How Park Street Surgery Recovered 51 Hours of Staff Time in 8 Weeks

Numbers are easy to put in a pitch deck. A 40 percent reduction in call volume. A 25 percent increase in capacity. A fraction of the cost of a new hire. Every vendor in the GP technology space has a version of these claims.

This post is about a specific practice, in a specific city, with a specific patient list and what happened when they deployed an AI voice receptionist and measured the results across eight weeks.

Park Street Surgery is an NHS GP practice in Liverpool with approximately 8,000 registered patients. In November 2025, they went live with Jackie, an AI voice receptionist built by Auxilis AI. This is what happened.

The Problem Park Street Were Trying to Solve

Park Street Surgery faced the same structural challenge that most practices with a busy patient list recognise immediately. Phone demand at 8am exceeded what the reception team could handle. Calls went unanswered. Patients waited in queues. Staff managed the pressure as best they could, but the administrative workload was crowding out other work outbound patient contact, care navigation, QOF activity.

The practice was not in crisis. But the gap between what the reception team could handle and what the patient list demanded was real, and it was not closing by itself.

Their decision to trial Jackie was driven by two practical questions: could an AI system actually hold a conversation well enough to replace a receptionist for routine inbound calls? And if it could, what would that mean for the team’s workload in practice?

The Setup

Jackie was integrated with Park Street’s existing telephony platform. No new phone number was issued to patients. No hardware was installed. The setup process took one to two weeks and involved configuring Jackie to follow Park Street’s specific protocols their sick note rules, their appointment booking eligibility criteria, their triage form requirements.

The practice chose to route a portion of inbound calls through Jackie initially, with the option to scale. The reception team remained fully staffed throughout the pilot. Jackie was not a replacement it was additional capacity sitting alongside the existing team.

What Jackie Handled

Across the eight-week pilot, Jackie handled over 1,200 patient calls. The requests covered the routine inbound range that any GP practice recognises.

• Appointment requests; Jackie collected the reason for the appointment, relevant symptoms, and preferred timing, and submitted a completed triage form directly into the practice’s system.
• Sick note and fit note requests; Jackie verified whether the patient’s current sick note had expired before processing a new request, applying the practice’s eligibility rules automatically. Patients whose notes had not expired were informed of the self-certification rule and sent an SMS link.
• Appointment cancellations; Jackie processed cancellations directly in EMIS, freeing up appointment slots in real time without requiring a receptionist to take the call.
• General admin queries; Jackie handled a range of routine questions, escalating to reception for anything that required human judgement or sensitive handling.
• Safety flagging; on calls where patients mentioned symptoms that warranted clinical escalation, Jackie identified the red flag indicators and advised the patient to contact 999 or offered to transfer them to reception immediately.

The Results

The 51 hours of recovered staff time represents the equivalent of approximately 0.25 FTE, one additional quarter of a full-time receptionist’s working time, generated without hiring anyone. For a practice paying a receptionist approximately £25,000 to £30,000 annually, that is £6,000 to £7,500 worth of capacity per year, returned to the team.

The 81 percent call absorption figure means that for every ten inbound calls that arrived during Jackie’s operating hours, eight were handled end-to-end by Jackie without a human receptionist being involved. The remaining two were transferred to reception, usually because the patient requested a human or the call involved a matter outside Jackie’s configured scope.

The 91.1 percent call completion rate means that nine in ten calls Jackie answered reached a complete resolution, a submitted form, a confirmed cancellation, a correctly applied eligibility rule, or an appropriate escalation. This is a meaningful figure because it reflects not just call volume but call quality.

What Surprised the Practice

Three things stood out to the Park Street team that were not necessarily anticipated before the pilot.

1. Patient acceptance was higher than expected

The practice anticipated some patient resistance to speaking with an AI system. In practice, the response was largely positive. Patient feedback, collected via SMS at the end of calls, noted the speed of answer, the politeness of the interaction, and notably, the absence of a queue. One patient feedback response described Jackie as ‘much more polite than the receptionist.’ Another said speaking to Jackie had ‘reduced my anxiety’ because they did not have to wait.

The key factor appears to be conversation quality. Jackie adapts in real time to what the patient says, asks contextually appropriate follow-up questions, and does not rush the interaction. Patients who expected a phone menu experienced a conversation instead.

1. The sick note eligibility logic created immediate value

One of the highest-volume request types at Park Street, as at most practices, is sick note and fit note requests. Jackie’s eligibility checking logic, configured to the practice’s specific rules, prevented unnecessary task creation from the first day of operation. Patients whose notes had not expired were correctly informed and sent an SMS with self-certification guidance. This alone removed a category of avoidable admin work from the reception team’s daily load.

Zero outages across eight weeks

The practice experienced no service interruptions during the pilot. Jackie was available throughout every operating hour it was scheduled to cover. For a practice evaluating a technology that sits at the front of their patient access pathway, reliability is not a minor consideration, it is the baseline requirement. Zero unplanned outages across eight weeks met that requirement.

What Did Not Change

The reception team at Park Street remained fully employed and fully active throughout the pilot. Jackie handled the high-volume, routine end of the inbound call range. The reception team handled complex queries, clinical escalations, sensitive conversations, and relationship-based interactions, the work that benefits most from a human presence.

This is the deployment model that makes operational sense. AI voice does not replace the judgement, empathy, or contextual knowledge that experienced reception staff bring to difficult calls. It handles the volume so that the team can focus on the work that genuinely requires them.

What Comes Next for Practices Considering a Pilot

Park Street’s experience represents a four-week free pilot model. Jackie is configured to the practice’s specific protocols during a one to two week setup period, then goes live with no commitment to a paid contract during the pilot phase.

The metrics measured during the pilot are call handling rate, time recovered, information quality, safety event rate, and patient and staff satisfaction. At the end of four weeks, the practice has live data from its own patient population, not aggregate benchmarks from somewhere else.

For practices evaluating AI voice receptionists, that live data is the most honest answer to the question of whether it works in your specific context.

Frequently Asked Questions

Did Park Street Surgery reduce its reception team after deploying Jackie?

No. The reception team remained fully employed throughout the pilot and beyond. Jackie added capacity alongside the existing team, not instead of it. Staff were redeployed from routine call handling toward care navigation, QOF activity, and outbound patient contact.

How long did it take to set Jackie up at Park Street?

Setup took one to two weeks. This covered telephony integration, protocol configuration, and compliance checks. No new hardware was required and no patient communication about the change was needed, patients continued to call the same number.

What happens if Jackie cannot handle a call?

If a call falls outside Jackie’s configured scope, or if the patient requests a human at any point, the call is transferred to reception immediately. Patients are always informed they can speak to a human. Jackie does not dead-end calls.

How is patient data handled during calls?

Jackie is GDPR compliant and the practice remains the data controller throughout. Calls are processed within the NHS Data Security and Protection Toolkit framework. Two-factor patient verification (phone number and date of birth, validated against NHS PDS) applies to all calls. Calls where identification fails are flagged for human review and are not charged.

See how Jackie performs in your practice with a free 4-week pilot. No contract, no setup cost, no risk. Book your demo at auxilis.ai.

Book a free 20-minute demo at auxilis.ai