af AuditFellow.ai
Log in Sign up
For internal audit teams

A senior audit team, assisting you.

Your audit methodology, inside ChatGPT, Claude and Gemini. Ask for a finding, a risk, a control or a workpaper in your own words. It comes back in your team's format, checked, in a minute.

  • Write a finding from one line, in the shape your reviewers expect.
  • Correct it once and the correction becomes a rule for the whole team.
  • Nothing new to learn: it works inside the chat you already use, one key per person.
Try it free Watch it run

Works in ChatGPT, Claude and Gemini today. Microsoft Copilot is next.

MCPs & Tools Audit Methodology Understand Research Plan Build Execute Review Finding Audit Report Testing Procedure Audit Memo Auditor Prompt Model + Harness Audit-ready Deliverables
How it works

A methodology layer around the model, with a learning loop behind it.

AuditFellow does not replace the chat your team uses. It attaches a structured methodology to each request at the moment the chat sends it, makes the model reason through it in a fixed order, checks the result against an output contract, and turns your reviewers' corrections into versioned rules. The model writes; the harness decides what a correct answer looks like.

1 · Request-level injection The harness travels with the message, not with the person The browser extension intercepts the request the chat sends to its own server and appends the harness: the reasoning core, the catalogue of skills with their output contracts, the global and team rules, and the knowledge files. The model your company already pays for does the writing. Nothing is sent to AuditFellow: the only traffic to us is a daily key check and the pack download. The full harness travels once per conversation; later messages carry a short reminder, and a compact variant fits chats with small context windows.
2 · The reasoning loop Eight phases before a word of the deliverable is written Understand the request; inventory the tools the session really has; recall what the conversation already settled; map the request to exactly one skill, with hard isolation so no other methodology leaks into the shape; choose the safest route; ask at most two questions, only for inputs the deliverable cannot exist without; execute under the contract; deliver in the person's channel and language. Any value the organization did not supply, a rating scale, an owner, a target date, is written as N/A. It is never invented.
3 · Skills with output contracts A skill is a specification, not a prompt Each skill holds the methodology for one deliverable (a finding, a risk statement, a control, a workpaper, a data strategy, an AI usage document), an output specification with the fields, their order and their labels, and a self-review checklist rendered from that specification. The checklist runs before delivery, and it names the known failure modes explicitly, for instance a "Risk" section where Business Impact and Risk Level belong. Every answer opens with the line "AuditFellow · task", so anyone can see which methodology was applied.
4 · The learning loop Corrections become rules only after they prove themselves A reviewer corrects an answer in plain words. A distiller turns the correction into a candidate rule, conceptual rather than a string match. A gate then re-runs a set of requests with and without the rule and scores both against the output specifications; only a rule that raises the score is applied. Rules are versioned with rollback, live in the team layer (encrypted storage we cannot read, or your own endpoint, holding distilled rules only), and reach every key of the organization with the next pack, within a day. Admins and coordinators feed the team layer; auditors keep their corrections on their device.

Every night we run our own prompts through the shipped harness against a model and score the shape of the answers, so a regression in a chat platform or a model update shows up on our side before it reaches your team. Changes are listed on the Updates page inside the app.

Why teams use it

A senior audit team at your side, from the first task.

100% of corrections verified applied
Your conventions, learned once Reviewers correct the draft as they always have. Each correction becomes a rule, and each rule is verified applied. Nothing to configure.
Conventions this team taught it
"observation", never "finding" severity on a 1 to 4 scale owner and target date on every recommendation
91% accuracy with the harness
Fitted to how your team works The deliverable comes back in your team’s structure and language, right the first time: a finding, a risk, or an executed analytic test.
Benchmark
Claude Fable 5 or Gemini 3.1 Pro, bare
70%
Sonnet 5 or Gemini 3.5 Flash + harness
91%
6xfaster 69%fewer tokens
Faster and cheaper Under the harness, a small model outperforms a frontier one. The expertise travels in the harness, not in the model, so you stop paying premium rates for structure.
Tokens per task
Bare model
100%
With the harness
31%
New expertise in every update Each update ships packaged review frameworks: today cybersecurity, GenAI applications, fraud and irregularities, controls, risk, workpapers and data strategy. Your team audits domains it has never staffed, with the depth a specialist would bring.
Auditors on your team, with specialist knowledge in
Payments Insurance Safety CommOps Cybersecurity GenAI Cybersecurity GenAI applications fraud and irregularities, data strategy

How we measure: the same requests run in our workbench with and without the harness and scored against the output specifications, field by field. Latest gate run: 9.4 against 3.4 out of 10 on the issue writer. Token and time figures come from the same runs. We publish each run on the Updates page inside the app.

See it run

Write a finding

Recordings of real runs, with sample data. All seven examples.

Try it on your next finding.

Seven days free, no card. Install, paste your key, run the same request with and without the harness. If you like it, use it. If not, go back to the cave.

Sign in with your work Google account. Your organization is created on the first login, and you are its admin. Start free trial Already have a key? Log in
Contact

Talk to the person who built it.

Team pricing, own-storage setup, or a live comparison on your own example. Answered within a business day.