AI Voice Agents and No-Show Reduction: What Health Systems Are Actually Seeing

Search “AI voice agents no-show reduction” and you’ll land on dozens of vendor blogs promising dramatic drops in missed appointments. Some of that is real. Some of it traces back to a single unnamed clinic with no control group. This piece sticks to named US and Canada health system deployments, peer-reviewed trials, and published data, and says plainly where voice-agent evidence is still thin compared with the older phone-reminder research.

Quick Facts

• Tampa General Hospital’s AI voice agent “Amy,” built with Hyro, drove a 21% increase in scheduled appointments and cut call abandonment 56%, from 34% to 14.9%, within two weeks of launch.
• A Hamilton Health Sciences-backed pilot at an Ontario clinic went from 70-80% of calls hitting voicemail to 100% of calls answered, half of them handled entirely by an AI voice agent.
• The strongest no-show-specific evidence still comes from older technology: Penn Medicine’s randomized trial of an automated IVR call plus text reminder cut no-shows from 11.3% to 9.6% across more than 244,000 patients.
• WellSpan Health’s voice agent “Ana” is expanding in 2026 into outreach to patients who miss imaging appointments, one of the first named US health systems pointing a voice agent directly at the no-show problem.

From IVR to Voice Agents: What’s Actually Different

The term “AI phone reminder” covers a wide range of technology, and that range matters when reading results. At the simple end is interactive voice response (IVR): a prerecorded message that plays and disconnects. One step up is predictive-model-driven calling, where a machine learning model flags which patients are statistically likely to miss their appointment and routes only those patients to a phone call, live or automated. At the newest end are conversational voice agents, built on large language models, that can hold a real two-way exchange, answer a question, confirm attendance, or reschedule on the spot without a human on the line. Most of the rigorous, controlled no-show research was done on the first two categories, before conversational agents existed. Here’s what’s actually running in US and Canada health systems right now.

AI Voice Agents in US and Canada Health Systems: What’s Actually Running in 2025-2026

Adoption of true conversational AI voice agents is real and moving fast across US and Canada health systems. What’s published so far is mostly about call handling and access, not an isolated no-show percentage, which is worth knowing before you compare a vendor’s promised number to your own clinic.

Tampa General Hospital: A Voice Agent Named Amy

Tampa General Hospital launched an AI voice agent called Amy, built with Hyro, to handle scheduling and call routing. Within two weeks of launch, appointments scheduled through its experience center rose 21%, daily call abandonment fell 56% (from 34% to 14.9%), and average wait times dropped 58% (from 6.2 to 2.4 minutes), according to Scott Arnold, the hospital’s chief digital and innovation officer.

Catholic Health: Voice AI Cuts Hold Times, Not Yet No-Shows

Catholic Health in New York deployed voice AI from Notable Health for its patient access and help desk lines. Call containment reached 54% on day one and rose to 64%, against a target of roughly 600,000 annual inbound calls, saving an estimated $60,000 in the first two months. The published results cover call handling and cost, not no-show rates.

WellSpan Health: The Clearest Signal Yet on No-Shows

WellSpan Health’s voice agent Ana, built with Hippocratic AI, already handles more than 160,000 patient calls a month. In its 2026 expansion, WellSpan named its next target directly: proactive outreach to patients who have missed imaging appointments. That makes it the clearest named example of a health system pointing a voice agent at the no-show problem specifically, though results aren’t published yet.

A Canadian Example: A Nobleton, Ontario Clinic

In a pilot linked to Hamilton Health Sciences, a family clinic in Nobleton, Ontario adopted an AI phone agent from Strello Health that answers calls, books appointments, and sends reminders. Before the rollout, 70-80% of calls went to voicemail; afterward, 100% of calls were answered, with half handled entirely by the AI. “The system was well received by my patients, so we expanded it,” said Dr. Eric Da Silva, one of the clinic’s physicians.

