The Real User Feedback: Praise From Staff, Frustration From Callers

When a surgery in Lincolnshire replaced part of its phone line with an AI receptionist named EMMA, the result was a story of two very different experiences: clinicians praised the information quality, while patients gave up and drove to the practice in person. That divide, captured in a BBC News investigation, is the only direct window into what real users think of the EMMA AI receptionist service — and it reveals a tension every business evaluating AI phone tools should understand.

On the clinical side, the feedback is genuinely positive. Dr. Satpal Shekhawat, a GP at Kirton Lindsey & Scotter Surgery, which serves roughly 10,500 patients, said: "The quality of the information I'm receiving [from Emma] is good, the relevant information to help me decide what clinically needs doing, is good." He framed EMMA as a support tool — one designed to free up human receptionists for other duties, not replace them. Under the system, most patients are contacted within two or three hours.

Patients, however, tell a very different story. Their complaints center on the AI failing at basic tasks: understanding dates of birth and describing health concerns. Those failures led to abandoned calls and in-person visits as workarounds.

• Sandra Dodsworth, after EMMA failed to recognize her date of birth, said: "I gave up, I've driven down. I'm not happy... I'd rather speak to a person who can understand me."
• Another patient, Alex, described being "in a lot of pain" and "having to scream at a robot," adding, "I don't think healthcare is the right place for it."
• Patients reported emotional distress and communication barriers, with a consistent preference for human interaction in a high-stakes setting.

The contrast matters because it maps exactly onto what broader research shows about AI receptionists. A comparison of AI versus human receptionists finds AI excels at high call volumes and 24/7 availability, while humans remain essential for emotionally sensitive or complex interactions. The hybrid model — AI handling initial response, humans handling escalations — is where both efficiency and caller experience hold up.

There's also a transparency dimension. A Twilio survey found 54% of consumers want to know when they're interacting with AI, and 84% want control over personalization settings. Services like CallMyLeads build disclosure in from the start — callers always know they're talking to AI, and every caller can reach a human, text, or book online. The EMMA case suggests that escape hatch isn't optional; it's what stands between a frustrated caller and a lost one.

Regional NHS officials say other practices are watching the trial with interest. Whether EMMA's refinements close the gap between staff satisfaction and caller frustration will determine whether this tension resolves — or defines the service.

Why the Market Loves AI Receptionists (And Where It Falls Short)

The numbers tell one story; the patient at Kirton Lindsey tells another. She drove to the surgery in person after an AI receptionist failed to recognize her date of birth — and that gap between market optimism and real-world experience is exactly what businesses need to understand before choosing an AI vendor.

The AI receptionist market is booming. Market research values it at $2.7 billion in 2025, projected to reach $31.3 billion by 2035 — a 27.8% annual growth rate. North America accounts for over 42% of that market, and small and medium businesses make up the largest share of buyers, drawn by the cost of hiring human staff.

Consumers are warming up, too. According to Salesforce survey data, 39% of U.S. consumers are comfortable letting AI agents schedule appointments, and 44% would use one as a personal assistant — rising to 70% among Gen Z. Those are the kinds of numbers that make an always-on, never-voicemail answering system look like an obvious win for lead-driven businesses.

But the BBC's investigation of EMMA at a UK surgery serving roughly 10,500 patients shows where the optimism glosses over reality. Clinicians praised the system — Dr. Satpal Shekhawat said the information EMMA collected was "good" and clinically relevant. Patients told a different story: abandoned calls, emotional distress, and one man in pain saying he was "screaming at a robot" instead of getting an appointment.

The pattern is predictable, and it maps to three fault lines:

• Natural language limits surface fastest in high-stakes settings, where callers use emotional, unstructured speech rather than clean booking requests.
• Healthcare and similar contexts amplify frustration because callers are anxious before the conversation even starts.
• Failed inputs cascade — one unrecognized date of birth turns into a lost call and a lost patient.

Industry analysis supports a hybrid answer: AI excels at high call volumes, 24/7 availability, and routine intake, while humans handle emotionally sensitive escalations. A comparison of AI versus human receptionists frames it simply — AI manages the first response and routine tasks, people step in when empathy matters.

That's the standard any business should hold vendors to, including how they handle disclosure — a Twilio survey found 54% of consumers want to know when they're talking to AI. CallMyLeads builds that transparency in from the start: callers always know they're speaking with AI, and every caller can reach a human, text, or book online instead of fighting a system that can't understand them.

The Fix: Support Humans, Don't Replace Them

The BBC investigation at Kirton Lindsey & Scotter Surgery revealed a clear pattern: patients abandoned calls when the AI couldn't understand basic inputs like dates of birth, while clinicians praised the quality of information EMMA collected for clinical decisions. The practice's own leadership emphasized the system was designed to support receptionists, not replace them — freeing staff for other duties while handling routine information gathering. This hybrid intent aligns with what consumers actually want: a Twilio survey found 54% of people want to know when they're talking to AI, and 84% want control over personalization settings.

