Case Study · Clinical AI · NHS

NHS trust referral triage — from manual to AI-automated in 12 weeks

An NHS trust processing 2,400 referrals per month was missing 18-week pathway targets due to manual triage backlogs. We built a DSPT-compliant AI triage system that prioritised clinical urgency without replacing clinical judgement.

Sector NHS & Public Sector

Engagement Project Delivery

Timeline 12 weeks

Compliance DSPT · GDPR · CQC

The Challenge

Clinicians buried in paperwork, not patients

The trust's referral management team was processing every incoming referral manually — reading, categorising, and assigning clinical urgency across 7 specialty pathways. With 2,400 referrals per month and a growing backlog, the trust was at risk of breaching 18-week RTT pathway requirements. Clinicians were spending up to 40% of their administrative capacity on triage alone.

The Problem

Manual referral triage was creating an unsafe, unsustainable backlog

2,400 referrals/month processed entirely by hand

18-week RTT pathway breach risk due to triage backlog

Inconsistent urgency scoring across the team

Unstructured referral letters — variable quality, missing data

NHS procurement constraints limiting speed to deploy

Strict DSPT, GDPR, and CQC obligations on every patient data touchpoint

Our Approach

Augment the clinician's judgement — never replace it

DSPT-compliant deployment on NHS-approved Azure HSCN infrastructure

AI reads referral letters, extracts structured clinical data

Clinical urgency scoring across 7 pathways — with XAI reasoning

Mandatory HITL override — clinician reviews every AI score before routing

Red-flagging system — automatic escalation for urgent clinical indicators

Full CQC-compliant audit log on every triage decision

NHS Compliance Framework

Every design decision was made with NHS information governance at the centre

DSPT-compliant Azure HSCN deployment

NHS-approved AI governance framework

GDPR Data Processing Agreement in place

CQC-compliant audit trail on every decision

Mandatory Human-in-the-Loop on all outputs

Data never leaves NHS-approved boundary

Delivery Timeline

Discovery to production in 12 weeks — including IG sign-off

Navigating NHS procurement and information governance without slowing the delivery timeline requires experience. We've done it before.

Weeks 1–2 · Discovery

Clinical workflow mapping and IG assessment

Worked alongside the referral management team to map the full triage workflow across all 7 specialty pathways. Completed a full Information Governance assessment in parallel — establishing the DSPT obligations and GDPR requirements that would shape the entire architecture.

Weeks 3–5 · Architecture & Governance

DSPT-compliant infrastructure and model design

Provisioned the Azure HSCN environment with DSPT compliance controls. Designed the clinical NLP extraction pipeline, urgency scoring model architecture, and the XAI framework. IG sign-off obtained by end of Week 5 — earlier than the trust's standard 8-week process.

Weeks 6–9 · Build & Clinical Validation

AI development and clinician-in-the-loop testing

Built the referral extraction pipeline and urgency scoring model. Ran a 3-week clinical validation exercise with 6 clinicians reviewing AI outputs against their own scoring — used to calibrate the model and refine the HITL interface. Achieved 89% agreement rate before proceeding to deployment.

Weeks 10–12 · Deployment & Handover

Parallel run, go-live, and team training

2-week parallel run alongside manual process to validate outputs at scale. Go-live with all 2,400 monthly referrals. 2 days of team training. Full system documentation and runbooks. Trust now operates the system independently, supported by a 6-month SLA.

"What impressed us most was that they understood NHS information governance before we had to explain it. They designed the system around our obligations, not around what was technically easiest."

Head of Referral Management · NHS Trust (name withheld)

Results

Clinical capacity freed. Pathways protected.

62%

Reduction in administrative time per referral — 12 mins to 4.5 mins

2,400/mo

Referrals processed automatically, all 7 specialty pathways

0 breaches

18-week RTT pathway breaches in 3 months post-launch (vs. 14 prior quarter)

89%

Clinician agreement rate on AI urgency scoring — validated pre-launch

100%

DSPT and GDPR compliance — all patient data on NHS-approved infrastructure

12 wks

Discovery to production — including full IG sign-off

Technology

NHS-approved infrastructure throughout

☁️

Azure HSCN

NHS-compliant cloud

🏥

Azure Health Data

Clinical data services

🤖

Azure OpenAI

Clinical NLP (NHS data boundary)

🔍

Azure AI Search

Pathway knowledge base

🔐

Entra ID + PIM

NHS identity & access

📊

Power BI

Trust performance reporting

🛡️

Microsoft Purview

Data governance & audit

📋

XAI Engine

Explainability for every decision

NHS trust referral triage — from manual to AI-automated in 12 weeks

Clinicians buried in paperwork, not patients

Discovery to production in 12 weeks — including IG sign-off

Clinical capacity freed. Pathways protected.

NHS-approved infrastructure throughout

Related work

Working in NHS or public sector?