UKCompliance
A compliance co-analyst for regulated firms
For a London compliance-technology firm, we are building an AI co-analyst that investigates problems the way a firm’s own team would — and leaves a complete audit trail behind it.
- Client
- A London compliance-technology firm serving regulated companies
- Sector
- Compliance · UK
- Dates
- 2025–26
- Headline result
- About 90% of each investigation done before a person decides
The problem
When two systems disagree — a payment that doesn’t match, a record that doesn’t reconcile — a compliance analyst has to chase evidence across five or six systems, follow the firm’s procedure step by step, and write it all up for the auditor. It takes days, and most of it is gathering, not judging.
What we built
A co-analyst in a single chat window. It pulls the evidence from connected systems (or messages the right person when there’s no system to ask), works through the firm’s own procedure, names the root cause, recommends the next step, and records every action as it goes. It answers policy questions in plain English, with the exact rule it relied on.
What changed
Analysts arrive at a finished investigation instead of a blank page. A person still makes every reportable decision — and their name and time are on the record.
How it works
- 1
Gather
connects to the firm’s existing systems; where there’s no connection, it asks the right person.
- 2
Investigate
follows the firm’s own written procedure, step by step, rather than a generic playbook.
- 3
Decide
a named person approves anything reportable; the AI recommends, people decide.
- 4
Record
every step lands in a permanent audit trail, so the audit pack builds itself. Runs on the firm’s own servers or private cloud.
Results
What it delivers.
Evidence, procedure, and root cause done by the co-analyst
A permanent audit trail, written as the work happens
Every reportable call is made and signed by a person
“Compliance teams don’t need AI that decides for them. They need one that does the legwork and shows its working.”
Have a problem
like this one?
In two weeks, a Discovery Sprint shows you what superintelligence would do with it.