UHSussex AI assistant cuts NHS patient call waits from 30 minutes to under three

TL;DR: University Hospitals Sussex NHS Foundation Trust has cut patient call wait times from over 30 minutes to under three since deploying Netcall’s conversational AI virtual assistant, Digital Health reports. Call abandonment fell 75%, digital engagement reached 86%, and roughly 1,500 patient queries a day are now handled without booking-centre involvement. Did Not Attend rates are at 4%, around 200 appointments are rebooked or cancelled by self-service daily, and weekly validation campaigns to 4,000 patients have cut the waiting list by 12%.

What the deployment looks like

The trust paired Netcall’s AI assistant with the firm’s patient engagement portal (Patient Hub) and waiting-list validation tool. The assistant handles natural-language queries, answers common questions, supports self-service for tasks like appointment changes, and routes the rest to the right team. Call abandonment is now under 10%, and 36% of inbound calls are diverted from the booking centre entirely.

James Allan, associate director for planned care at UHSussex, told Digital Health: “We have a strategic commitment to embrace technology to improve the care of our patients. We are now able to deal with routine requests more effectively and quickly, giving our teams more time to focus on supporting patients when they do need to speak to someone directly.”

The trust is now operating a single phone number across all patient-facing services and is implementing a multi-language AI chatbot. In January, UHSussex selected Alcidion as preferred supplier for its forthcoming electronic patient record system.

Looking forward

Concrete deployment outcomes are scarce in NHS AI reporting; this one is unusually clean. The headline numbers — 90% wait reduction, 75% abandonment reduction, 12% list reduction from validation — are the kind of figures Trust boards can defend to commissioners and patient groups. They also land alongside this week’s Oxford Internet Institute paper showing that warmth-tuned chatbots become 30% less accurate. UHSussex’s deployment is patient-facing administrative work — appointments, redirection, validation — rather than clinical advice, which is the right scope given current model reliability evidence.

For NHS digital teams evaluating similar deployments, the questions worth asking Netcall and comparable vendors are: what fallback exists when the model misroutes a patient with an urgent clinical concern, what audit trails are kept, and what is the multi-language assistant’s accuracy benchmark — particularly for older patients and those with low digital confidence, where the retention of human-staff capacity remains the safety net.