violated.”
   • [claimed-docs] “The uri will be a pre-signed S3 URL allowing you to upload the receipt securely... we will try to match the receipt with existing expenses.”
   • [claimed-docs] “Upload receipt and match automatically.”
   • [claimed-docs] “Physical and virtual cards issued on the leading worldwide network, Mastercard, with acceptance in 210+ countries and territories.”

   proof ↗⚑ flag

   Navan~partialclaimed6/10

   Navan's own docs describe unified handling of card transactions, manual expenses, receipts, and reimbursements within a single platform/API (navan-docs-4, navan-docs-8, navan-docs-9, navan-docs-24), implying expenses converge into one system, but no evidence explicitly confirms automatic itinerary-linked report assembly from bookings+cards+receipts without manual reconciliation steps. Missing for 10: explicit documentation of automatic itinerary-to-expense-report linkage, and independent/hands-on confirmation that receipts and card charges auto-match to trip itineraries without manual effort.

   • [claimed-docs] “Retrieve and update Navan card transactions, Connect card transactions, manual expenses, payroll, repayments, fees, adjustments, daily rebat…”
   • [claimed-docs] “Manage expenses and issue reimbursements”
   • [claimed-docs] “Connect existing corporate or business credit cards”
   • [claimed-docs] “Receipt scanning for on-the-go expenses”
   • [claimed-docs] “Unlimited policy and approval workflows”

   proof ↗⚑ flag