AI for clinics: what's actually worth automating in 2026

Two things are worth automating in a practice right now: consultation notes, and the twelve questions your front desk answers every day. Almost everything else being sold to clinics is not ready.

Ask a practice owner when they write their notes and you will often hear: after dinner.

Patients until six. Eat. Then write up the whole day from memory, because the shorthand from the chair is unreadable to anyone else. It rarely gets described as a problem. It is just what the job is, the way you would describe a commute.

The two things worth automating

1. Consultation notes

Most of what makes charting slow isn't judgement. It is transcription. Someone says something out loud, and later someone types it up properly.

That part is solved. The consultation gets recorded, the notes come back structured, and the clinician corrects the two things that are wrong instead of writing all of it from scratch. We built exactly this for a clinical platform: notes generated automatically from the visit itself.

The time saved is not marginal. It is the difference between finishing at six and finishing at nine.

2. The front desk phone

Your front desk answers the same twelve questions every day. Opening hours. Do you take my insurance. What should I bring. Can I move my appointment. Where do I park.

None of it needs a person. All of it interrupts one, usually while a patient who actually wants to book is on hold deciding whether to wait.

Those questions can answer themselves, from your own policies, in your own wording. The calls that genuinely need a human then get a human who isn't already worn down.

What to ignore for now

Diagnostic AI, automated coding, anything that makes a clinical decision without a clinician signing it off. Not because the technology cannot do parts of it, but because the liability, the regulation and the failure modes are not worth it for a practice your size. Let hospital systems absorb that risk first.

Be sceptical of anything sold as an all-in-one practice AI. The value is concentrated in two narrow, boring, high-frequency tasks. A product that claims to do fifteen things is usually doing none of them well.

What it costs

An assistant trained on your own documents and policies starts around $8,000 and takes about four weeks. Most of that four weeks is not the AI, it is cleaning up the policies so there is one current version of each, which is work worth doing regardless.

You can estimate your own range with the cost calculator.

The one question to ask a vendor

Where does the data live, and can you show me an answer with its source attached?

If an assistant cannot tell you which document a statement came from, you cannot audit it, and in a clinical setting you will need to.