A HIPAA-aware chatbot that answers patients and knows when to stop.
We build patient chatbots for clinics and healthtech products: answers drawn from your approved documents with citations, PHI kept to the minimum, refusal and escalation rules, and a person on anything clinical. At a two-location dermatology clinic, repetitive front-desk inquiries fell 62% in 8 weeks.
What makes a chatbot safe to put in front of patients.
No chatbot is HIPAA compliant on its own. Compliance comes from the safeguards around it and how it is run: a BAA for every service that touches PHI, the minimum patient data collected, encryption, access controls and a log of what was said. We build those in, and we keep the bot on topics where it cannot hurt anyone.
For a dermatology clinic with two locations, the assistant answers only administrative and preparation questions (hours, directions, insurance, forms, pre- and post-op instructions) from a clinician-reviewed set of documents, cites its source, refuses medical advice and hands off to staff with a summary. In 8 weeks repetitive front-desk inquiries fell 62%, the median time to answer fell from 4:10 to 1:20, and 40% of after-hours conversations were resolved without staff.
What goes into a patient chatbot.
Narrow on purpose: it answers what your documents answer, and passes the rest to a person.
Answers from approved sources
Retrieval over your SOPs, prep guides and insurance FAQs, versioned and reviewed by clinicians, with the source cited in each answer.
Guardrails
Not-medical-advice framing, refusal rules, keyword and intent escalation triggers, and rate limits.
PHI kept to a minimum
Only the details a task needs are collected, sensitive fields are redacted, and data is encrypted in transit and at rest.
Web, text and phone
A website widget, SMS and WhatsApp, and deflection from the phone line to self-serve, with an after-hours reply.
Handoff to staff
Anything outside its topics goes to a person with a summary of the conversation, so the patient does not repeat themselves.
Reporting
Deflection, time to answer, escalations and satisfaction, with weekly transcript reviews that decide what it learns next.
How a patient chatbot goes live.
Agree the topics
What it may answer, what it must refuse, and who takes over when it stops.
Build the source of truth
Your documents collected, reviewed by clinicians and versioned before the bot reads them.
Test before patients see it
An evaluation set of real questions, including the ones it must refuse, run before launch.
Launch and review weekly
Transcript reviews and dashboards decide what it covers next.
Published prices, mapped to the job.
Every price here is the same one on the pricing page. Work with no published price gets a fixed price in writing before it starts.
Figures in the notes are what past clients paid, from the published case studies; they show the range, not a rate card. Work with no natural end, such as ongoing development, is priced per engineer per month with no lock-in.
When to hire us for this, and when not to.
- Your front desk answers the same administrative questions all day.
- The answers already exist in documents your clinicians trust.
- You want after-hours answers without after-hours staff.
- You want it to give medical advice or triage symptoms. We keep chatbots to administrative and preparation topics.
- You need it to book into your calendar. Our AI receptionist page covers booking agents.
- You want it live this week with no review of where patient data goes.
A patient assistant in production.
A two-location dermatology clinic, eight weeks from start to measured results.
A RAG patient assistant with clinical guardrails cut repetitive front-desk inquiries 62% and time-to-answer 68% in 8 weeks.
These are past engagements at the prices those clients paid. For build price, monthly cost and payback side by side, see what ten client projects cost.
HIPAA-compliant chatbot, answered
Is a chatbot HIPAA compliant?
Not by itself. HIPAA compliance comes from the safeguards around the chatbot and how it is run: a business associate agreement with every service that touches PHI, the minimum necessary data, encryption, access controls and audit logs. We build those in and sign a BAA for engagements that touch PHI. We do not claim a certification we do not hold.
What can a patient chatbot answer?
In our published build: hours, directions, insurance, forms and pre- and post-procedure instructions, all from documents clinicians reviewed. It refuses medical advice and passes anything clinical to staff with a summary.
What results has it produced?
At a two-location dermatology clinic, after 8 weeks: repetitive inquiries handled by the front desk fell 62%, the median time to answer common questions fell from 4:10 to 1:20, satisfaction rose 18%, and 40% of after-hours conversations were resolved without staff.
Where does patient data go?
As little as possible is collected, sensitive fields are redacted, and data is encrypted in transit and at rest. We can deploy in your cloud, and every service in the chain that touches PHI needs a BAA.
How much does it cost?
A patient chatbot is a Gigabit Agents build, a flat fee from $8,000 agreed before work starts. Running it afterwards, with evaluations and transcript reviews, is Managed AI Operations.
Often scoped alongside.
AI receptionist
Answers chats and calls and books into your scheduling system.
AI receptionistHealthcare software development
The industry hub for care providers and healthtech.
Healthcare software developmentHealthcare app development
Patient apps, telehealth and provider dashboards.
Healthcare app developmentFrom the field notes
Want patients answered after hours?
Send the questions your front desk answers most. A senior engineer replies with what the chatbot should cover and a fixed price.