Internal Agent Platform

Brex built a Retool-based platform where operations staff build, test and deploy agents for KYC, disputes, quality assurance and collections. Its systems engineering team maintains the builder; human involvement differs by workflow.

Company: Brex
Approach type: Platform
Work supported: Finance ops, Support, Customer success
Interfaces: Slack, Internal ui
Deployment stage: Scaled
Evidence strength: Secondary only
Entry reviewed: 2026-09-16

Copy as MD Ask ChatGPT

Capabilities and architecture

Model

Not reportedDeployed-agent models are not documented; the platform provides multi-model testing and evaluation at build time.

Harness

Retool-based builder with prompt management and multi-model testing/evaluation; built by a ~25-person systems-engineering team^{[Source 1]}

Sandbox

Retool-hosted runtime (no bespoke execution env described)^{[Source 1]}Retool hosts the runtime; the capture describes no further isolation boundary.

Tool access

An MCP server exposes product capabilities to internal agents. Separately, Slack /c1 requests access to AI tools through ConductorOne; it is not evidence of invoking an operations agent.^{[Source 1]}The MCP bridge and immediate internal availability are supported at line 40; the Slack /c1 clause describes provisioning configuration, not agent invocation (flagged).

Knowledge

Standard operating procedures uploaded as a knowledge base, such as 100-page dispute guides; customer account data^{[Source 1]}

Context management

Not reportedNo mechanism for maintaining agent context over time is documented; the QA feedback loop improves performance rather than managing context.

Credentials

SSO via internal Retool proxies (no per-user accounts); ConductorOne access management; Okta auth; data classified by risk; ≤30-day retention, no training on inputs^{[Source 1]}SSO via Retool proxies, Okta provisioning, and the retention and segregation controls are supported at lines 60–62; the ConductorOne vendor name is not in the capture (flagged).

Interfaces

slack
internal-ui

^{[Source 1]}

Mechanisms

Prompt + eval studio

Non-technical ops staff design prompts, test across models, deploy with QA oversight^{[Source 1]}

MCP tool bridge

Product features exposed to internal agents through one MCP server^{[Source 1]}

Documented uses

Two documented agent runs on the platform: dispute-submission preparation, and collections response drafting for the servicing team.

Prepare the dispute submission

The dispute agent applies the 100+ page dispute guide and the BPO operating procedure to produce a more robust, better-articulated submission^{[Source 1]}

Draft the collections response

The agent analyzes the customer reply and layers in account context and negotiating guidelines, offering the servicing team drafts in several tones to select, customize, or send^{[Source 1]}

Access and controls

See documented credential and access boundaries in the architecture above. Scoped human-review assessments for individual uses remain in research details.

Reliability and validation

Apply the quality rubric

A QA agent reviews every support interaction against the quality rubric and feeds response trends into a closed loop improving agent and human performance^{[Source 1]}

Adoption and operating evidence

These results come from the sources. Notes below each result explain what it covers and any reported limits.

Headline claim

Dispute-agent use example: Dispute processing time fell from three hours to three seconds^{[Source 1]}

Reported by: Brex
Scope: Dispute-submission preparation using the internal agent platform; not end-to-end chargeback resolution

The source does not report the denominator of this figure.

Observed in 2025

About this result

Category: Effectiveness
Evidence type: Reported measurement
What it covers: Dispute-submission preparation time on the internal agent platform

Key observation

50%+ of customer-support cases resolved by chatbot as first touch^{[Source 1]}

Reported by: Brex
Scope: Customer-support cases resolved by the chatbot at first touch (Brex customer-facing product AI, adjacent to this internal platform)
Denominator: Customer-support cases; exact sample size not provided

Observed in 2025

About this result

Category: Adoption output
Evidence type: Reported measurement
What it covers: Customer-support cases resolved by the chatbot at first touch (customer-facing product AI, adjacent to this platform)

Key observation

QA covers every support interaction; one person using AI instead of five QA specialists^{[Source 1]}

Reported by: Brex
Scope: Quality assurance of customer-support interactions
Denominator: Every support interaction
Method: Agent applies the quality rubric to every response; one person oversees instead of five QA specialists

Observed in 2025

About this result

Category: Effectiveness
Evidence type: Reported measurement
What it covers: Support-interaction QA coverage and the staffing change from five specialists to one

Key observation

KYC adverse-media accuracy 85% → 88%^{[Source 1]}

Reported by: Brex
Scope: KYC adverse-media classification; human versus agent accuracy

The source does not report the denominator of this figure.

