Pricing models, claims and vulnerable customers. Every decision attributable.
An engineer joins your team for four weeks and ships one job. What stays is a record of which model or agent did what, what it was allowed to do, and which named person released it.
Automation moved faster than the evidence.
- Claims triage is automated, and the reasoning behind an individual outcome is hard to reconstruct.
- Consumer Duty outcomes are evidenced with operational metrics used as a proxy.
- Communications in a customer’s chosen format are promised once and honoured unevenly.
- A pricing model changed, and the record of who approved that change sits in a deck.
- Model risk expectations ask for attestations, and the underlying evidence is assembled by hand.
- From December 2027, pricing sits in a high-risk category with technical documentation attached.
The regulator asks which decision this was, who was accountable for it, and on what basis.
In. Build. Leave. Prove.
One job, four weeks, in production. Then it keeps proving itself.
In
The job, the rules it has to respect, and the people who release each decision.
Build
Built in your environment, with the old way and the new way running side by side.
Leave
Your team runs it, and owns the rules and the approval points.
Prove
Every decision checked before it takes effect and recorded, replayable any day.
Three jobs a four-week deployment could take.
- Claims triage and first-notification drafting, with every customer-facing draft released by a named handler and the decision path recorded.
- Vulnerable-customer identification and renewal communications in the format the customer asked for once, honoured everywhere.
- A model change record: which model version priced which quote, approved by whom, and against which rules.
- Whichever we start with, the record is the same: what was allowed, what happened, who signed.
Which decision was this, who was accountable for it, and on what basis?
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.
That is what a deployment leaves running for your job. In the accessibility product today, a person on your team accepts every finding before it reaches your record.
Who this is for.
- Insurers, managing general agents and underwriting-decision firms automating customer-facing work.
- Risk and compliance leaders who attest to model governance.
- Teams preparing for Annex III technical documentation ahead of December 2027.
Not for
A model monitoring dashboard. This is the evidence of permission underneath the decisions, and the job that produces it.
Two people, on every call and in your standup.
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 every deployment: what the job is, what it is worth, and that it lands. He is on the first call and every one after.
Patent pending, UK application GB2620101.2. Meet the team
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. Replay it any day. We maintain it, or you run it without us.
What you keep
- The job, live, in your process.
- The rules and approval points, owned by your team.
- The code, assigned to you in writing.
- The record of every decision, replayable any day.
What people ask.
Does the European Accessibility Act apply to insurers?
Does the European Accessibility Act apply to insurers?
Not as the consumer banking services list is written, so we do not claim it does. In the UK the hooks are Consumer Duty and the Equality Act, and neither names a digital standard.
How does this sit with model risk expectations?
How does this sit with model risk expectations?
The PRA’s model risk principles have been in force since May 2024 for firms in scope, with senior-manager attestations. What this gives you is the evidence underneath one: which model version made which decision, approved by whom, and when.
What about the EU AI Act?
What about the EU AI Act?
Insurance pricing sits in Annex III, and those obligations apply from 2 December 2027 after the delay. Building the record now is what makes the technical documentation a by-product rather than a project.
Can AI decide a claim?
Can AI decide a claim?
Not here. It can triage and draft. A named handler releases every customer-facing decision, and the record says which decision was theirs and what it rested on.
How does this help the Consumer Duty board report?
How does this help the Consumer Duty board report?
The evidence for outcomes stops being assembled from operational metrics used as a proxy. Each customer-facing action carries who allowed it, on what basis, and when, which is what the report is trying to say.
What does it cost?
What does it cost?
Scoped on the call, because it depends on the job.
Which decision would you have to explain first?
Thirty minutes with Simon and Jason. Bring the process or the attestation you are preparing, and you leave knowing what it would take.