The situation
Assessment reports were a throughput bottleneck, with KPI and SLA compliance below target on both the ADHD and ASD pathways.
The constraint
This was clinical documentation in a regulated, NHS-facing service. Any AI had to sit inside clinical governance, not around it.
What I did
- Identified and evaluated AI report-drafting tools, defining success criteria and validating the return on investment.
- Took the chosen tool through the AI governance framework I'd set up, rather than around it.
- Unblocked a disagreement about patient consent across clinical, legal and operational stakeholders.
- Led the rollout across both assessment pathways.
With hindsight
The patient-consent question took far longer than it should have, because one stakeholder's position shifted more than once, and each shift reopened a discussion several senior people had already spent time closing. I should have insisted on a shared decision log from the start: what was decided, why, and who agreed. People should be able to change their minds, but when the record is visible and signed off, a change has to be acknowledged as one and argued for. It would have saved days of senior time.
The outcome
Throughput and compliance improved together on both pathways, with clinicians spending around 60% less time writing reports.