Unpuzzled advises chief executives, boards and senior teams on AI: whether to adopt it, how to deploy it, and what to do if a deployment runs into trouble. We’ve seen what goes right and what goes wrong, and we’ll tell you which one you’re looking at.
You don’t need to know how the model works. You need to know whether you can rely on it, what it costs if you can’t, and what to do next.
Who we advise
Chief executives · Chairs and boards · Finance, operations and technology leaders · General counsel. In every kind of business.
Where we help
Stage 1 · Considering it
We test the proposals in front of you against what we’ve seen succeed and fail elsewhere, and tell you which are worth taking forward.
Stage 2 · Buying or building
We read the proposal and the contract from the technical side, and show you what the supplier can and can’t promise.
Stage 3 · Running it
We review the deployment as it runs today against what it was meant to deliver, and tell you what to measure from here.
Stage 4 · Something’s wrong
We find out what happened and why, quickly and quietly, and give you options you can defend.
From experience
Trouble often starts with an assumption nobody checked. These four come up again and again.
“The pilot went well.”
Pilots run on cases someone chose. Production brings the ones nobody chose.
“It gives the same answer every time.”
These systems can give multiple answers to the same question. That’s harmless in a draft, and serious anywhere the output is relied on.
“Our supplier has it covered.”
Your supplier usually builds on a model from someone else, who can change it, reprice it or retire it on their own timetable.
“We’ll measure the return once it’s running.”
Without a baseline taken before go-live, there’s nothing to measure it against.
And when it works
The deployments that pay their way tend to share the same few things.
How it works
Four steps, at whatever pace the situation needs.
Call or write, and tell us what’s on your mind. No preparation, no obligation, and nothing goes further.
We look at what’s in front of you: the proposals, contracts, systems and people involved.
Written for the board, in plain English: what’s happening, what it means, and your options.
If you want us there while decisions are made and put into practice, we stay on call.
Confidentiality
Executives come to us with questions they can’t yet raise inside the business. What you tell us goes no further.
Practices
AI advice is only as useful as its grasp of the business it’s applied to.
Retail
Origination, pricing and collections, where automated decisions reach customers directly.
Models that decide who gets credit, and the explanations regulators expect for them.
Global book building, structured products, munis and more
Pricing, documentation and surveillance across complex, data-heavy instruments.
Post Production & Visual Effects
Production pipelines, rights, and the use of generative tools on screen.
Including real-time data
Presenting complex information so it can be grasped at a glance, from data analysis to live dashboards a board can act on.
Including RAG retrieval and message routing
Avoiding the non-deterministic triple lock, making RAG retrieval meaningful, and routing each message to the model that should answer it.
The web is moving
Most of all, the web is moving: search is turning into answers, and more of your visitors are AI agents. We help you build for the web that’s coming, not the one that’s going.
We are a team of seasoned engineers, business leaders, and successful entrepreneurs. We’ve built pioneering global systems across multiple industries.
If you’re struggling with AI, either in practice or preparation, we offer best practice, healthy scepticism, and a route out of trouble, if needed.
Contact
Tell us a little about where you are with AI. Keep it general if you’d rather. We can go into detail once we’ve spoken.