Our front desk gets roughly 40 phone calls a day. About half of those are questions that could be answered without a human: “What time is my appointment?” “Do I need a referral?” “How do I clean my CPAP mask?” “Where do I park?”

So when AI chatbots started popping up in healthcare marketing materials, I was interested. Not because I wanted to replace our reception staff — they’re brilliant — but because freeing them from repetitive queries would let them focus on the patients standing right in front of them.

Here’s what I’ve learned after researching (and eventually implementing) chatbot technology in a sleep medicine setting.

The Case For

Phone hold times are a real problem in medical practices. A study published in Health Affairs found that long wait times for scheduling were one of the top drivers of patient dissatisfaction in outpatient settings. When a patient calls at 8:30am and sits on hold for six minutes, they’re annoyed before they even speak to anyone.

A well-configured chatbot on your website or messaging platform can handle a surprising volume of inquiries instantly:

• Appointment scheduling and rescheduling — integration with your practice management system means patients can book without a phone call
• Pre-appointment instructions — what to bring, fasting requirements for split-night studies, medication guidance
• CPAP supply reordering — mask cushions, filters, tubing
• Insurance and billing FAQs — coverage questions, Medicare eligibility for sleep studies
• Directions and parking — sounds trivial, but this accounts for a lot of calls

One an AI consultancy we consulted estimated that a well-trained chatbot can handle 60-70% of inbound patient queries without human escalation. That number matched what we eventually saw in practice.

The Case Against (Or at Least, for Caution)

Healthcare chatbots carry risks that general business chatbots don’t. The big one is clinical advice. A chatbot should never diagnose, recommend treatment changes, or interpret test results. Full stop. Patients will absolutely try to get clinical guidance from a chatbot, and the system needs clear guardrails.

Privacy is the other major concern. If your chatbot handles protected health information — appointment details, insurance info, medical history — it needs to comply with relevant privacy legislation. In Australia, that means the Privacy Act and Australian Privacy Principles. In the US, HIPAA applies. Make sure any vendor you’re considering can demonstrate compliance, not just claim it.

Then there’s the uncanny valley problem. Patients who think they’re talking to a human and later discover it was a bot tend to feel deceived. Transparency matters. Label your chatbot clearly: “Hi, I’m the SleepMedicineService virtual assistant. I can help with scheduling, directions, and general questions. For medical advice, please contact your sleep physician directly.”

What It Actually Costs

Pricing varies enormously. Simple rule-based chatbots — basically glorified FAQ pages with a conversational interface — can run as low as $50-100 per month. These work fine for small practices with predictable query patterns.

AI-powered chatbots with natural language understanding, appointment integration, and learning capabilities typically cost $300-800 per month for a small to mid-size practice. Enterprise solutions can run into thousands.

The ROI calculation should factor in:

• Reduced phone call volume (track this before and after)
• Staff time freed for higher-value tasks
• After-hours availability (patients can get answers at 11pm)
• Appointment no-show reduction through automated reminders
• Patient satisfaction scores

Our Experience

We went live with a chatbot six months ago. The first month was rough — the bot didn’t understand common sleep medicine terminology and kept misrouting queries. “I need a new mask” got interpreted as a cosmetic surgery question twice.

After training the system on sleep-specific vocabulary and common patient phrasing, things improved substantially. By month three, we were deflecting about 55% of incoming queries. Our phone hold times dropped from an average of four minutes to under two.

Patients over 65 use it less than younger patients, which isn’t surprising. We made sure to keep phone access equally prominent so nobody feels forced into a digital interaction they’re uncomfortable with.

My Honest Recommendation

If your practice handles more than 20 calls a day and you have a website, a chatbot is probably worth exploring. Start with a basic version that handles scheduling and FAQs. Don’t try to build a clinical triage system — that’s a different product with much higher regulatory risk.

Get your staff involved early. They know the questions patients actually ask, and they’ll spot gaps in the chatbot’s training data faster than any developer will.

And please, for the love of clear communication: don’t pretend the bot is human. Patients respect honesty. They don’t respect feeling tricked.

The technology isn’t perfect, but it’s genuinely useful when scoped correctly. For a sleep clinic juggling CPAP queries, referral processing, and study scheduling, even modest automation makes a noticeable difference to daily operations.