Service · Governance

AI Governance for financial services.

In financial services, governance is not a later phase. It is part of the product. Airclerk designs the controls, approvals, audit trails and operating model required for production Claude workflows.

01 / The framework

Ten questions every Claude workflow must answer.

G-01
What can Claude access?
G-02
What can Claude recommend?
G-03
What must Claude not do?
G-04
When is human approval required?
G-05
What gets logged, and where?
G-06
How is evidence captured?
G-07
How are outputs reviewed?
G-08
How are incidents handled?
G-09
How are workflows evaluated?
G-10
How is AI usage monitored over time?
02 / Principles

Our governance principles.

P-01

Human-in-the-loop by design

Critical decisions are reviewed by people. The model proposes, humans dispose — and the system records both.

P-02

Permission-aware access

Claude inherits the user's permissions. No bypassing the existing access model.

P-03

Auditability as a feature

Every action, prompt, tool call and output is recorded with traceable provenance.

P-04

Evidence capture

Outputs cite their source documents and the reasoning path that produced them.

P-05

Clear system boundaries

What Claude can read, write and trigger is defined per workflow, not per organisation.

P-06

No black-box production decisions

If we cannot explain why a workflow produced an output, it does not ship to production.

First step

Assess your Claude production readiness.

The Readiness Sprint includes a full governance and controls model for the recommended first workflow.

View the sprint