The plan
useful first
Be useful first. Earn the right to predict.
We do not need the prediction maths working to be worth paying for. The first version saves clinicians time on paperwork. Only once that is in real use, and the data behind it holds up, does the prediction get switched on - and even then it runs silently at first.
- 01Working now
Be useful
Save clinicians time today
Taking notes from the conversation, transcribing, and keeping it all encrypted on the machine. Worth paying for on its own, before any prediction exists.
- 02Next thing we build
Start measuring
Write every promise down
Record what was promised after a hospital stay, and for long-term conditions, as something trackable. This builds the data we would need later, without claiming to predict anything yet.
- 03Silently at first
Turn on prediction
OLRHM + TRIS
Only build the ranking engine once people can agree what counts as a miss. Then run it where nobody can see it until the accuracy, workload, fairness and regulatory checks have all passed.
- 04After it is proven
Scale up
Across a whole hospital
Shared infrastructure and connections to hospital systems - once the core, the evidence, and the controls can survive someone checking them properly.
Where the money goes
fund the proof
Fund the proof, not the costume.
No amount is named here. The briefing sets out what each stage costs and what has to be true before the next one starts.
- 01
Foundation
The shared safety core, traceable records, and written requirements
- 02
Protection
Patent scope, and finding out whether regulators treat this as a medical device
- 03
Evidence
Getting experts to agree what counts as a miss, gathering data, independent checking
- 04
Continuity
More engineers, someone owning quality, and a second site to test at