All projects

Confidential (audit firm) · Audit and compliance

AI compliance agent generating clause-level tracked-change findings.

Private deployment, client project

Category

AI

Year

2025

Capabilities

AI, Cloud & Automation

Industry

Audit and compliance

Compliance Review Agent on a laptop
Compliance Review Agent on a phone

The problem

Manual compliance and audit reviews against established standards were slow and inconsistent between reviewers, with each report taking about a week to complete.

Our approach

We built an agent that checks uploaded reports against an established standards dataset, then generates a findings report with clause-level tracked changes and AI-written comments explaining every flagged issue.

Inside the product

More than one screen.

3 more screens from Compliance Review Agent, captured at device size.

Compliance Review Agent: Review queue on a laptop

Review queue

Documents split into clauses and checked against the selected standards.

Compliance Review Agent: Findings report on a laptop

Findings report

Findings by severity, exported as tracked changes and margin comments.

Compliance Review Agent: Standards library on a laptop

Standards library

The paragraphs every finding cites, with the firm’s own guidance alongside.

On the phone

Compliance Review Agent: Home on a phone

Home

Compliance Review Agent: Review queue on a phone

Review queue

Compliance Review Agent: Findings report on a phone

Findings report

Compliance Review Agent: Standards library on a phone

Standards library

How it is built

Architecture

A Python and FastAPI backend deployed on AWS handles report ingestion and comparison against the standards dataset. Claude powers the review and comment generation, returning findings tied to individual clauses for a traceable review workflow.

Key features

Report upload and comparison against a standards dataset
Clause-level tracked-change findings
AI-generated comments for every finding
Consistent, repeatable review logic across reports

Challenges & solutions

Where it got hard.

Challenge

Broad AI feedback was not precise enough for reviewers who needed to see exactly which clause triggered a finding.

Solution

Grounded each check in the standards dataset and returned findings as clause-level tracked changes with an explanation attached.

Challenge

Reviewer judgement varied between reports, making outcomes difficult to compare and slowing final sign-off.

Solution

Encoded the review as one repeatable agent workflow so every report is evaluated against the same standards and output structure.

Results

What shipping it changed.

50%shorter audit turnaround, from about one week to roughly half that.
2xfaster review process with more consistent outcomes.

Stack

  • Backend

    • Python
    • FastAPI

  • Cloud

    • AWS

  • AI

    • Claude (Anthropic)

Services involved

Let's talk

We can usually tell within one call whether the approach above transfers to your problem, and what would need to change.

Start a projecthello@feinwerk.software