A healthcare chatbot is software that answers patient questions in a chat window and, in its more capable form, completes administrative tasks such as booking or moving an appointment. The safe scope for most clinics is administrative: scheduling, practice policies, accepted insurance plans and directions. Clinical questions, including symptoms, medication and test results, belong with licensed staff.
This guide explains where a chatbot fits in a clinic, where it should stop, and what a practice manager should settle before launch. It is general information and not legal, compliance or medical advice.
In brief: a clinic chatbot earns its place by taking routine front-desk volume off the phones, provided its scope is narrow and its handoff to a person is reliable. Handling of protected health information (PHI) is a separate decision that should be made with a privacy officer, not assumed from a vendor's marketing.
What does a healthcare chatbot actually do?
Most clinic conversations are not clinical. Patients and prospective patients ask whether a practice accepts their insurance, how to move an appointment, what to bring to a first visit, where to park, and what a consultation costs. These questions are repetitive, they have documented answers, and they tend to reach the front desk at the moments when staff are already attending to patients in person.
A basic chatbot answers these from a library of approved content. A more capable AI agent can go one step further and act. In the case of Bund AI, the healthcare configuration is limited to scheduling and administration: it books an appointment against real calendar availability and emails a confirmation, reschedules or cancels within the change window the practice sets, confirms the details of an upcoming appointment, and answers questions about services, pricing, insurance and practice policies from the clinic's own content. It also captures a prospective patient's name and contact details as a lead when someone is exploring the practice and is not ready to book.
The distinction between a chatbot that explains and an agent that acts matters operationally. A bot that tells a patient to call the clinic to reschedule has simply moved the work. For a longer treatment of that difference, see our explainer on chatbots versus AI agents.
Where should a healthcare chatbot stop?
The boundary should be explicit and enforced in configuration, not left to the model's judgment. A clinic chatbot should not diagnose, interpret symptoms, comment on medication or dosage, or discuss test results. Each of these requires a licensed clinician who knows the patient.
Bund AI does not give medical advice. Its healthcare setup is designed to decline clinical questions and route them to the practice team, passing along the conversation so the patient does not have to repeat themselves. Management teams should test this boundary directly before launch by asking the agent the questions a worried patient would ask, and confirming that it declines and escalates rather than improvising.
Urgent situations deserve separate attention. A clinic should publish clear emergency guidance in its knowledge content and in the chat greeting, directing anyone with an urgent symptom to emergency services rather than to a chat window. A chatbot is not an emergency channel and should never be presented as one.
How does a healthcare chatbot handle patient scheduling?
Appointment logistics are the strongest use case, for a practical reason: they are rule-bound. A patient wants a slot, a change or a cancellation, and the clinic has availability and a change window. Bund AI connects to calendars such as Google Calendar and Cal.com for direct booking, and can share a Calendly link for practices that prefer it. Our overview of AI appointment scheduling describes the mechanics in more detail, and the appointment booking solution page lists the supported calendars.
Rescheduling deserves particular attention because of its link to no-shows. Patients who find it awkward to move an appointment frequently do not attend, and an empty slot cannot be recovered. Making a change possible at any hour, in the patient's own language (Bund AI replies in more than 40), removes one avoidable cause of missed visits. The effect on any given practice depends on its patient mix and existing reminder process, and a clinic should measure its own no-show rate before and after rather than rely on a general claim.
Is a healthcare chatbot HIPAA compliant?
A chatbot is not compliant by default. In the United States, compliance depends on a signed Business Associate Agreement (BAA) with any vendor that handles PHI on a covered entity's behalf, technical safeguards that meet the Security Rule, and the way the practice configures and uses the tool. There is no official HIPAA certification for chatbots.
Bund AI is not HIPAA compliant, does not sign BAAs, and should not be used to collect, store or process PHI. Its suitability is therefore limited to administrative and public-facing conversations, such as hours, location, accepted insurance and practice policies, where the visitor is not asked to identify themselves or describe their health. Many compliance teams treat even a named appointment request at a provider as PHI, so a practice that wants identity-verified scheduling inside a chat should use a vendor that signs a BAA and should consult its privacy officer first. A fuller buyer's checklist, including vendors that publicly offer a BAA, is in our guide to the HIPAA compliant chatbot.