Observed in 2025

About this result

Category: Effectiveness
Evidence type: Reported measurement
What it covers: KYC adverse-media classification accuracy, agent versus human

Lessons

Reported opinion: Brex's CTO recommends automating a useful portion of a workflow before pursuing complete automation.^{[Source 1]}

Brex assigns a systems engineering team to maintain its internal agent platform and shares capabilities with the customer product.^{[Source 1]}

Brex automates high-confidence fraud cases and sends the remaining cases to analysts with AI-generated findings.^{[Source 1]}

Brex operations staff and engineers translated KYC procedures into individual decisions and instructions, retaining human quality checks.^{[Source 1]}

Brex groups legal approvals by data retention, training use, and segregation requirements, with controls agreed by product and legal teams.^{[Source 1]}

Back to infrastructure

Improve this record

Give your coding agent the context and instructions to prepare a source-backed update.

Suggest on GitHub

Paste the prompt into your coding agent, then describe the correction or add a public source.

Preview prompt

Help me update the Internal Agents Map record for Internal Agent Platform at Brex.

Repository: https://github.com/steel-experiments/internal-agents-map
Record: data/agents/brex-agent-platform.yaml
Published record: https://internal-agents.com/agents/brex-agent-platform
Readable context: https://internal-agents.com/agents/brex-agent-platform.md

Use the correction or public source I provide in this conversation. If I have not supplied either, ask what I want changed before editing.

Use an existing checkout of this repository, or clone it if needed. Read the current record from the repository. Follow AGENTS.md, CONTRIBUTING.md, and the relevant instructions in data/schema.md.

Verify that the evidence supports the requested change. Preserve uncertainty and distinguish reported facts from catalog judgments. Follow the repository process for capturing new sources. If evidence is insufficient, explain the gap instead of inventing details.

Make a focused change on a branch. Regenerate derived files and run the required checks. Do not edit generated content by hand.

Show me the change, supporting evidence, and check results, including any checks you could not run. Ask whether to submit a pull request unless I already requested one. Do not merge the pull request. If you cannot edit the repository, prepare a source-backed issue draft.

Sources and research details

Citations link to the original publisher. Each source also keeps a preserved copy in the repository, so a changed or removed page stays checkable.

Agent, Human, Ops: How Brex Is Changing Roles and Workflowshttps://www.firstround.com/ai/brexCase study · Independent secondary · Last source verification: 2026-08-31
Preserved copy
The end of the trade-off: How AI agents broke the onboarding trilemmahttps://www.brex.com/journal/rebuilding-onboarding-ai-nativeEngineering blog · First party · Last source verification: 2026-09-17
Preserved copy

Question coverage and scope

purpose: Reported
workflow: Reported: The prompt studio and the MCP bridge are build-time mechanisms. Platform-wide capability and the L1/L2/L3 restructure stay out of the run description.
human involvement: Not applicable: There is no single platform-wide agent review boundary. Scoped downstream operating-model claims remain in research details; access controls are described separately.
implementation: Reported
validation: Reported: The quality-rubric agent checks every support interaction; the KYC human checks are cited with human involvement.
observations: Reported
lessons: Reported

