-
Healthcare
-

AI Health Bus Puts Berkshire's Productivity Target to the Test

By
Distilled Post Editorial Team

A bus parked outside a community hospital usually means a health check or a vaccination. At Prospect Park and West Berkshire Community Hospital, Berkshire Healthcare NHS Foundation Trust has used its Health Bus for something less familiar. Staff step on, sit at a screen and watch Microsoft Copilot turn a rambling email into a tidy one or condense a long report into a page. A further session is planned at Wokingham Hospital. The scene is modest, and it matters because of the number hanging over it.

The trust is working to an estimated £15 million planning deficit for 2026/27 and a requirement to raise productivity by at least two per cent a year. Those figures explain the bus. Administrative time is among the few resources a trust can release without closing a ward or a clinic, and generative AI promises to return some of it. Teaching staff in person, rather than circulating a licence and a guidance note, is a sensible start, because NHS digital programmes have often failed on adoption before technical limits are reached.

Arithmetic still needs care. Minutes saved on drafting are real, but they are spread across hundreds of people in fragments too small to redeploy. A clinician who saves ten minutes a day does not create a vacancy that can come out of the budget or an extra appointment that appears on a rota. Unless someone redesigns the work around the released time, the gain dissolves into a slightly easier working day. That has value for morale and retention, which cost the NHS dearly, yet it does not close a deficit. The two per cent target will be measured in activity and expenditure.

The trust's own roadmap points to where larger gains sit. Ambient voice software that drafts notes during a consultation and automated triage of referrals reach the work that consumes clinical capacity directly, with summaries feeding the patient record. They also carry real risk. A system that mishears a medication or misjudges the urgency of a referral creates a patient safety event. That demands clinical safety assurance, data protection impact assessments, supplier due diligence and clear accountability when an automated summary is wrong. National guidance on ambient voice has begun to set expectations, but much of the assurance burden still falls on individual trusts, and few have spare capacity for it.

Procurement deserves equal scrutiny. Copilot is licensed per user, and the cost of broad rollout is easy to underestimate when the savings are hard to demonstrate. A trust that buys widely before measuring anything may add to the deficit it hopes to reduce. Narrower deployments would produce better evidence, provided each has a measured baseline and a named owner who decides what the released time is for.

There is also a workforce question that training sessions cannot answer. Staff who are told that AI will free their time may reasonably ask whether that time will be returned to them or handed to the waiting list, and whether their roles will look different in three years. Unions and professional bodies are watching how automation is introduced, and leaders who treat it as a pure efficiency measure will meet resistance. Candour about intent is cheaper than repairing trust afterwards.

Berkshire also illustrates a wider problem for policymakers. Government has made digital and AI adoption central to its reform agenda, and productivity targets assume that technology will carry much of the load. Trusts are left to turn that assumption into local plans using their own money and staff. Where leadership is stable and the digital team is strong, it may work. Elsewhere, sessions will be delivered, enthusiasm will be recorded and the savings will fail to arrive.

Patients will judge the programme by waiting times. If referral triage speeds up and clinicians spend less time typing, the benefit will appear in shorter queues and more attentive consultations. If it does not, the bus will have been a pleasant afternoon. Berkshire deserves credit for starting with its workforce, since staff who distrust a tool quietly stop using it. The harder test comes next, when the trust must show which hours came back and where they went.

‍