Relay

What Relay is

An AI coworker for auditors.

An AI coworker for auditors is AI that works inside the audit engagement: it knows the client's files, the prior year and the firm's methodology, and it does the manual audit work — keying figures, tying them to source, building workpapers — while the auditor reviews, judges and signs. Relay is built as exactly that.

The engagement

A colleague who knows the whole engagement.

A general AI tool answers the question you bring it, with the context you paste in. Relay works inside the engagement, with the files, the prior year and your firm's methodology already in front of it, and it carries all of that from one task to the next. The manager who already knows explains what that changes on a job.

Knows

  • The client and the engagement folder
  • The current-year and prior-year workbooks
  • Your firm's templates, referencing and documentation standards
  • The client's emails and the source documents behind each balance

Does

  • Finds the figure and the document behind it
  • Keys the figures into the workbook
  • Ties each one back to its source
  • Lays out the testing to your methodology
  • Updates the file when the figures change
  • Flags what does not reconcile

Stays with you

  • Reviewing the work
  • Investigating the exceptions
  • The professional judgement
  • The sign-off and the opinion

Four ways

Four ways to put AI on an audit.

The difference between them is what each one leaves for the auditor to do.

Ask a general AI

ChatGPT or Claude answers the question you bring it. You supply the context, and you carry the answer back into the file yourself.

Audit software

CaseWare, Draftworx and their peers hold the file and manage the engagement. The work inside the file is still done by hand.

Automation tools

A script or a point tool automates one defined task, one workpaper at a time, with no knowledge of the rest of the engagement.

An AI coworker

You give it the engagement and tell it what needs doing. It works through the groundwork — keying, tying, building the workpaper — and gives you back the part that needs your judgement.

In practice

What you say to Relay.

Relay takes instructions the way a colleague does: in a sentence, in the order the audit runs, with your firm's standards assumed.

  • “Help me plan the procedures for this engagement.”
  • “Build the workbook from our template and use our referencing standards.”
  • “Tie the current-year figures back to the trial balance and the financial statements.”
  • “The client sent revised figures. Update the file.”

One engagement, start to finish. In the demo on the homepage, Relay is pointed at an engagement folder holding the trial balance, the draft and prior-year financial statements and the general ledger. It reads the files and reports back the revenue profile, the balances that matter and its first observations. Asked for the payables work, it proposes the procedures and a risk-based sample, and on approval builds the workpaper: a lead sheet reconciled to the trial balance, supplier testing, subsequent-payment testing and a source index tying every figure back to the client's own documents. Watch the demo, or read how Relay builds AI audit workpapers.

FAQ

Questions auditors ask.

What is an AI coworker for auditors?

An AI coworker for auditors is AI that works inside the audit engagement. It has the client's files, the prior year and the firm's methodology in front of it, it does the manual audit work — keying figures, tying them to source, building and updating workpapers — and it hands the result back to the auditor, who reviews, judges and signs. Relay is built as one.

How is Relay different from ChatGPT or Claude?

ChatGPT and Claude answer the question you bring them, with whatever context you paste in. Relay works inside the engagement: it already holds the files, the workbooks, the prior year and your firm's methodology, and it does the work in the file. Relay runs on large language models underneath; the difference is everything built around them so they can act inside an audit engagement, with the work recorded and every figure traceable.

Is Relay an AI audit agent?

In the technical sense, yes: an agent is a language model given tools, workflows and guardrails so it can act on a task instead of describing it. Relay's harness does that for audit work — read the Excel, pull the figures from a PDF, apply the firm's methodology, draft the workpaper, record what it did. We say coworker because that is what it is like to use: you hand work over, and it comes back for your review. How the harness works is explained in the blog.

Does Relay replace auditors?

No. Relay does the manual half of the audit, which is where most of the hours and most of the quality failures sit. Risk assessment, professional scepticism, the judgement on each exception and the audit opinion stay with the auditor, and the file carries their name. We have written about what auditors will do when the workbook builds itself.

Does Relay remember the engagement between sessions?

Yes. Relay carries the engagement: the prior-year working papers, the trial balance, the client's accounting policies and the supporting evidence, connected in one knowledge graph that grows as new evidence arrives. When figures change, it knows which work is affected, and nobody has to re-explain the file at each step.

Does Relay conclude or sign anything?

No. Relay never concludes. It does the work and shows everything behind it — every figure traceable to its source, every test recorded — and when the evidence is thin or the numbers do not reconcile, it flags that for you. The judgement, the sign-off and the opinion stay with you.

Demo

Run Relay on your next engagement.

Leave your work email and we'll be in touch with your firm to arrange a time.