Research details for every claim on this page

Summary
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
- ContextualizesThe end of the trade-off: How AI agents broke the onboarding trilemmaSeparate documented implementation: see dedicated agent record; not evidence of identical builder version.
Headline claim
Statement type
Metric
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
Reported by
Brex
Scope
Dispute-submission preparation using the internal agent platform; not end-to-end chargeback resolution
Denominator
Not reported
Method
Not reported
Observation date
2025
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 179–189
Sandbox
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
Harness
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
Interfaces
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
Tool access
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
Knowledge
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
Credentials
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
Prompt + eval studio
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
MCP tool bridge
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and Workflows
Prepare the dispute submission
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 179–189
Draft the collections response
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, line 173
Apply the quality rubric
Statement type
Fact
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 80–82
Key observation
Statement type
Metric
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
Reported by
Brex
Scope
Customer-support cases resolved by the chatbot at first touch (Brex customer-facing product AI, adjacent to this internal platform)
Denominator
Customer-support cases; exact sample size not provided
Method
Not reported
Observation date
2025
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 76
Key observation
Dispute processing: 3 hours → 3 seconds
Duplicate representation of Headline claim. Identical subject, values, period, scope, and locator as the headline; the shorthand adds nothing.
Statement type
Metric
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
Reported by
Brex
Scope
Dispute-submission preparation using the internal agent platform; not end-to-end chargeback resolution
Denominator
Not reported
Method
Not reported
Observation date
2025
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 179–189
Key observation
Statement type
Metric
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
Reported by
Brex
Scope
Quality assurance of customer-support interactions
Denominator
Every support interaction
Method
Agent applies the quality rubric to every response; one person oversees instead of five QA specialists
Observation date
2025
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 80
Key observation
Statement type
Metric
Provenance
Reported
Confidence
Not assessed
Confidence reason
No explicit confidence assessment is recorded for this claim.
Reported by
Brex
Scope
KYC adverse-media classification; human versus agent accuracy
Denominator
Not reported
Method
Not reported
Observation date
2025
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 163–167
Lesson
Statement type
Opinion
Provenance
Reported
Confidence
Medium
Confidence reason
The interview contrasts a 40% target with 100% and describes fraud cases still routed to analysts. These are the CTO's priorities, not measured optimal targets.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 153-159
Lesson
Statement type
Fact
Provenance
Reported
Confidence
Medium
Confidence reason
The interview identifies the platform team and explains how new external MCP tools become available internally.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 38-40
Lesson
Statement type
Fact
Provenance
Reported
Confidence
Medium
Confidence reason
The fraud example names the automation boundary. It does not establish that all complete automation fails on accuracy.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 159
Lesson
Statement type
Fact
Provenance
Reported
Confidence
Medium
Confidence reason
The interview describes a joint, line-by-line review of operating procedures and subsequent human review of LLM work.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 195-205
Lesson
Statement type
Fact
Provenance
Reported
Confidence
Medium
Confidence reason
The approval tracks and defined controls are described in the interview; vendor identity alone is not the stated basis for those tracks.
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 54-60
Operating model assessment
Level 5 for high-confidence operations case (support first touch, fraud recommendation) → automated completion; low-confidence remainder → analyst review; human attention boundary: exception-only.
Statement type
Inference
Provenance
Catalog judgment
Confidence
Medium
Confidence reason
The report documents confidence-threshold automation for fraud recommendations and describes L1 process work as fully automated in most cases, with the remainder reviewed by an analyst.
Observation date
2025-09-25
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 76, 159
Operating model assessment
Level 3 for KYC application → completed verification steps; human attention boundary: work-product review.
Statement type
Inference
Provenance
Catalog judgment
Confidence
Low
Confidence reason
The report documents human checks inside the KYC flow and humans reviewing LLM work, but a later passage shifts humans toward reviewing trends, so the per-run boundary is mixed in the evidence.
Observation date
2025-09-25
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, lines 143, 205
- ContextualizesAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, line 211 (humans shift toward reviewing trends)
Operating model assessment
Level 3 for collections follow-up interaction → drafted response options the servicing team selects, customizes, or sends; human attention boundary: work-product review.
Statement type
Inference
Provenance
Catalog judgment
Confidence
Medium
Confidence reason
The report documents the servicing team picking among agent-drafted response options before anything is sent; initial outreach itself is automated.
Observation date
2025-09-25
- SupportsAgent, Human, Ops: How Brex Is Changing Roles and WorkflowsPreserved content.md, line 173

Capabilities and architecture

Mechanisms

Documented uses

Access and controls

Reliability and validation

Adoption and operating evidence

Lessons

Related reading

Agents using this

Related implementations

Sources and research details