How we work.

Boring AI in production. By day twenty-eight.

Fixed fee

no day rates

2–4 weeks

calendar duration

Founders deliver

no juniors between

Most AI engagements fail in the shape of the contract, not the technology.

Six-month proposal. Slide-deck commitments. Real problem discovered at month three.

We do it in reverse.

The four-week journey

Week 1

Map and Connect

Pick the workflow. Map your systems.

Week 2

Build and Govern

Build it on your real data.

Week 3

Harden and Train

Layer governance. Train your team.

Week 4

Prove and Decide

Run live. Decide your next move.

Shorter problems run as 2-week sprints. Same shape, fewer weeks.

What you walk away with

A live workflow

Architecture design

Solution design

Requirements

Risk assessment

Best practice training

The live workflow stays only if you subscribe. The other five are yours regardless.

You bring:

one workflow.

one exec sponsor.

two or three SMEs.

one set of credentials.

That's it.

You don't need an existing AI strategy.

The decision at the end

Subscribe

  • · System stays live
  • · Add more workflows
  • · Ongoing support included

Cancel

  • · System decommissioned
  • · Five deliverables retained
  • · No further commitment

The sprint fee is paid either way.

What “governed AI” actually means

Immutable audit trail

Every action — every call, every decision, every data touch — logged and unalterable. Your regulator can reconstruct exactly what happened.

Policy at the boundary

Rate limits, approval thresholds, what an agent can touch — enforced as policy, not asked for in a prompt. Prompts are suggestions. Policy isn't.

Versioned contracts

Every outbound action operates under a versioned contract — schema, policy, and audit in one reviewable artefact. The same control surface your regulator already looks at.

A rollback path

Every action is reversible, or has a defined point where reversibility stops. When something goes wrong, there's a defined way to undo it.

Explicit data boundaries

Your data stays in your systems. Nothing trains vendor models. We store scrubbed telemetry for audit — nothing else.

Governance should be invisible when everything is working — and obvious the moment it isn't.

Two markets, same spine

Specialty insurance

Our beachhead. Regulation is the hard part — we've made it boring.

Small and medium firms

Same spine, shaped for firms our size. HubSpot, Xero, Microsoft 365, plus the traditional systems nobody talks about on stage.

Small by design

The people scoping the work are the people building it.

Agents do the work that doesn't need judgement. Humans do the work that does. Capacity scales with architecture, not headcount.

Boring AI. Built for business users.

15 minutes. No slides. A conversation about whether there's a real problem here.

Book a call
Ask Moss about our approach