How we work

Define. Judge. Deploy.

The edge is no longer who has the intelligence. It is where, how and why you use it. Three steps. Every engagement. No exceptions.

Define business reality

The Map · people, technology, data, cost

We sit with the work until we see how it really happens. Golden path and every exception. Who touches it, which systems, what data, what it costs today.

We also build the Mirror: a parallel AI team on approved data, read only, no production access, under NDA. You watch your own work run the new way before anyone commits to building anything.

What you get

Process record, written down and agreed
Baseline: volume, cycle time, error rate, hours, cost
Decision rights: who initiates, approves, reverses, owns
Data position: classification, residency, what agents may never touch

Where we stop

If we cannot write down how the work happens today, nothing gets built.

Apply expert judgement

Where the agent acts · where the human approves

We decide where AI belongs, and where it does not. Deterministic where it can be. Intelligent where it must be. Human where it matters.

Then we sequence: what is worth doing now, what waits for the data, and what should not be automated at all.

What you get

Intervention map, every step classified with the reason
No go list: what we will not build, and why
Economics: expected payback per candidate, before you spend

Where we stop

If the honest answer is do not build it, that is the deliverable.

Deploy the AI that works

Judged on ROI · economics · resilience

Built on what you already run. No forced migrations. Ownership transferred to you. Three gates, and nothing moves without clearing the one before it.

Nobody gets replaced. Your team moves up to judgement, ownership and approvals while agents carry the repeatable work underneath.

Proven

Real cases, failure taxonomy, pass threshold, cost per successful task. Zero unauthorised actions. No demo alone goes live.

Shadow

Runs against live inputs beside your people. Parallel, no consequences. Autonomy widens by evidence only.

Owned

Monitoring, runbook, incident owner, rollback. Your people hold the approvals. Exceptions feed the evidence set.

Why this shape

Governance first. Not bolted on later.

Most firms deploying AI are engineers learning governance last. We came the other way. Twenty five years in cybersecurity, risk and governance, then deployed engineering.

If your industry has a framework, we have been assessed against it. Defence, government, critical infrastructure, financial services and health does not leave many we have not had to prove.

Your board will ask who approved a decision an agent made. The evidence you need that day is the evidence we produce on the way through.

What it costs

Not day rates.

Foundation first. Define the business reality, build the Mirror, hand you the judgement and the sequence. One fixed fee, set when we scope the work.

Then Operating Tokens. One blended allowance for models, builds, workshops, governance and human led delivery. Only approved work consumes it. One monthly statement.

See the packages

Start here

The intelligence is commodity now. The judgement is not.

Humans lead. AI extends.

Start your engagement Email us