Operational and documentation work. Clinicians decide.

There is a great deal of valuable, low-risk AI work in healthcare that is not diagnosis. Most of it involves giving clinicians back the hours they currently spend on documentation and administration.

01 / What is hard here

The problem underneath the problem.

Clinical decision support carries a regulatory and liability burden that is entirely appropriate and that most organisations asking for it have not costed. It is a medical device question before it is a machine learning question.

Meanwhile the operational layer — scheduling, coding, documentation, capacity planning, prior authorisation — is where staff time is genuinely being lost, and it carries a fraction of the risk. That is where we concentrate.

02 / What we build

Work we take on in healthcare.

Clinical documentation support: Structuring consultation notes and correspondence, with the clinician reviewing and signing everything before it enters the record.
Coding and billing assistance: Suggesting procedure and diagnosis codes from documentation with confidence scores, reducing both rework and denied claims.
Operational forecasting: Admission volume, length of stay, theatre utilisation and staffing demand — genuinely useful and carrying no clinical risk.
Triage and routing support: Prioritising referrals, messages and administrative queues so that limited clinical attention reaches the right place sooner.
Records retrieval: Search across historical notes and correspondence so a clinician can find the relevant history in seconds rather than minutes.

03 / Constraints

What shapes the build.

These are design constraints agreed at the start, not compliance work bolted on before launch. Retrofitting them is expensive and usually incomplete.

Clinical safety governance: Anything touching clinical decisions goes through the appropriate safety and regulatory process. We will tell you when your request crosses that line.
Patient data handling: Minimum necessary access, strong audit trails, defined retention, and an explicit decision about whether data may reach any external model provider.
On-premise where required: Self-hosted models are a standard option here, not an exception, and frequently the only acceptable architecture.
Clinician review: Generated clinical text is a draft for review, never a record. The workflow enforces this rather than relying on habit.

04 / What we advise against

Where we would say no.

We would advise against autonomous diagnostic or triage decisions without the appropriate regulatory route, and against any deployment that puts generated text into a patient record without clinician sign-off.

We would rather lose the work than build something that creates a liability for you or a harm for the people on the other side of it. If that rules out what you had in mind, it is better established now than at delivery.

05 / Other sectors

Adjacent work.

Different domains, the same delivery discipline — a measured baseline, a real evaluation set, and monitoring that catches decay early.

Next step

Working in healthcare?

Tell us which process is slow, expensive or error-prone. The diagnostic will tell you whether this is a problem worth solving with machine learning — including if it is not.

Request a diagnostic All industries