Try for Free

AI customer service

AI customer service for telehealth: automate the admin, route the clinical

Virtual care teams answer the same administrative questions all day: when is my visit, why can't I connect, was my prescription sent, why was I charged. Aftersales Agent resolves those end to end and sends every symptom or medication question to your clinicians. Pair it with the telehealth helpdesk setup for your human team, or see how it fits across healthcare.

What do telehealth patients contact support about?

Most telehealth support volume is administrative: getting into a visit, getting a prescription to the right pharmacy and making sense of a bill. Agent can resolve most of it with read access to scheduling, payment and prescription-status data, and hands every clinical question to your clinicians.

Conversation typeCan Agent resolve it?Hand off whenData Agent needs
Book, reschedule or cancel a visitYesUrgent symptoms, or no slot inside the window set for that visit typeOpen slots, visit type, the patient's current state, cancellation policy
Can't join a video visit, or the clinician hasn't appearedYes, for device, browser and link problemsStill not connected five minutes after start timeStart time, join link status, device and browser
Prescription statusYes, status onlyAny question about dose, side effects or interactionsE-prescribing status, pharmacy on file, time sent
Refill and pharmacy change requestsPartly: records and routes the requestAlways for the decision, which is a clinician'sPrescription record, last visit date
Insurance eligibility and copaysYes, when the eligibility check answers itDenials, pending prior authorization, coverage disputesEligibility response, plan on file, copay rules
Unexpected charges and refundsYes, within a refund limitAbove the limit, chargeback or legal language, written estimate requestsInvoices and payments (Stripe), refund policy
Membership or subscription changesYesFinancial hardship or a care complaint given as the reasonSubscription status, next billing date
Login, identity checks, dependent accessPartly: resets and re-sent linksProxy access, failed identity checks, mismatched detailsAccount status, verification result, proxy rules
Visit summaries, work notes, receipts, records requestsYes for re-sending existing documentsFormal records requests, or a request to change a noteDocument store, request log
Symptoms, side effects, test resultsNoAlways, at once; emergencies get the emergency message firstNone: Agent routes without opening clinical notes

Example set-up. Your clinical leadership decides every threshold.

What does an AI procedure for a failed video visit look like?

Write the procedure the way you would brief a new care coordinator. Agent turns each line into steps, and each step uses a connector with permissions you can revoke. This example covers a patient who can't get into a video visit, where minutes matter because a late connection can cost the whole slot.

Example procedure: “I can't join my visit”

  1. Verify the patient with date of birth and the phone or email on the account before discussing the visit.
  2. Look up today's visit: start time, clinician, join link status and the state the patient is in now. If your clinicians can't see patients there, hand off to scheduling.
  3. If the visit is more than 15 minutes away, text a device check link and a fresh join link to the number on file.
  4. If the visit has started, work through the fixes in order: supported browser, camera and microphone permissions, close other apps using the camera, reload, switch between Wi-Fi and mobile data.
  5. Still not connected five minutes after start time: post the name, visit time and steps tried to the coordinators' Slack channel, and offer audio only if the visit type allows it.
  6. If the clinician records a no-show caused by a technical failure, offer three open slots and waive the late-cancellation fee up to $50.
  7. If the patient mentions chest pain, trouble breathing, stroke signs, self-harm or any emergency, stop, send the emergency message and hand off to the on-call clinical team.

Guardrails: no clinical advice; no diagnosis, visit reason or medication name in texts; fee waivers capped at $50 per patient per month; handoff on negative sentiment or words such as “lawyer” or “medical board”. Tagging device and browser lets Insights flag a spike after one browser update.

Example conversation (illustrative, not a real patient)

Patient, portal chat, 2:03 pm: My appointment was at 2 and the screen says waiting for camera.

Agent: I can help. To confirm it's you, what's your date of birth?

Patient: April 12, 1961.

Agent: Thanks, Maria. Your clinician is in the room. Your browser is blocking the camera: tap the camera icon in the address bar, choose Allow, then open the link I've just texted you.

Patient: OK, I can see myself now.

Agent: You're connected. If the call drops, reply here and I'll send a new link.

What must a telehealth AI agent never do?

A telehealth AI agent must never give medical advice. Agent doesn't assess symptoms, interpret results, suggest doses or say whether a medication suits someone. Those questions go to a licensed clinician every time, with the conversation, the data Agent looked up and what it already tried attached.

Agent never

  • Reassures about symptoms. “That sounds normal” is clinical advice.
  • Approves or promises a prescription or refill. That includes controlled medications, which DEA lets clinicians prescribe by telemedicine without an in-person visit under a temporary extension through December 31, 2026.
  • Shares visit details before verification, or with a relative who isn't an authorized proxy.
  • Puts a diagnosis, medication name or visit reason in a text or email subject line.

Agent always hands off

  • Symptom, side-effect and test-result questions.
  • Emergency or crisis language, after telling the patient to call 911, or to call or text 988 for the Suicide and Crisis Lifeline.
  • Complaints about care, privacy concerns and legal language.
  • Billing disputes above your refund limit.

How patient data is handled

A HIPAA covered entity needs a business associate agreement with any vendor handling protected health information on its behalf. Aftersales offers HIPAA support with a signed BAA on the Enterprise plan. Signing it doesn't make your platform compliant; your privacy officer still owns policy and configuration.

