In a healthcare services company AI arrives in three places before anyone decides anything: the clinician's phone that drafts the visit note on the drive home, the billing office that writes the appeal and reads the payer's denial, and the front desk that answers the patient's message. The owner's job is not to pick the tool. It is to decide three things before a tool touches a patient's record, a claim or a care note: who in the company owns the question, what may never go into an outside tool, and what the company will never let a machine do alone. This article is for the owner of a home health or hospice agency, a therapy or dental group, a physician practice, a behavioral-health provider or a medical staffing company with a few dozen to a few hundred people.
I was COO and then President of a public company that closed more than forty acquisitions, and I ran the integration on most of them. I am a lawyer by training. Now I write fixed-fee AI plans for CEOs of private-equity-backed and owner-led mid-market companies through 013 Labs, and healthcare services companies are the group where a tool's mistake lands on a patient and a regulator may ask for the record.
Where AI already is in your company
Ask a clinician what the free tools on the phones are doing this week and the honest answer is: drafting the visit note from a voice memo, summarising the referral packet, writing the letter to the referring physician. Ask the billing office and it is the first draft of the appeal, the summary of the payer's denial, the answer to the prior-authorization request. Ask the front desk and it is the answer to the patient's portal message, the reminder text, the intake summary. Ask the office and it is the job posting, the collections note and the staff schedule. None of it was approved. Some of it has already told a patient something about their care with nobody's name on it, and some of it has carried a patient's record, a diagnosis or a payer's contract rate onto a vendor's server under no agreement at all.
The three decisions, in the order a healthcare company makes them
Who owns the question. Not the IT provider, not the records-system vendor and not a new hire. In most companies it is the clinical director, the compliance officer or the practice administrator, because they are the people the clinicians and the front desk already ask when something new shows up. Give that person a written scope of half a page: what they may approve alone, what comes to the owner, when the page is reviewed. The article on AI for a small business with no IT department draws the line between their column and yours.
The data rule. One sentence. For a healthcare company it usually reads: no patient's record, note, image, claim, appointment detail or message, no payer contract rate, and nothing about an employee go into a tool the company has not approved, and no tool is approved to hold patient information without the signed agreement that makes the vendor responsible for it. The owner writes it, because the patient's record is protected by law and the vendor's agreement is the owner's to sign. This page is general information for owners, not legal or medical advice; your healthcare lawyer and your compliance officer decide what the law requires of your company, and your clinicians decide what is right for a patient.
The never-build list. The three things the company will never let a machine do alone. For a healthcare services company the list writes itself: no diagnosis, care-plan change or medication decision is made by a tool; no visit note, claim, prior authorization or appeal is signed or submitted without the clinician or the biller reading it; no message about a patient's care goes to the patient from a tool alone. Everything else is a question of fit. These three are not.
What fits at fifty people, and what does not yet
The first project is the work your people already do by hand, the same way, every week: the visit note drafted from the clinician's dictation for the clinician to correct and sign, the referral packet summarised for the intake nurse, the appeal letter drafted from the denial for the biller to finish, the reminder text, the job posting, the schedule. The article on picking your first AI project has the five tests. What does not fit yet, in most companies this size, is anything that decides about a patient or a payment: the diagnosis, the care plan, the medication, the claim, the authorization, the answer to the patient. A tool can prepare any of those for a person. It should not finish them.
The payer's audit, the surveyor, the insurer and the buyer will ask
The AI question reaches a healthcare company from four directions that have nothing to do with technology. The payer's audit asks who wrote and signed the note behind the claim, and a note a tool drafted that nobody corrected is a claim the payer can take back. The accreditation surveyor asks which systems hold patient information and under what agreement. The malpractice and cyber insurers' renewal forms ask the same question in their own words. And a buyer, strategic or private equity, asks it in diligence, because healthcare services is a sector buyers roll up; the article on whether AI makes a business worth more when you sell has that version, and the one on what to tell your bank about AI covers the bank. One dated page answers all of them.
Picture a home health and hospice agency of two hundred twenty people across three offices. A dozen field clinicians have been drafting visit notes from voice memos in a free tool on their own phones for a season, and the notes have been finished the same day for it. The owner finds out when a payer's audit asks for the record of who wrote and signed a sample of notes, and then finds that the patients' names, diagnoses and medication lists went into a tool under no agreement with the vendor. The fix takes two weeks and one call to the company's healthcare lawyer. The compliance officer is named as the person who owns the question, with a half-page scope. The owner writes the data rule: no patient information into a tool without the signed vendor agreement, and nothing at all into a free one. The never-build list goes on the page, starting with no note signed unread. The free tool is replaced by one under a signed agreement the same month, dictation goes through it with the clinician correcting and signing every note, the records are deleted from the old vendor's account with the deletion confirmed in writing, and the auditor gets the page with the answer. This is a hypothesis, a composite and not a client.
The one page
What is in use, by role and task, including the free tools and the features the records-system vendor switched on. Who owns the question, with the written scope. The data rule in one sentence, with the vendor agreements listed. The never-build list, the three items above. The review date and who signs. The owner signs it, the compliance officer keeps it with the privacy policies, and the administrator attaches it to the next payer or surveyor request. The board-briefing template has the same page in a form a healthcare company can fill in. Not sure which of the three is missing? The AI readiness scorecard asks ten questions across the whole business and shows which one.
Mistakes that repeat
Letting the records-system vendor switch it on. A drafting feature arrives in an update, defaults to on, and nobody decides whether notes may start from it. The owner decides, feature by feature.
Patient information in a free tool. The one mistake that puts the company in front of a regulator and the patient's lawyer. The data rule exists for this.
A note signed nobody read. A tool drafts the visit note; a clinician signs at the end of a long day. The payer does not care who wrote it. The clinician reads every line before signing.
The answer to the patient. An answer to a portal message that a tool wrote and nobody checked is advice the company gave. A person answers anything about care.
No review date. Payers change their audits, surveyors change their standards, people change offices. A page without a date is a page nobody opens.
Frequently asked questions
Should a healthcare services company use AI?
It already does, through the clinicians' phones, the billing office and the front desk. The decision worth making is who owns the question, what may never go into an outside tool, and the three things the company never lets a machine do alone: a clinical decision, a note or claim signed unread, and a message to a patient about their care.
Who should own AI in a healthcare company?
The person the clinicians and the front desk already ask when a new tool shows up, usually the clinical director, the compliance officer or the practice administrator, with a written scope of half a page and a review date. Not the IT provider and not the records-system vendor.
Can AI write our clinical notes?
It can draft from the clinician's own dictation, for the clinician to correct and sign, inside a tool the company has approved under a signed vendor agreement. It should not finish the note, because the signature is the clinician's and the payer's auditor will ask who checked it.
What about patient information in these tools?
It is first in the data rule. No patient's record, note, image, claim or message goes into a tool the company has not approved, and no tool is approved to hold patient information without the signed agreement that makes the vendor responsible for it. If patient information has already gone into a free tool, delete it from the vendor's account, get the deletion confirmed in writing, and tell the company's healthcare lawyer the same day. This is general information, not legal or medical advice.
Why does my payer, my surveyor or my insurer ask about AI?
Because each carries the risk when a note, a claim or an answer to a patient that a tool touched turns out to be wrong, and each wants to know whether a person stands behind it. One dated page with names on it is the answer, and the same page answers a buyer and the bank.
Not sure where you stand? Take the 3-minute AI readiness scorecard.