AI agents that survive contact with your compliance team.
We build the agent into the process you already run, with what it is allowed to do written down first, a named person on every consequential action, and a record you can replay.
The pilot works. It is still a pilot.
- Nobody could say what the agent would be allowed to do in production, so it never went.
- A vendor's log proves their platform recorded the action. It does not prove anyone allowed it.
- The escalation rules live in a config screen nobody can show an auditor.
- The agent's mistakes are found afterwards, by the person they landed on.
- Compliance said no. Not to AI, to AI nobody could account for.
An agent goes to production when the answer to “who allowed that?” exists before it acts, not after.
In. Build. Leave. Prove.
One job, four weeks, in production. Then it keeps proving itself.
In
The process, the permission list, and who signs. Written down before anything runs.
Build
The agent runs in shadow first, then live with approvals, inside your environment.
Leave
Your team owns the permissions and the approval points, and can change them.
Prove
Every action it took, every one it was refused, and who allowed the rest.
What you are left with.
- One agent in one process: triage, drafting, classification, reconciliation or intake.
- The permission list, written before it runs and owned by your team afterwards.
- Approval points with named people, at the steps that carry consequence.
- The record of every decision, replayable any day.
- The measure of what it did against the old way, taken the same way both times.
Can you show that this action was allowed before it happened, and by whom?
What the AI is allowed to do is written down first.
Every decision is checked against those rules before it happens.
A named person signs it off. The check produces the evidence; a person judges.
You can replay the whole history any day and get the same answer.
If anyone changes it later, it shows. Patent pending, UK application GB2620101.2.
The request
Show us that the Aurora Banking app meets EN 301 549.
From a procurement team · 2 Sep
What you show them
Account overview
42 checks
Signed off
QA lead · 4 Sep
Payments
38 checks, 2 fixes re-checked
Signed off
Product owner · 6 Sep
Onboarding
29 checks
Signed off
QA lead · 9 Sep
That is what a deployment leaves running for your job, and what the record holds. In the accessibility product today, a person on your team accepts every finding before it reaches your record.
Who this is for.
- Operations, digital and risk leaders in regulated businesses.
- Teams with a process that is high volume, rule-bound and currently manual.
- Anyone whose compliance team has already refused an agent once.
Not for
A chatbot on a marketing page. This is an agent inside a process that carries consequence.
Who you deal with.
You deal with Simon and Jason from the first call. One engineer sits inside your team, full time, for the four weeks.
Simon Milner, Founding Architect
He designed the record: what the AI is allowed to do, checked before it acts, and replayable afterwards. Twenty-five years in Silicon Valley before that.
Jason Crispin, Founder
He owns the customer side of a deployment: what the job is, what it is worth, and that it lands. He is on the first call and every one after.
Four weeks, then it keeps proving itself.
Week 1
The baseline
What the job is, what allowed means for it, and who signs. Written down before anything runs.
Weeks 2 to 4
The build
Our engineer works in your codebase next to your developers. The old way and the new way run side by side.
Week 4 on
The proof
Every decision checked and recorded, on AUDITSU, on a monthly subscription. Replay it any day. Your team runs the job; we maintain it if you ask.
What you keep
- The agent, running in your process.
- The permission list and the approval points, owned by your team.
- The code, assigned to you in writing.
- The record of every decision, replayable any day.
What people ask.
What can the agent not do?
What can the agent not do?
Whatever you have not allowed. The permission list is written before it runs, in your words, and anything outside it is refused and recorded as refused.
Who approves what it does?
Who approves what it does?
People you name, at the points you choose. The agent can propose anything; only a named person can make a consequential action happen, and the record says who did.
What happens when it gets something wrong?
What happens when it gets something wrong?
The approval point catches it before it takes effect. The record shows what was proposed, what was allowed and what was refused, and you can replay it any day. Shadow mode runs first, so the first mistakes land on nobody.
How much of our team's time does this take?
How much of our team's time does this take?
The process owner, a developer who can review, and the people who will approve. Week one is agreeing the process and the permission list in writing. From then on, approvals are the work those people already do, now with a record.
Where does the data go, and what does your engineer see?
Where does the data go, and what does your engineer see?
We build in your environment, on your accounts, with access you can revoke any day. The agent sees what the process already sees and nothing more. If the data cannot go to a hosted model at all, that is the sovereign deployment.
How is this different from the agent builder inside the platform we already pay for?
How is this different from the agent builder inside the platform we already pay for?
Their guardrails live in their platform and their log proves their platform recorded the action. Ours live with you, the check runs before the action, a named person signs, and the record can be replayed by anyone you show it to. You can change model or platform later without rewriting the rules.
What does it cost, and what is fixed?
What does it cost, and what is fixed?
The scope and the four weeks are fixed. The price depends on the process and is set on the first call, before anything starts. You leave the call knowing what it would take and what you keep.
What happens after the four weeks?
What happens after the four weeks?
Your team owns the permissions and the approval points and can change them. The check and the record keep running on AUDITSU, on a monthly subscription, and the next process goes on the same record. We maintain the agent itself if you ask.
Where to go next.
Which process would you hand over first?
Thirty minutes with Simon and Jason. Bring the process and the rule it has to respect. If the honest answer is not to use an agent for it, we will say so.