01
The decision nobody structures
Organisations have a process for procurement, for hiring and for clinical governance. Deciding whether to build something is usually left to conviction and a business case.
First edition
Before You Build
The de-risking playbook
Twenty-three pages on how to decide whether a health AI or technology investment is worth building, before the budget is committed and the answer gets expensive.
No sales pitch, no case for hiring us, and nothing you need a technical background to follow.
The central idea
Most of the value is decided after the prediction.
Technology creates value through a chain, and a break anywhere along it means nothing changes. A model can be accurate and the programme can still fail, because nobody decided what should happen once it knows something, or because the person who would have to act never saw it.
Need
Something is wrong, or could be better
Prediction
The system spots it, or predicts it
Decision
Someone works out what to do about it
Action
Someone does something differently
Outcome
The result you wanted in the first place
What is inside
Eighteen chapters, and the tools to run it yourself.
The ground it covers, in summary. It includes the diagnostic, the assumption register and the evidence ladder in full, so a team can work through the method without us.
02
How technology creates value
Need, prediction, decision, action, outcome. Most investment goes into the second step. Most of the value is decided in the third and fourth.
03
Why good organisations get this wrong
The biases that make a weak build decision feel like a confident one, including why stopping feels more expensive than continuing.
04
The five uncertainties to resolve
Need and adoption, workflow and accountability, action logic, commercial viability, and regulation and governance.
05
Testing assumptions without a production system
How to get real answers from real people before there is anything to deploy, and why that is cheaper than the alternative.
06
Validation rather than confirmation
How to design a test that could actually tell you no, rather than one that gathers support for what you already decided.
07
Regulation and governance questions
Whether it is a medical device, how personal data would be used, and which frameworks you should be building towards from the start.
08
The Before You Build diagnostic
The questions to work through, and what a weak answer to each one tells you.
09
The assumption register and the evidence ladder
A way to write down what you are betting on, and a way to judge how strong the evidence behind each bet actually is.
10
Go ahead, rework, or stop
A scorecard for the decision itself, including the case where the honest answer is to buy or partner rather than build.
11
Examples
The method applied end to end, then four shorter ones. All drawn from real work, with details changed and combined so no client is identifiable.
12
What a decision-ready pack contains
What you should have in your hands at the end, and what makes it defensible to a board or an investor.
Would rather not fill in a form?
The Before You Build page explains the method, free and with nothing in the way. The playbook goes further: the diagnostic, the assumption register, the evidence ladder and worked examples, in a document made to be shared with a board.
Read the method