The pattern holds across other named deployments. MUSC Health’s voice agent Emily now handles 25% of all patient calls, having passed 2.2 million calls since its 2024 launch. Sutter Health rolled out 24/7 scheduling, refill, and billing support via chat, voice, and text across its 25 hospitals and 200-plus clinics, though it hasn’t published outcome numbers yet. Intermountain Health automated 44% of repetitive calls and cut call abandonment by 85%. None of these have published a no-show percentage tied specifically to the voice agent, which is the honest state of the evidence right now: real adoption, real operational gains, and no-show-specific numbers still to come.

Quick-Reference: AI Phone Reminder and No-Show Studies at a Glance

The Older Evidence: What IVR and Predictive Models Showed Before Voice Agents

Penn Medicine’s NEJM Catalyst Trial: Adding a Call to a Text

The most rigorous recent evidence comes from a randomized trial run by Penn Medicine’s Patient Access team with its Center for Health Care Transformation and Innovation. Patients who received an automated IVR call in addition to a text reminder had a no-show rate 1.7 percentage points lower (down to 9.6% from 11.3%), and appointment completions 1.9 percentage points higher (up to 77.8% from 75.9%), compared with text reminders alone. The effect was not evenly distributed. The intervention had the greatest effect among patients in the highest no-show risk quartile, and it increased completion rates the most among Black patients, narrowing a preexisting equity gap. After the trial, the health system rolled the program out at scale, and follow-up data from over 244,000 high-risk patients over six months showed the improvement held.

Changi General Hospital, Singapore: A Model, Not a Voice Agent

One older example worth a brief mention: radiologists at Changi General Hospital in Singapore built a predictive model to flag MRI patients at highest risk of missing their scan, then had human technologists call that shortlist. The no-show rate for contacted patients fell from a 19.3% pre-implementation baseline to 15.9% (a 17.2% improvement, p<0.0001), while patients the team couldn’t reach by phone still no-showed at 40.3%. The “AI” here is the targeting model, not the phone call itself, which is the opposite setup from the voice-agent deployments above.

Massachusetts General Hospital: Targeted Calls, Not Blanket Calls

A randomized trial at MGH’s Internal Medicine Associates clinic, which schedules 80,000 appointments annually with an average 7% no-show rate, tested whether adding a live coordinator call to patients flagged as high-risk by a prediction model, on top of the standard automated call, would help. It did: the no-show rate in the intervention arm was 22.8%, a significant absolute risk reduction of 6.4 percentage points compared with automated-only reminders for that same high-risk group. Note the baseline here is much higher than the clinic’s overall 7%, because this trial deliberately isolated the hardest-to-reach patients, exactly the group where targeted outreach shows up most clearly.

Where the AI Phone Reminders and No-Show Reduction Evidence Gets Murkier

When Adding a Human Call Didn’t Move the Needle

Not every intervention works, and that matters for anyone building a business case. A quality improvement project at a 359-bed community hospital’s infectious disease clinic added staff-performed phone calls on top of existing automated reminders and found the no-show rate barely moved, from 7.93% to 6.54%, a change that was not statistically significant. A follow-up at the same clinic tried shifting the reminder call from one day before the appointment to two or three days before, and the result went the wrong direction: the no-show rate actually increased by 4.49 percentage points after the earlier timing was introduced. Timing and channel choices that work well in one clinic don’t automatically transfer to another.

Live Contact Still Beats a Message, When You Can Get It

A continuous quality improvement study of 250 primary care patients with depression found a stark gap by delivery method: live reminders had a 3% no-show rate, compared with 24% for message or voicemail reminders, and 39% when the call wasn’t answered at all. That pattern echoes older research: a randomized trial of 13,505 outpatient appointments found a lower no-show rate with a clinic staff reminder compared with an automated appointment reminder. This is the core tension AI phone systems try to resolve: automation scales, but a real conversation still tends to outperform a recording, which is exactly why conversational voice AI is being pitched as a middle path, even though independent trial data on that specific format remains limited.

