Where to start with AI if you run a healthcare practice.

I get the same question from practice owners almost every week: “Everyone says we should be using AI — where do we even start?” My answer is always the same. Don't start with the shiny clinical stuff. Start with the front desk, because that's where you're quietly losing patients and burning out your staff every single day.

The problems I actually see in practices

When I sit down with a clinic, the pain is rarely about medicine. It's operational, and it looks like this:

• The phones never stop, and they still go unanswered. Refill requests, appointment changes, insurance questions, new-patient calls — all hitting one or two people who are also checking patients in. Calls roll to voicemail, and a chunk of those callers simply book somewhere else.
• No-shows eat the schedule. Every empty chair is revenue you can't get back, and most no-shows trace straight back to a missed or forgotten reminder.
• Your best staff are doing robot work. Highly capable people spend their day reading back hours, repeating directions, and playing phone tag instead of caring for patients in front of them.
• Nights and weekends are a black hole. Patients are most motivated to act when you're closed, and right now that demand goes nowhere.

None of that requires a futuristic algorithm to fix. It requires answering the phone reliably and never dropping a routine task — which is exactly what AI is good at today.

Where AI genuinely helps (and where it doesn't)

Let me be clear about something, because there's a lot of hype: I do not recommend pointing AI at diagnosis, treatment decisions, or anything clinical as a starting point. That's where the risk lives and the value is hardest to prove. The fastest, safest wins are administrative:

• Call handling and routing. An AI voice agent answers every call in one ring, understands what the caller needs, and routes them to the right person or department — or just handles it.
• Scheduling and reminders. Booking, rescheduling, and confirmation reminders by call and text, which is the single biggest lever on no-shows.
• Intake and routine questions. Hours, location, what to bring, insurance accepted, prep instructions — answered instantly and consistently.
• Knowledge at the front desk. A private assistant trained on your policies so staff get the right answer in seconds instead of interrupting a provider.

The goal isn't to replace your team. It's to delete the repetitive 70% so your team can do the part that actually needs a human.

Where to start: a phased plan

Here's the sequence I'd run if it were my practice. Each phase pays for itself before you move to the next.

Phase 1 — Answer every call and cut no-shows

Put a voice AI agent on your main line to catch overflow, after-hours, and rush-hour calls, and wire up automated appointment reminders. This is the highest-ROI move in healthcare, full stop. You recover booked appointments you were losing and you take the worst pressure off the front desk in week one.

Phase 2 — Automate intake and the repetitive questions

Once the phone is handled, automate new-patient intake, form collection, and the FAQ that eats your staff's day. Now the people at your desk are handling exceptions, not reading back the same five answers.

Phase 3 — Put your knowledge to work

With the front desk under control, give your team a custom assistant trained on your protocols, payer rules, and SOPs. This is where you start compounding — faster onboarding for new hires, fewer interruptions for providers, and consistent answers across the practice.

The mistakes I'd tell you to avoid

• Don't lead with clinical AI. Prove value and build trust on the administrative side first. The clinical conversation can come later, carefully.
• Don't ignore compliance. Anything that touches patient information needs the right safeguards and a signed BAA with vendors. We scope this before anything goes live — it is not an afterthought.
• Keep a human in the loop. Urgent, sensitive, or unusual calls should warm-transfer to a person with full context, never hit a dead end.
• Don't try to boil the ocean. A six-month “AI transformation” usually transforms nothing. One fix that works beats ten that are “in progress.”

The practices that win with AI aren't the ones chasing the most advanced tool. They're the ones who fixed the unglamorous front-desk problems first, freed up their people, and then kept going from a position of trust.

What this looks like in the real world

Let me make it concrete. A mid-size practice I'd typically work with misses 15 to 25 calls on a busy day — new patients, reschedules, refill questions. Say even 6 of those were new patients worth, conservatively, a few hundred dollars in first-visit value and far more in lifetime value. Recover half of them and you've added meaningful weekly revenue without spending another dollar on marketing. Now layer in no-shows: if you run 8 hygiene and provider hours a day and a reminder system trims no-shows from 12% to 5%, you've quietly reclaimed several appointment slots a week that were pure loss.

None of that is exotic. It's the same boring math every practice already feels in its gut but never quantifies. The AI just makes sure the calls get answered and the reminders actually go out, every time, without depending on whether the front desk had a spare second.

How you'll know it's working

Don't measure AI by how impressive it sounds. Measure it by numbers you already track: percentage of calls answered, new-patient bookings per week, no-show rate, and how often your front-desk team has to interrupt a provider. If those move in the right direction within the first month, you're on the right path. If they don't, something's misconfigured — and that's a conversation to have early, not six months in. I'd rather a practice start small, watch one number move, and expand from proof than buy a big platform on faith.

Questions I get from practice owners

Is AI in a medical practice actually HIPAA-compliant?

It can be, and it has to be — this is the first thing we scope. Anything that touches patient information runs on platforms and configurations built for healthcare, with a signed business associate agreement in place. Compliance isn't a feature you bolt on later; it's the foundation, and I won't put anything live that doesn't meet it.

Won't patients be annoyed talking to AI?

The opposite, usually. Patients are annoyed by hold music, voicemail, and not getting a callback. A natural-sounding agent that answers in one ring and actually helps them — or gets them to the right person fast — beats the experience most practices give today. And anyone who needs a human can reach one immediately.

How long until it's live?

Weeks, not months, for the front-desk pieces. We start with one fix — usually call handling and reminders — get it working, and expand from there. You shouldn't have to wait on a giant rollout to see the first results.

More from the founder: Where to start with AI · AI & operational efficiency · All resources