Keep advertising pixels off chat and portal pages. HHS has issued guidance on tracking technologies (a 2024 court ruling vacated part of it, so check the current version with counsel), and direct-to-consumer platforms outside HIPAA can fall under the FTC's Health Breach Notification Rule; the FTC's 2023 order against BetterHelp banned that counseling service from sharing health data for ads. See Aftersales security and our roundup of chatbots for HIPAA-regulated teams.

Which support channels matter most to telehealth patients?

Portal chat and SMS matter most, because both reach the patient on the device they are about to use for the visit.

  • Chat inside the patient portal or app. The patient is already signed in, so verification is quick and Agent knows which visit they mean.
  • SMS. Reminders, join links and “your clinician is running 10 minutes late” updates, kept free of clinical detail and sent under the texting consent rules your counsel sets.
  • Phone. Many older patients call rather than type, and video isn't always realistic for them: a JAMA Internal Medicine study using 2018 data estimated that 38% of US adults aged 65 and over were not ready for video visits. Aftersales phone, on Growth and above, includes transfers and recording consent.
  • Email and WhatsApp. Email for receipts, superbills and visit summaries; WhatsApp where patients already message. Agent replies in 31 languages.

Every channel lands in one shared inbox, so a patient who starts in chat and then calls doesn't explain the problem twice.

Which telehealth systems should an AI agent connect to?

Agent needs read access to the systems your coordinators already check, plus a short list of write actions. Stripe and Salesforce have built-in connectors. Scheduling, video, e-prescribing, EHR and eligibility systems connect through the Aftersales API or webhooks; confirm specific connectors on a call.

SystemWhat Agent readsWhat Agent may do
Scheduling or practice management systemOpen slots, visit type, patient's stateBook, reschedule, cancel
Video visit platformRoom status, join linkSend a fresh link
E-prescribing and pharmacy statusSent or not, pharmacy, timeLog a pharmacy change for review
EHR or patient recordAdministrative fields onlyRe-send a visit summary
Eligibility and billing systemCoverage result, copay, balanceExplain a charge
Stripe (self-pay visits and memberships)Charges, subscriptionsRefund within limits; pause or cancel
Salesforce, if your CRM runs on itEmployer or health plan sponsorUpdate the record

Scope permissions so Agent can't open clinical notes at all. No conversation above needs them, and a narrow scope is easier to defend in a security risk assessment. If your human team stays on Zendesk, Freshdesk or Salesforce Service Cloud, Agent can also run inside it with no seats and no migration, billed per resolution with a minimum monthly commitment.

How much does AI customer service cost for a telehealth platform?

A mid-sized virtual care platform pays for seats on Scale plus $0.99 for each conversation Agent resolves. Price on the Enterprise plan if conversations will contain protected health information, because that plan carries the BAA.

Example assumptions: 8 support seats, 5,000 patient conversations a month, Agent resolving 50% early on (below the 71% median Aftersales sees after 90 days, since clinical questions always go to people), Copilot for every rep, annual billing at list prices.

Line itemCalculationMonthly cost
Scale seats8 × $109$872
Agent resolutions2,500 × $0.99$2,475
Copilot add-on8 × $29$232
Conversations handed to people2,500 × $0$0
Total$3,579

That is about $0.72 per conversation across all 5,000. Agent never takes a seat. Clinical leads and your privacy officer can join as Lite seats to view, comment and approve; Scale includes 50. See every plan on the pricing page.

How do you roll out an AI agent for telehealth in 30 days?

Start narrow: one channel, one procedure, clinic hours only. Widen each week once the handoffs look right.

  1. Days 1 to 7: contracts and connections. Sign the BAA, connect scheduling, payments and prescription status, and have your medical director approve the escalation words and the emergency message.
  2. Days 8 to 14: visit access in portal chat. Turn Agent on for joining problems and scheduling during clinic hours. Coordinators read every handoff and fix procedures the same day.
  3. Days 15 to 21: prescriptions and billing. Add prescription status, refunds under your limit and membership changes, then open SMS and email.
  4. Days 22 to 30: after hours and review. Extend Agent to evenings and weekends. Review resolution rate and CSAT, and review every clinical handoff with your clinical lead before widening to after-hours.

Measure progress with the resolution rate guide, or browse other AI customer service guides by industry.

Frequently asked questions

Can an AI chatbot give medical advice to telehealth patients?
No. Aftersales Agent is set up with escalation rules that pass symptom, medication and test-result questions to your clinical team with the conversation attached. It handles the administrative work around care.
Will Aftersales sign a BAA for a telehealth platform?
Yes, on the Enterprise plan. Aftersales offers HIPAA support with a signed BAA on the Enterprise plan, but a BAA alone doesn't make a platform compliant; your own policies and configuration still matter.
What happens if a patient describes an emergency in chat?
Agent stops the procedure, tells the patient to call 911 or to call or text 988, and hands the conversation to your on-call clinical team. Your medical director approves the trigger words and that message before launch.
Can an AI agent tell patients whether their prescription was sent?
Yes, if it can read your e-prescribing or pharmacy status: whether the prescription went out, to which pharmacy and when. Refill approvals and questions about the medication go to a clinician.
How much does an AI agent cost for a small telehealth practice?
Each conversation Agent resolves costs $0.99, on any plan, and a conversation it passes to a clinician costs nothing. Practices that need a BAA move to the Enterprise plan, with custom pricing from sales.
AS
The Aftersales teamCustomer service product specialists

See which patient conversations Agent can resolve

Start a 14-day free trial, move to the Enterprise plan and sign the BAA before real patient data flows, then measure resolution on your own conversations.