• AI handles initial response, FAQs, and structured data collection
• Human staff take escalations, complex questions, and emotionally sensitive situations
• Clear AI disclosure at the start of every interaction
• Instant handoff options — text, callback, or online booking — when the caller prefers a person

When patients like Sandra Dodsworth drove to the surgery after failing to get her date of birth recognized, and Alex described "screaming at a robot" while in pain, they weren't rejecting automation — they were rejecting a dead end. The fix isn't better AI alone; it's designing for the handoff. CallMyLeads builds this into every deployment: inbound calls answered 24/7 with honest AI disclosure, qualification rules set by the client, and immediate routing to human staff, text follow-up, or self-serve booking the moment the AI hits its limit. The lead that gets a reply first usually wins — but only if the reply actually works for the person on the other end.

What to Check Before You Choose an AI Receptionist

The EMMA story at Kirton Lindsey & Scotter Surgery offers a clear lesson: an AI receptionist succeeds or fails on a handful of details you can test before signing up. Patients abandoned calls when the AI couldn't recognize a date of birth, while clinicians praised the quality of information it collected — a split you can avoid by evaluating the right things upfront (BBC News).

Start with understanding. Ask any provider for a live demo using the exact inputs your callers actually use — names, addresses, appointment reasons, dates of birth. The BBC investigation found that failure to understand basic inputs was the single biggest driver of patient frustration, with one patient driving to the surgery in person after the AI couldn't recognize her date of birth. If the system stumbles on your most common caller phrases during a demo, it will stumble in production.

Confirm a human escape hatch. One patient in the BBC report described "screaming at a robot" while in pain, and the practice itself stressed that EMMA was meant to support, not replace, human receptionists. A hybrid approach — AI handling routine calls, humans handling escalations — is widely recommended as the model that balances efficiency with caller experience. Ask exactly when and how calls route to your team, and what happens after hours.

Verify AI disclosure upfront. A Twilio survey found 54% of consumers want to know when they're talking to AI rather than a human, and 84% want control over personalization settings. A provider that hides the AI behind a human-sounding voice creates the trust gap that made EMMA's patients feel deceived. Disclosure should be a feature, not an afterthought.

Match the tool to the use case. EMMA's healthcare deployment shows the risk of deploying AI in emotionally sensitive contexts — one patient said plainly, "I don't think healthcare is the right place for it." AI receptionists are strongest where speed and routine matter: lead response, missed-call recovery, and appointment booking. That's the same principle behind services like CallMyLeads, which focuses AI on fast lead response and booking rather than sensitive conversations.

Before you commit, run this checklist:

• Test the AI with your real callers' most common phrases and inputs
• Confirm a clear escalation path to a human on every call
• Verify the AI discloses itself upfront and offers alternative channels
• Ask which conversations route to humans — sensitive topics should never sit with the AI
• Check what happens when the AI fails: does the caller get a callback, or a dead end?

The market is growing fast — from $2.7 billion in 2025 toward a projected $31.3 billion by 2035 — which means more providers and more variation in quality. The businesses that win with AI receptionists aren't the ones that pick the flashiest tool. They're the ones that test understanding, demand human escalation, and keep the AI pointed at what it does best: answering fast so no lead, and no caller, goes unheard.

Getting the Speed Benefit Without the Backlash

The lesson from EMMA's rollout isn't that AI receptionists fail — it's that speed only wins when callers never feel trapped. At Kirton Lindsey & Scotter Surgery, the same system that frustrated patients still got most of them contacted within two or three hours, according to BBC reporting. The problem wasn't the response time. It was what happened in the moments before that response.

That distinction matters for any business evaluating lead response tools. The core value proposition of an AI receptionist is real: answering every call 24/7/365, capturing leads in seconds, and never letting a form submission sit overnight. Market data shows the category is growing fast — the global AI receptionist market sits at USD 2.7 billion in 2025 and is projected to reach USD 31.3 billion by 2035. Businesses are adopting because the speed benefit is genuine.

But EMMA's patient backlash reveals the failure mode to avoid. When callers can't get the AI to understand basic inputs, they abandon the call and drive to the office in person — the exact opposite of what the technology is supposed to achieve. The fix isn't better AI alone. It's design choices that respect the caller.

Based on the research, here's what separates a fast system from a frustrating one:

• Disclose the AI upfront. A Twilio survey found 54% of consumers want to know when they're talking to AI rather than a human, per market research. Hiding it guarantees backlash when discovered.
• Offer an immediate human path. Patients at the surgery wanted "a person who can understand me" — the system needs an escape hatch before frustration builds.
• Give callers alternative channels. Text and online booking routes let people who struggle with voice AI complete their goal anyway.

This is how CallMyLeads approaches it. Every caller knows they're talking to AI, and every caller can reach a human, switch to text, or book online — disclosure is treated as a feature, not something to hide. The speed stays: inbound calls answered around the clock, first replies in seconds, nothing going to voicemail.

The hybrid model is where the research lands too. Industry analysis recommends AI handling initial response and routine tasks while humans handle escalations and emotionally sensitive interactions. That balance captures the efficiency without the "screaming at a robot" experience patients described.

Done right, nobody gives up and drives down to your office. They get an answer in seconds, a clear next step, and a way out if the AI isn't working for them.