The last major session of The National ran from 3:20 to 5:00 on Saturday afternoon, and it was not the one anybody would have predicted from the title. "Artificial Intelligence. Real Implications. What Every DC Needs to Know NOW" turned out to be almost entirely about money. Specifically, about chiropractic insurance AI: the software sitting between a practice and its reimbursement, and what a practice can do about it.
A member of our team was in the room. This is what was presented, including a warning about AI-written clinical notes that deserves wider circulation, and the practical recommendation the presenter kept returning to. It closes our series from Orlando, after the day one report on AI and practice operations.
Table of Contents
The Session That Closed the Week
David Klein presented. His background is the relevant part: he is a certified professional coder and compliance officer, founder of a coding and compliance consultancy, and co-founder of a practice management system. Not a technologist talking about the future, in other words, but someone who works on getting practices paid.
Where it sat in the week
It ran opposite the ownership and transitions session in Celebration 12-13, which meant attendees had to choose between the two Saturday sessions with the most direct bearing on the money side of a practice. Our notes on the second are therefore drawn from the event's published materials rather than from the room.
What it actually covered
By our attendee's notes, the material was insurance, documentation and reimbursement rather than the marketing applications most AI sessions default to. How payers use AI to review claims, profile providers, identify outliers and trigger denials, and how a practice can use the same category of tool on its own side of the wall.
Why that framing mattered
Most conference AI content is aimed at what a practice could gain. This was aimed at what a practice could lose, which is a harder sell from a stage and a more useful hour for the audience. Rooms respond differently to a warning delivered by someone who spends their working life on audit defence than to a pitch delivered next to a booth.
How Insurers Are Using Chiropractic Insurance AI
The presentation's core claim was that the review process has already moved, and that most practices are still writing for the reader it used to have.
Profiling the provider, not only the claim
Per the session, payers are using AI to identify provider patterns rather than assess each claim in isolation. That is a meaningful shift. A single claim can look unremarkable while a provider's overall pattern flags as an outlier, and the practice never sees the comparison being made or the population it is being compared against.
Why practices cannot see it happening
The uncomfortable feature of chiropractic insurance AI at this level is that it is invisible from the practice side. You see your own denials. You do not see the peer group you were measured against, the threshold that triggered the review, or whether your pattern shifted because of something you changed or because the model did. That opacity is not incidental; it is what makes the system difficult to argue with.
Predicting necessity before a human looks
The session also described AI being used to predict medical necessity and flag claims accordingly, with denials capable of being generated before any human meaningfully reviews the case. We covered the mechanics of that in more detail in our report on AI and chiropractic claim denials from Friday's programme. What Saturday added was the provider-profiling dimension, which is the part a practice can least easily see.
The Warning About AI-Written Notes
This was the part of the session with the widest implications, and the part least likely to be on anyone's radar.
The category of error that matters
It is worth separating two failure modes that get discussed as one. Software that transcribes badly produces text a practitioner will catch on any read-through. Software that generates confidently produces text a practitioner will not catch, because there is nothing in the sentence to signal that it is wrong. The second is the one Klein's session was warning about, and it is specific to generative tools rather than to automation generally.
Notes that describe things that never happened
Klein's presentation warned that AI-generated clinical notes can invent information: examinations, diagnoses or details that were never actually performed or documented. Not garbled text, which is obvious and correctable, but plausible clinical content describing care that did not take place. The presentation cited research finding that AI-generated notes contained more unique errors per note than human-written ones.
Why this is a different order of problem
A typo is visible. An invented examination finding is not, because it reads exactly like a real one, and it enters the legal and clinical record under a practitioner's signature. The exposure is not that the software made a mistake. It is that a human attested to it.
This sharpens the review-first discipline that ran through every documentation session at this event, which we set out after Friday's panels in our piece on AI and chiropractic documentation. Reading the draft is not a quality step. It is the only thing standing between an efficiency tool and a fabricated record.
💡 Pro Tip
If you use any AI scribe, change what you are checking for. Most practitioners proof-read for wording. The thing to look for is the opposite: content that reads perfectly well and describes something you did not do. Scan specifically for examinations, findings and measurements you do not remember performing, because those are the errors that will never look wrong.
