How Park Street Surgery recovered 51 hours of staff time in 8 weeks

Park Street Surgery is a busy general practice that, like most, measured its mornings by the phones. The eight o'clock surge set the tone for the whole day — and for eight weeks the team ran a supervised pilot to see whether an AI receptionist could change that.

The challenge

The pattern was familiar and stubborn. Demand did not arrive evenly through the day; it arrived all at once. Every line lit up at eight, a queue built behind the engaged tone, and patients redialled, gave up, or turned up at the desk having already waited twice. The reception team was neither under-staffed nor under-skilled — it was simply outnumbered by a surge that no rota can absorb.

The cost was not only the queue. Mornings spent triaging the telephone meant less attention for the patients standing at the desk, more interrupted admin, and a steady, low-grade pressure that is hard to staff away. The practice wanted to know whether the routine, repeatable calls — repeat prescriptions, appointment requests, fit notes, simple enquiries — could be handled in full, instantly, without taking anything away from the patients who needed a person.

The pilot

Jackie was configured to match how Park Street already triaged, and writes structured notes straight into the practice's existing clinical system — no parallel inbox, no new record to learn. Information governance was signed off before a single call was taken. From go-live, Jackie answered on the first ring, in parallel, every line at once: no hold music, no engaged tone, no patient left waiting.

The boundary mattered as much as the capability. Jackie is an administrator, not a clinician: she does not diagnose, give medical advice, or decide how urgent a problem is. She listens for safety indicators throughout a call, and any patient can ask for a person at any point — Jackie then hands cleanly to the reception team. The pilot ran supervised, so the practice could watch every escalation and satisfy itself before widening Jackie's remit.

The results

Across the eight weeks, Jackie handled 81% of inbound calls, and completed 91.1% of those she handled end to end — the whole task done, with structured notes filed and admin completed, no human touch required. The remainder were either escalated to the team by design or passed to a person at the patient's request, exactly as intended.

In real terms that returned 51 hours of reception time over the eight weeks — capacity that had previously been spent answering, holding and re-triaging the morning queue. The surge did not disappear; it was simply answered.

“The phones used to define our mornings. Now the team starts the day on the patients in front of them.”

What it freed up

The headline figure is the recovered time, but the change the team felt was the shape of the morning. With the routine calls handled in full, reception could start the day on the patients in front of them rather than on the switchboard. Admin arrived finished rather than as a backlog to catch up on, and the human option was protected for the people who most needed it — shorter queues for everyone, not the removal of a person from the line.

Eight weeks is a short window, and Park Street treated it as a measured trial rather than a leap of faith: governance first, supervised go-live, the practice's own data as the scorecard. The result was straightforward enough to act on — the phones stopped defining the mornings, and the team got the mornings back.