Question 1

What is Relay?

Accepted Answer

Relay is an AI platform for audit. It automates the execution layer of an audit — keying in figures, tying out, building workbooks and workpapers, testing populations, and compiling evidence — so auditors can spend their time on professional judgement instead of data entry. Relay is built by a team that includes ex-auditors.

Question 2

Is Relay an AI copilot for auditors?

Accepted Answer

Yes. Relay works as an agentic copilot for audit teams: its agents run the grind of audit execution and bring every step back to you to review and sign off. Audit becomes a relay of agents and human expertise — the agent does the grind, the auditor does the thinking.

Question 3

Can AI actually do audit work?

Accepted Answer

AI can do the execution layer of audit — the repetitive, time-intensive grind of building workbooks, performing testing, and compiling evidence, which is where most audit-quality failures actually originate. It cannot and should not do the judgement: assessing risk, exercising professional scepticism, and forming the opinion remain with the auditor.

Question 4

Does Relay sign or conclude the audit opinion?

Accepted Answer

No. Relay never concludes. It does the work and shows you everything behind it, but the judgement, the sign-off, and the opinion stay with you. The one thing you can't delegate is the thing you're liable for, so Relay isn't even trying to.

Question 5

Can you trust Relay's output, and is it reviewable for inspection?

Accepted Answer

Everything Relay produces is checkable, and checkable fast. Every figure is traceable to the source it came from, and every test is recorded — what it looked at, against which standard, and what it found. When the evidence is thin or the data won't reconcile, Relay flags it for a human rather than papering over it. When a regulator inspects the file, it's your name on it, and you can follow exactly how every conclusion was reached.

Question 6

What can Relay automate in an audit?

Accepted Answer

Relay automates audit execution: planning preparation, evidence requests, building draft workbooks and workpapers from a client's trial balance and supporting documents, sampling and population testing, exception analysis, formula and consistency checks, and reviewer-ready write-ups — with the risks and materiality set during planning carried through testing, and the whole file rebuilt in seconds when the client changes their figures.

Question 7

What is an AI audit harness, and why does it matter?

Accepted Answer

An AI harness is everything around a large language model that lets it work inside an engagement rather than just answer questions about one: the tools to take action (read Excel, pull figures from a PDF, query the firm's methodology, draft a workpaper), the skills that string those tools into repeatable workflows, the loop in which the agent inspects context, acts, observes the result, and decides what to do next, and the guardrails — permissions, review requirements, audit trails, source citations, and human sign-off — that keep it safe. A large language model plus a harness is what Relay calls an audit agent. The harness is what turns a clever reply into actual audit work, which is why it matters even more in audit than in most other fields.

Question 8

Does Relay start each audit from scratch, or does it remember the engagement?

Accepted Answer

Relay carries the engagement rather than starting cold each time. It reads the prior-year working papers, the trial balance, the client's accounting policies and the supporting evidence, and builds them into a single connected knowledge graph — which policy governs which account, what support matches which amount, what last year's team concluded and why. That context map evolves as new evidence arrives during the engagement, and when figures change Relay knows which work already performed is affected. It applies your firm's own methodology — your templates, sampling rules, and documentation standards — consistently across every engagement, so no one has to re-explain the file at each step.

Question 9

What is Audit Bench?

Accepted Answer

Audit Bench is Relay's AI evaluation framework for real-world audit work. It uses lifelike audit scenarios — synthetic but grounded, designed with chartered accountants — to benchmark how different models (GPT-5, Claude, Gemini, Llama, custom agents) plan, inspect evidence, apply methodology, handle exceptions, and produce reviewer-ready output. It benchmarks audit judgement, not chat.

Half your audit can be automated by AI.Relay does that half.

What Relay is built on.

Large Language Models

Agentic Harness

Data and Integrations

Context and Knowledge

Audit Capabilities

Product and Interface

Security and Governance

Audit execution, run by agents — reviewed by you.

Benchmarking audit judgement, not chat.

Real audit files, rebuilt as controlled cases

Designed with chartered accountants

Compare agents on audit execution

Run Relay on your next engagement.