A common and conservative pattern is to let a general chatbot answer public questions and direct patients to the practice's covered scheduling system or patient portal for anything personal. The right design for a particular clinic is a decision for its privacy officer and counsel.
What should a clinic check before launching?
Several decisions precede the technology. The first is data scope: the practice should write down exactly what the chatbot is permitted to collect and where that information is stored. Bund AI masks card numbers at storage and can also mask email addresses and phone numbers, and its sessions are signed and revocable, but masking does not replace a data-flow review.
The second is the knowledge base. The agent answers from content the clinic provides, whether crawled web pages, uploaded PDF, DOCX or Markdown files, or pasted text. Accuracy therefore depends on the currency of that content. Insurance lists, fees and policies should have a named owner and a review date. The agent also flags questions it could not answer, which gives the practice a running list of content gaps.
The third is the handoff. Escalation should reach a named person with the conversation attached, and staff should know what to expect. The escalation and handoff page describes how this works, and our piece on when AI should hand off to a human sets out criteria worth adapting. Related considerations for other regulated professions appear in our discussion of the legal chatbot.
What does a clinic chatbot cost?
Bund AI prices on flat monthly plans with no per-resolution fees. The Basic plan is $0.99 per month for 50 AI replies, Starter is $49 per month for 1,000 replies and three seats, Growth is $99 per month for 5,000 replies and 10 seats, and Scale is $199 per month for 10,000 replies and 25 seats. Enterprise is custom. An AI reply is every message the agent sends, and a single conversation often takes two or three. A practice can estimate its monthly volume from its current call and email counts, and the pricing page has the current detail.
Early rollouts of Bund AI have shown 86.7% of tickets resolved with no human touch and a 1.8-second median first response. These figures come from early rollouts across industries, not from clinics specifically, and a practice should treat them as indicative. The healthcare industry page sets out the configuration in full.
When Bund AI is not the right fit
Bund AI is a poor fit for any clinic whose main need is identity-verified patient interaction. If the requirement is to look up a patient's record in an electronic health record, discuss results, process prescription refills or handle intake forms containing health history, the clinic needs a vendor that signs a BAA and integrates with its EHR. Bund AI does not offer a native EHR integration, and it does not operate a phone line, SMS or WhatsApp channel; its channels are the embeddable web widget and an email inbox.
It is also not a substitute for clinical triage. Practices that expect a chatbot to assess urgency should reconsider, because that judgment belongs with trained staff. Larger health systems with complex scheduling across many locations and providers will likely need a purpose-built patient access platform.
Frequently asked questions
What is a healthcare chatbot used for?
A healthcare chatbot is used mainly for administrative patient contact: booking and rescheduling appointments, answering questions about services, pricing, insurance and policies, and capturing inquiries from prospective patients. It should hand clinical questions to staff rather than answer them.
Can a healthcare chatbot give medical advice?
It should not. Bund AI declines medical and clinical questions and routes them to the practice team with the conversation attached. Anyone with an urgent symptom should be directed to emergency services, not a chat window.
Is a healthcare chatbot HIPAA compliant?
Not automatically. Compliance depends on a signed BAA, the vendor's safeguards and the practice's configuration. Bund AI does not sign BAAs and is not suitable for PHI, so it fits anonymous, administrative conversations only.
Can a chatbot book patient appointments?
Yes, for administrative scheduling. Bund AI books against real calendar availability, reschedules and cancels within the practice's change window, and emails a confirmation. Whether a named booking counts as PHI is a question for the practice's privacy officer.
How long does it take to set up a clinic chatbot?
Bund AI is designed to go live in about a day once the knowledge content and calendar are connected. The longer task is reviewing the content for accuracy and agreeing the escalation process with staff, which should happen before launch.