How does AI patient scheduling work?

What the agent does on a scheduling call

Your best schedulers run these calls from memory. The agent runs them from your rules.

• Finds the patient by name and date of birth.
• Works out the visit type from what the patient says, in their own words.
• Checks provider availability for the locations your rules allow.
• Checks insurance where your EHR exposes it, such as Epic's real-time insurance check.
• Books, reschedules or cancels, and the change shows on your calendar straight away.
• Texts a confirmation once the call ends.

The same agent can call or text patients before their visit and rebook anyone who can't make it, so a cancellation turns into a new booking instead of a no-show.

A new patient who says "my knee's been killing me since Saturday's game" and an established patient who says "I need to move Thursday" are making two different calls. The agent hears the difference and asks the one or two questions that decide where each booking goes. Nobody presses a key.

Built on your EHR, not beside it

When we built our ModMed integration, we chose to book in real time rather than in overnight batches, because a patient on the phone can't wait for tomorrow's sync. When a call starts, the agent identifies which practice the patient is calling, finds the patient, searches live slots and creates the appointment over FHIR R4, with every request logged for audit. We tested the ModMed integration in a ModMed sandbox before it handled a real patient call.

Epic connects through SMART on FHIR for scheduling and insurance, and the same approach covers athenahealth, NexHealth and Raintree. If your scheduling runs on something else, the agent connects to it through its API.

For your IT team: agents act in your systems through functions you control, built in Dialog Studio (formerly Agent Studio) or as code with the Agent Development Kit (ADK). Patient data runs under HIPAA Security Rule controls examined within our SOC 2+ report, and a BAA is available. It stays in the region you choose (US, UK, EU or APAC) with no cross-region transfer, and it's never trained on without your opt-in and never sold.

Your scheduling rules, applied on every call

Scheduling templates are where generic tools break. You decide which providers see new patients, how long a follow-up runs and when a same-day slot is allowed. The agent applies those rules the same way at 8am on a Monday and at 6pm on a Friday. Start with the visit types that keep your phones busiest, then add the rest once you trust it.

When the rules change, your team changes the agent. You test the change on a branch against real call scenarios before it goes live, and you can roll it back if you need to. Wren, the agent inside Dialog Studio, reviews live calls, flags bookings that went sideways and drafts a fix that a person approves before anything changes.

What patients and teams get

Howard Brown Health, a federally qualified health center that serves more than 40,000 people a year in Chicago, cut average handle time on routine requests by 72% and reached 30% call containment, against a 20% target. Its planned next phase connects the agent to Epic so patients can create, reschedule and cancel appointments themselves.

Audibel runs more than 400 hearing care clinics. After launch, its call abandonment fell from 46% to 2%, and appointment volume rose 2% in a year when the wider hearing aid market shrank.

Where your schedulers stay in the loop

The agent does not bend a rule to squeeze someone in. A surgeon who wants to see a scan before booking, a patient asking for an exception, a request your rules don't cover: those go to a person, with the patient's details and what they asked for, so your scheduler starts at the answer rather than at hello.

Questions teams ask