The Systematic Review Picture

Zooming out from single-site studies, a rapid systematic review of predictive model-based interventions found high certainty evidence that predictive model-based text message reminders reduced no-shows, and phone call reminders and patient navigator calls were probably effective, though the number of available studies was small. An earlier, broader Cochrane review of mobile phone messaging reminders similarly found low to moderate quality evidence that reminders improve appointment attendance, a more cautious rating than most marketing pages imply. Practices already using automation elsewhere in the front office, as covered in our guide to AI patient intake for wellness clinics, tend to treat reminder calls as one layer of a broader access strategy rather than a standalone fix.

Vendor Case Studies on AI Phone Reminders: What’s Verifiable and What Isn’t

Named, official case studies exist and are worth citing, with caveats. Health Catalyst’s own published case study on Memorial Hospital at Gulfport describes an analytics-driven no-show reduction program that increased revenue by roughly $1 million annually, a real, named, single-site figure, though the exact percentage reduction in no-shows isn’t broken out in the public summary. That’s a meaningfully different tier of evidence than the many blog posts circulating specific, unverified figures tied to unnamed or loosely sourced deployments. When a statistic can’t be traced to a named organization or a published methodology, treat it as a plausible pattern, not a guaranteed outcome for your own clinic. This is the same caution we apply in our broader look at AI in healthcare operations ROI, where payback claims vary widely by how a vendor defines “savings.”

Practical Takeaways for Clinics Evaluating AI Phone Reminders and No-Show Tools

• Target, don’t blast. Every strong result in this article came from routing calls to patients flagged as high-risk, not calling everyone the same way.
• Expect single-digit to low-teens percentage point gains, not a miracle. The best randomized trials show improvements in the 1.7 to 6.4 percentage point range, not the dramatic swings some vendor pages advertise.
• Watch what happens to patients you can’t reach. Changi General’s data shows unreachable patients still no-show at very high rates, so a phone-based strategy needs a backup channel.
• Check for equity effects, both directions. Penn Medicine’s trial narrowed a racial gap in completion rates; any new system should be monitored to confirm it isn’t quietly widening one instead.
• Ask vendors for the study, not the stat. If a case study doesn’t name the organization or link a methodology, treat the number as a marketing anecdote.

For teams weighing whether a full conversational assistant is worth the investment beyond simple reminder calls, our AI digital assistants in healthcare ROI case studies piece walks through where those tools have and haven’t paid off.

FAQ

Do AI phone reminders actually reduce no-shows more than a simple text message?

The best trial evidence says yes, but the gain is modest, not dramatic. Penn Medicine’s randomized study found adding an automated IVR call on top of SMS reminders lowered no-shows from 11.3% to 9.6% and raised completion rates from 75.9% to 77.8%, sustained across more than 244,000 patients.

How much can hospitals expect their no-show rate to drop with AI phone reminders?

It depends heavily on how targeted the calls are. Changi General Hospital’s AI-targeted phone reminders for MRI patients cut no-shows by 17.2% relative to baseline, while Massachusetts General Hospital’s targeted calls to high-risk patients produced an absolute drop of 6.4 percentage points, both far more than untargeted, blanket calling.

Are vendor case studies claiming 40-50% no-show reduction accurate?

Some individual clinics do report reductions in that range, but these single-site, self-reported figures are a different tier of evidence than peer-reviewed trials and often lack a control group. Treat any specific percentage from a company blog as a data point, not a guarantee, unless it points to a named, published study.

Does AI phone reminder technology help or worsen equity gaps in appointment attendance?

In the strongest studies available, it helped. Penn Medicine’s trial found the intervention increased appointment completion rates the most among Black patients, reducing a preexisting equity gap, and a separate safety-net system trial found predictive-model-driven outreach lowered no-shows without worsening racial or ethnic disparities.

Do real AI voice agent deployments show a measured drop in no-shows?

Not yet published for most. US and Canada health systems are deploying conversational AI voice agents fast, and named results so far cover call handling and access, such as Tampa General’s 21% jump in scheduled appointments and a Nobleton, Ontario clinic going from mostly-voicemail to 100% of calls answered. WellSpan Health is the clearest signal of movement toward the no-show use case specifically, expanding its voice agent into outreach for missed imaging appointments in 2026, but no health system has yet published a no-show percentage tied directly to a voice agent.