Everyone will tell you what to do with AI. We tell you what it's doing to you.
We look at your IT, your data and your AI, and we show you where it breaks — before it does.
The method
Three questions, every time
Economics
Expensive is fine if it earns a multiple. We set the measures so you can prove it.
Reversibility
Can you still change your mind — and what would leaving cost?
Exposure
Who reaches your data, and how much rests on one provider.
Most organisations can tell you what their AI cost. Very few can tell you what it returned.
We don't allege anything about your vendors. We read what you already agreed to, and show you what you cannot get back.
Scope
Three layers, because the problem is rarely where you think
We define the ambition, then the solution that matches it.
We build the foundation everything above depends on.
We test whether the decisions hold, and name the partners worth talking to.
Foresight
Why AI investments fail
Not because the technology fails. Because nobody agreed what it was for, what good would look like, or what it should cost.
We settle those three before you spend, and we model where your version breaks.
If data is the new oil, you have drilled the well and loaded it onto ships owned by someone else — and you are paying them for the privilege.
Modelled scenarios, not predictions.
When to call us
At the start, ideally
The cheapest decision to change is the one you have not made yet. We would rather help you design it than explain it afterwards.
Before the ambition is set, the stack is chosen or anything is signed. We define what good looks like, and design so you can still change your mind.
What you actually bought, what it returned, and what moving would take.
Money spent, nothing agreed about what good looks like.
Deliverables
What you walk away knowing
- What every layer costs you, and what leaving would take. A dependency map scored on economics, reversibility and exposure — yours to re-run each year.
- What to stop, and what to back. Every review names at least one of each.
- What you are handing your vendors, and how to hand over less.
Nothing to sell you.
- No money from AI vendors. The buyer is the only party paying us.
- No review without a kill list.
- Nobody learning on your account.
- Your name is never published. Findings, never fingerprints.
In this region
Why this matters here, now
Regulators here are moving faster than most of the world, and on exactly these questions. Korea's framework reaches any business whose AI affects Korean users. Japan is evaluating domestic against foreign models, with control of infrastructure an explicit criterion.
The practice
Judgement, not headcount
Staffed by practitioners who have run these functions, scaled around the problem rather than a bench you would pay for.
- Advisory. Technical and business strategy leaders — data and AI across financial services, insurance, infrastructure, energy and manufacturing, and with governments and public-private institutions.
- Security. Serving CISOs, in the room when it matters.
- Engineering. Senior engineers and analysts across Asia and Europe.
You get the people who have done it, and none of the people learning on your account.
Meet the partners — the record, with links →
We take on a small number of senior associates each year — people who have run these functions rather than advised on them. If that is you, write to us.