Get a quote →Fighting AI With AI
The session's strongest practical recommendation, and the one our attendee flagged as the takeaway, was essentially symmetrical: if the payer is running software against your claims, run software against them first.
Check the claim before it goes
The specific applications named were pre-submission checks for code and diagnosis mismatches, missing modifiers, and documentation gaps. This is the practical answer to the asymmetry that has been the theme of the whole week. A denial costs the payer almost nothing to issue and costs the practice an afternoon to appeal; the only way to rebalance that is to reduce the number of claims that give the algorithm something to catch.
Why prevention beats appeal every time
Worth stating the economics plainly. An appeal recovers money you had already earned, at the cost of staff hours you cannot bill. A clean submission earns the same money with none of that cost. Every hour spent on pre-submission checking is therefore worth more than an hour spent appealing, and it is the one place in this whole arrangement where a practice holds the faster process rather than the payer.
Appeals faster than a person can draft them
The second application was generating appeals. Appeal writing is exactly the kind of work that eats staff time out of proportion to the judgment it requires: structured, repetitive, and dependent on assembling the same categories of evidence each time. Speeding it up does not make the appeal more likely to succeed on its merits, but it makes more appeals get written at all, which in a system that profits from practices giving up is the point.
The caveat he kept attaching
By the notes, the slides repeatedly warned against blind trust in automation and insisted on a human reviewing AI output. Given the preceding section on invented clinical content, that is not a disclaimer bolted on at the end. It is the same argument twice: these tools produce drafts, and a draft that nobody reads becomes a liability the moment it is submitted.
Want us to do this for your clinic?
See the quality before you spend a dollar. Tell us about your clinic and we send a clear monthly price for the videos you need, within one business day, so you can judge the fit on your own practice rather than on our promises.
Get a quote →Chiropractic Insurance AI Is Also a Valuation Story
Saturday's programme carried a second theme in a different room, and it is connected to the first more closely than the schedule suggests.
Ownership, valuation and who is buying
The event's business track addressed how ownership trends are changing practice value and transitions, including a figure presented at the event that close to 70% of chiropractors are expected to retire within the next twenty years, and that private equity and larger group ownership are increasingly affecting valuations and sales.
The connection to the AI material is not obvious until you hold them together. A practice being valued for sale is valued partly on the reliability of its revenue, and revenue reliability now depends on how well its documentation survives an automated review. Clean, consistent, defensible records have quietly become an asset on a balance sheet rather than an administrative chore, and a practice with a denial problem it never diagnosed is worth less than one without it, whether or not anyone names the reason during the sale.
What the Whole Week Actually Said
Three days, and every session that touched artificial intelligence drew the same line without appearing to coordinate.
Three days, one boundary
What makes that convergence worth noting is that these were different presenters, in different tracks, with different commercial interests, speaking to different rooms. When a national event produces the same rule from every direction without anyone appearing to coordinate it, that is a professional norm forming in real time rather than a trend somebody is selling.
The line
Operations, administration and documentation drafting are fair territory. Clinical judgment, compliance responsibility and patient data are not. Thursday's practice management course said it, Friday's morning panel said it, Friday's afternoon session said it, and Saturday's session said it hardest, because Saturday's presenter was describing what happens when the line is crossed rather than recommending it be respected in principle.
What a practice should do on Monday
Nothing dramatic, and nothing that requires buying anything. Read the last month of notes for content describing care that did not happen. Log denials with their codes and stated reasons so a pattern can become visible. Check claims for mismatches before submitting rather than appealing after. And treat chiropractic insurance AI as a permanent feature of the environment rather than a wave to wait out, because nothing presented this week suggested it recedes.
None of that requires a purchasing decision this quarter, which is the useful thing about it. A practice can improve its position against chiropractic insurance AI considerably before it buys any software at all, simply by tightening what it already produces. The tools help after that, not before it.
That is the honest summary of The National 2026 from where we sat. The technology conversation in this profession has moved past whether to use these tools. It is now about who is already using them on you, and what a practice does about that.
