The Agent does the fieldwork.You sign the opinion.
The Agent turns the client's documents into finished workpapers, with each cited figure one click from its page.
Request demo
Take on more engagements at once. The Agent splits large jobs across helpers that work side by side and hands back workpapers ready for review.
Get more done with the Agent
Your team's hours go to judgment and to the client, while the Agent works through the documents in the background.
Helpers in parallel
The Agent splits a large job across up to five helpers that work at once, then brings their results together. Open any helper to follow its work, write to it or stop it.
Picks the work back up
Long runs carry on in the background. Ask the Agent to come back at a set time, after a while, or as soon as the client's files land in a folder.
Any document in, Excel and Word out
PDFs, scans, photos, emails and spreadsheets go in. The Agent delivers into your workbook and your Word report.
Stay in control of the work
The Agent never waits for permission to work: the only stop is a question the documents cannot settle, and you can steer the work at any moment.
Asks when judgment is needed
When the decision is yours, the Agent asks: one card, a few clear options and room to write your own.
Steer the work as it runs
Write while the Agent works and your message is read between one step and the next, without waiting for the turn to end.
Stop anything, any time
Stop one helper, one background run or the whole turn whenever you want.
Built to your standard
The Agent works in your templates and to your criteria, and every save checks the cited figures against their sources.
Your workbook, filled in place
The Agent fills the client's own workbook, in its sheets and its columns, and writes the report beside it in Word.
Figures traced to their page
A figure taken from a document carries its source inside its own cell. Click it and the page opens with the passage marked.
Checked on every save
Every save recalculates the workbook and checks each cited figure against its source. Anything that does not match goes back to the Agent.
Memory
What the Agent learns, it does not forget
A session ends; what the Agent learned in it stays. The Agent remembers what your firm knows and how each of you works, so nothing has to be explained twice and every engagement starts where the last one left off.
- The firm's memory
- Shared by the whole team: audit criteria, each client's particulars, how every kind of engagement is resolved. What the Agent learns with one of you, it applies for all.
- Each auditor's memory
- Yours alone: what you decide, the Agent remembers for next time, without changing what anyone else sees.
How teams use the Agent
Every invoice against its payment, paired automatically, with the evidence always in view.
See the use case
Security built into how the Agent works.
Every session is isolated, every request is checked and every save is kept.
More about securityAn isolated machine per session
Each session runs on its own machine, and the place where the Agent works has no network access.
Your organisation and nobody else
Access is membership of your organisation, checked on every request before it reaches the Agent.
Every save, a new version
Nothing is overwritten: each save lands as a new version of the file.
Questions about the Agent.
What exactly is the Agent?
An agent that carries out the document fieldwork of an audit engagement: it classifies the supporting documentation, extracts its data and matches it across documents. It is not a chatbot that summarises documents. It delivers the workpaper, with every figure cited to its evidence.
What do the helpers do, and why are there several?
An engagement has tasks that can move at the same time. The Agent opens helpers for them, up to five at once, and each one reports back with its result. The Agent can write to a helper, stop it or pick it up again. Each helper gets one precise brief; how many the Agent opens, and for what, depends on the engagement.
Who do I talk to?
Always the Agent. You say what you need from the engagement and the Agent decides what runs, in what order and when the work is done. You orchestrate nothing. The Agent only stops to ask you when it is missing a judgement that belongs to the audit.
How is it different from a general purpose AI?
Its room to act is bounded and auditable. The Agent works on your organisation's documentation with a closed set of tools, every figure it writes carries its source inside the cell, and every save checks that figure against its source. A summary that says "it matches" is not audit evidence; a figure with its source page is.
Where are my documents processed?
In the European Union. G2-F is hosted on AWS in Frankfurt (eu-central-1), processing stays in the EU, and your documents are never used to train or fine tune any model.