Keel · how it works

Proprietary framework

KEEL

The keel goes down first. Everything else is built on it.

The operating model is rarely decided. It's assumed.

Nobody sits down and chooses whether one team builds everything, or business units build their own, or something between. It just happens, by default, by whoever moved first. Then eighteen months in, nobody can agree what the CoE was actually meant to be.

An agent, not a questionnaire

What it reads, not what people remember

The usual way

Asks whether your gates exist.

The Keel agent

Checks that they were followed, on each use case.

Step one

What it connects to, read-only

Use-case intakeApproval and risk toolingDelivery pipelineModel registerTraining records
Keel agent

Read access only. Nothing is changed in your systems.

Step two

How it investigates

connectintake, approval and delivery tooling: read-only
discover23 use cases in flight. 14 went through intake. 9 did not.
scoreinvoice-triage scored at intake by a fixed rule: ships on its own
checkcv-screening: owner named, tested, fallback missing
routecv-screening sent for human review. Risk sign-off outstanding.
trackfirst ninety days: 7 of 12 gates met.

Every line is something a person can check afterwards.

Step three

What it produces

Use caseDecisionGate status
invoice-triageShips aloneMet
cv-screeningHuman reviewFallback missing
claims-summaryHuman reviewAwaiting sign-off
meeting-notesShips aloneMet

A use-case gate register.

How it is held to account

Rules, not vibes

01

One fixed rule at intake

Each use case either ships on its own or needs human review. The same inputs always give the same decision.

02

Gates are checked, not assumed

The agent looks for the owner, the test and the fallback, instead of asking whether they exist.

03

The foundation stays the same

Size, sector and market change which controls are switched on, not the model underneath.

The agent is held to the same questions it asks of everyone else.

Configured to your regulator

The same agent. Your market.

Controls for SM&CR ownership, model risk and Consumer Duty review.
Controls for FINMA's expectations and data protection under the Federal Act.
Controls for the EU AI Act risk classes and its human-oversight duties.

Choose a market. Other markets are added as packs.

How an engagement runs

From first conversation to a working agent

1

Scope

Which functions, which tooling, who decides.

2

Connect

Read-only connections to intake and delivery, reviewed first.

3

First gate register

Every use case, its decision and its gate status.

4

Run the gates

Live use cases are watched for drift and incidents.

A typical shape. The sequence and timings are set with you.

One platform

Keel does not work alone

KeelSets the policy and the gates for every use case.
SentinelQuestions what each agent can see, change and cause.
SoundingsSees what is actually running across the organisation.
BearingProves it to the supervisor, obligation by obligation.
The human layerHALO, Squall and Ensign measure what no system log can.

See the whole system on the Solutions page.

That's Keel

Want to know what shape your AI programme should actually take?

A demonstration of the method is one click away. The deployed Keel is built for your environment and configured to your regulator.

Discuss a deployment