Healthcare

Intake triage across four clinics

Four sites, four different intake habits, and a shared referral inbox that ran two to three days behind. We built classification and routing on top of the record system already in place, under clinical governance review throughout, then trained the intake teams to run it themselves.

Discuss something similar All case studies

Sector: Healthcare
Engagement: AI Automation, then AI Enablement
Duration: 8 weeks, then 2 weeks
Constraint: UK GDPR, clinical governance sign-off
Stack: Existing patient record system, client-hosted

annualised operating cost removed: £1.7M
referral acknowledgement: 2–3 days → 20 min
on one written triage standard: 4 sites
urgent referrals missed since cutover: 0

The problem

What was actually wrong.

Referrals arrived as email, fax and portal submissions across four sites, each of which had evolved its own triage conventions over a decade. The same referral could be categorised three different ways depending on where it landed.

The shared inbox ran two to three days behind, and urgent referrals were being found late rather than never — which is the failure mode that actually worries a clinical director.

Nothing could be sent to a third-party processor without a completed data protection impact assessment, and clinical governance had to sign off the routing logic itself, not just the security posture.

What we did

The approach, in the order it happened.

Agree the triage rules before automating them

The first fortnight produced something the group had never had: one written triage standard across the four sites. Automating four inconsistent processes would have preserved the inconsistency.

Classify and route, do not diagnose

The system sorts referrals by urgency band and speciality and routes them. It makes no clinical judgement, and the boundary was written into the scope so it could not drift during the build.

Urgent cases escalate on low confidence

Anything the classifier is unsure of goes to a human immediately rather than into a queue. The asymmetry is deliberate: a false escalation costs minutes, a missed urgent referral does not.

Enablement so the clinic teams own it

A two-week follow-on trained the intake coordinators at each site to read the confidence flags, correct misroutes, and feed those corrections back as training signal.

Outcome

What changed. Measured against the baseline agreed at scoping, over at least a full quarter of production running.

Results

Roughly £1.7M of annualised operating cost removed across the four sites, measured against the baseline agreed at scoping.
Referral acknowledgement moved from two to three days to under twenty minutes.
All four sites now run one written triage standard, which turned out to be as valuable as the automation itself.
No urgent referral has been missed since cutover, and every routing decision is traceable for governance review.

“The part I did not expect was being made to write the triage standard down. We should have done it years ago.”

Clinical Director

The engagement

How this was delivered. The services this project ran through, and what each one covers.

AI Automation

4–8 weeks

The right path when the bottleneck is already visible and you mainly need the solution scoped, built, and implemented. Not automation for its own sake — removing drag so the work runs better.

Explore AI Automation

AI Enablement

1–3 weeks

The right path when the tools are there but the habits, confidence, and role-specific usage aren’t. We help people use AI in ways that are practical, useful, and grounded in real work.

Explore AI Enablement

More work