gBell is software that carries patient calls through scheduling, intake, system updates, and staff handoffs for clinics and hospitals. It is a software product, not an agency or a staffing service.

Small and mid-size medical and dental practices where the front desk is understaffed, and every missed call is a patient who books somewhere else or a no-show nobody confirmed.

Is gBell HIPAA compliant, and does it sign a BAA?

gBell operates as a business associate under HIPAA and signs a Business Associate Agreement with a practice before any patient information moves. The BAA governs how protected health information is handled and takes precedence over the standard terms of service. A vendor claiming its terms of service already cover HIPAA is not describing how the law works.

Yes. The agent states at the start of every call that the caller is speaking with an automated assistant and that the call is recorded. A practice can adjust the wording and the voice but cannot remove the disclosure or move it later in the call. If a caller asks not to be recorded, the agent offers to take a message or transfer to a person. Several states, California among them, require every party to a confidential conversation to consent before it is recorded.

Does gBell use patient data to train AI models?

No. Protected health information is not used to train, fine-tune, or evaluate any AI model, and agreements with model providers prohibit them from doing so with data gBell sends. Where a provider offers zero data retention it is enabled. Where a provider retains data briefly for abuse monitoring, that retention is covered by a business associate agreement and the data is still not used for training.

What happens if a caller has an emergency?

The agent is instructed to tell a caller describing an emergency to hang up and call 911, and to flag the call to staff at the highest priority. gBell is not an emergency service, cannot dispatch help, and must not be relied on as a triage mechanism. A practice remains responsible for an emergency plan that does not depend on the service being available or correct.

What happens when a call needs a human?

The call escalates to staff, who receive the transcript and a patient summary before they pick up rather than after. Practices configure when, how, and to whom calls escalate.

Can patients book without calling?

Yes. gBell provides an embeddable self-scheduling widget a practice can place on its own website, and patients who booked by phone can reschedule or cancel through a link sent to them.

How does gBell protect patient information in emails and text messages?

Patient-identifying detail is not placed in email or SMS message bodies. Those messages carry a link, and the patient verifies with a one-time passcode before anything is shown, so a forwarded message or an unlocked phone does not expose someone’s appointment.

Does gBell verify insurance?

Yes. gBell sends a 270 eligibility request, reads the payer’s 271 response, explains the returned coverage and benefits, and routes missing detail to the next action.

Does gBell post EOBs and ERAs?

Yes. gBell Payment Posting matches an EOB or ERA to the right claim lines, applies approved posting rules, and reconciles the remittance to its check or EFT before clean work reaches the practice system.

Does gBell include a clinical scribe?

Yes. gBell Scribe turns post-visit clinician dictation into a case-specific draft with exact source quotes, uncertainty review, clinician approval, and export.

Does a practice keep its own phone number?

Yes. Practices keep their own number, their own calendar, and their own tone. The agent answers in the practice’s name, not the vendor’s.

Practice operations · August 2026 · 6 min read

Appointment reminders: AI calls vs texts vs email

Every reminder channel gets sold on the same promise, fewer no-shows, and each one actually delivers something different. A text gets seen. An email gets archived. A call gets an answer. The right question isn't which channel "works," it's what information each one gives you back, because a reminder program's value is mostly in what it learns, not what it sends.

What each channel actually returns

Text messageHigh visibility, near-zero friction, and the default choice for good reason. Its weakness is what silence means: an unanswered text is unreadable. Maybe they saw it and are coming, maybe the number is dead, maybe it sat unread. "Reply C to confirm" helps, and plenty of patients who fully intend to come never reply to anything.

EmailFine as a paper trail, with the confirmation details, directions, and forms. As the primary nudge it competes with everything else in an inbox and loses. Use it as the record, not the reminder.

A phone callThe only channel that ends in a definite state. A "yes" is a real confirmation. A "Thursday doesn’t work anymore" is a cancellation you learned about early enough to fill the slot. A voicemail is at least a delivered message. And unlike a text, a call surfaces the conversation nobody planned: the patient who mentions they never got their referral, or asks what the visit will cost.

The traditional problem with calls was labor: nobody staffs a front desk generously enough to phone tomorrow's entire schedule. That constraint is what changed. An AI agent making reminder calls costs roughly what a text blast costs and returns what a human caller returns, an answer, which is why the channel is worth reconsidering in 2026 even by practices that abandoned call reminders years ago. Reminder calls with exactly this behavior, including the early-cancellation catch feeding the waitlist, are how we built gBell, so read our enthusiasm for the channel with that in mind.

What "confirmed" should actually mean

Audit your current reminder tool by asking what its "confirmed" column literally measures. Often it means "message delivered," which measures your phone bill, not the patient's intent. A useful confirmation state machine has real states: reached and confirmed, reached and cancelled, message left, and unreachable. Each drives a different action, and the last two are work queues, not outcomes.

Reached and confirmed: done. Attach the copay amount to this touch if you know it; a patient told "see you Thursday, your copay is $25" arrives ready.

Reached and cancelled: the best bad outcome. The slot is open a day early; the waitlist should hear about it within minutes, not at the next staff lull.

Message left: partial signal. Voicemail plus a text fallback covers most of these.

Unreachable: a data problem wearing a scheduling costume. Dead numbers found by reminders are how patient records get cleaned for free.

Cadence, and the diminishing-returns cliff

More reminders help until they annoy, and annoyed patients start ignoring the channel entirely, which poisons future reminders too. A sane default for routine visits: one touch a few days out, one the day before, done. Long-lead specialist appointments earn an extra early touch at the one-week mark. New patients earn slightly more attention than established ones, since their no-show risk is higher and their directions questions are real.

Whatever the cadence, the reminder is also the cheapest moment to move other work forward: confirm the phone number, state the copay, mention what to bring. Every question answered in the reminder is a question that doesn't queue at the desk on Thursday.

Common questions

Are text reminders enough by themselves?

For a large share of patients, yes, and texts should stay the base layer. The gap is the silent remainder: unanswered texts hide both no-shows and dead phone numbers. A call layer, human or AI, is how you turn silence into answers for the appointments where an empty slot actually costs you.

When should reminder calls go out?

Day before, generally late morning or early evening when answer rates are decent, plus a few-days-out touch for longer-lead appointments. More important than the hour: leave enough runway that a cancellation discovered by the reminder can still be refilled.

Do patients find AI reminder calls annoying?

A badly built one, absolutely. The tolerable version identifies the practice immediately, states its business in one sentence, takes yes or no gracefully, hands off to a human on request, and never calls twice for the same appointment. Same etiquette a good human caller follows, enforced by code instead of training.

gBell

gBell is the AI operations layer for medical practices. Calls answered and booked, reminders run, after-hours covered, and the intake and insurance work behind the phone carried. It's the product these notes come from.

See how it works

Sources