Insights
Honest, practical notes on the experience your patient actually has, and on the decisions that come before a practice spends money on AI.
43 posts · 13 topics · new writing most days
The piece we would hand you first, and the arcs that run end to end.
The clinic down the road can buy the same models you can, this quarter, at the same price. What they cannot buy is how it feels to be your patient. That is where the differentiation moved, and most clinic AI budgets are still aimed at the layer that is becoming commodity.
Read articleThe six most recent, newest first.
The front desk, the clinician and the biller can each hit their target while the patient has a terrible eleven days. You measure the boxes and they experience the lines between them, and buying clinic AI one department at a time usually makes those lines worse.
Read article
A clinic pilot with no stated ending does not end, it fades, and keeps billing while it fades. Deciding the number, the date, and who calls it costs nothing on day one and is nearly impossible to introduce in month nine.
Read article
Data ownership, training rights, exit, price, and liability are settled the day you sign and renegotiated almost never. In a clinic one of them is not just commercial, it is a privacy obligation you already hold. Here is what each clause decides, what the standard wording quietly grants, and the version worth asking for instead.
Read article
Your staff are already using AI, so the policy question is not whether to allow it. It is whether the rules exist in writing or in each person's private judgement. Exposure comes in three shapes, and in a clinic the one most policies address is not the one most likely to reach a patient.
Read article
Ordinary software fails loudly and someone files a ticket. AI fails quietly and keeps answering. That difference is why a clinic AI system needs a named owner, a number, a cadence, and the authority to switch it off, and why most practices discover they have none of the four.
Read article
The same AI project reads as a threat or a plan depending entirely on how it is announced. Four facts make the difference, and they are the same four at the partners' meeting, at the staff huddle, and in anything a patient ever sees.
Read articleNothing in Latest matches that filter.
Everything else, 36 posts, newest first.
Caregivers to visits, patients to rooms, studies to scanners, therapists to a day of clients: it looks like four problems and it is one. Ask a language model to solve it and you get a fluent, infeasible answer. A short quiz sorts your problem into language, search, or both.
Most clinic AI reads text, but a great deal of what a practice knows arrives as pixels: faxed requisitions, handwritten intake forms, wound photos, scanned consult letters. A short quiz sorts your workflow into three honest tiers, and the failure modes of vision are not the ones you are watching for.
Every clinic product renamed itself an agent this year, and the word stopped carrying information. A five-scenario quiz trains your eye, and five questions expose what a vendor actually built, because the label decides the price you pay, the way it fails, and the oversight your practice owes it.
The hype says automate everything. In a clinic the discipline is knowing where the line goes, because the worst axis is always the one with a patient on it. A task belongs to a person, not a model, when it is hard to undo, needs real judgment, or puts health, money or safety at stake. Grade any task on those three axes here.
A second kind of visitor is reading your clinic site: one that does not scroll, does not wait for your scripts, and never phones to ask whether you are taking new patients. It takes a record and leaves. Here is what it can and cannot see, why most clinic sites fail at the same step, and the fixes that pay off whether or not the agents ever arrive.
Patients ask an assistant now, and no rank tracker can see inside that conversation. The honest instrument is a sample: ten questions a patient would really ask, run every month, scored on one axis. Build your probe set here and score your first run in about fifteen minutes.
Nobody gives a new receptionist chart access and a signature stamp on day one, yet practices hand both to a week-old AI tool, or grant nothing and get nothing. The fix is the oldest tool a clinic already owns: a scoped job, a named supervisor, probation, and promotions the tool has to earn.
The most-asked technical question in clinic AI has a decision tree for an answer, not a winner. Facts that change want retrieval. Voice and format want fine-tuning, rarely. And the real gap is usually a third, cheaper thing nobody argues about, plus one healthcare answer that is simply no.
Your next patient is asking ChatGPT which clinic to try, not scrolling Google. The answer names two or three, and either you are one of them or you do not exist. A scorecard shows whether an answer engine would cite your practice today, and the fixes are more honest than any SEO trick.
Every vendor pitch quotes the licence and goes quiet on what the thing actually costs to run. The running cost is tokens times price times volume, and a live calculator shows why it is almost always the smallest number in the room, next to the staff minutes it replaces.
Vendors spend weeks wordsmithing prompts when answer quality is mostly decided by what the model can see: the right chart entries, protocols, and definitions at the right moment. What context engineering is, why it quietly replaced prompt engineering as the real work, and why in a clinic it is also the privacy question.
The moment a clinic assistant reads text somebody outside wrote, a patient message, a faxed referral, that text can try to give it orders. A five-scenario quiz teaches you to spot prompt injection, and the three design moves that contain it even when detection fails.
More clinic AI projects stall on data than on models, but 'get our data ready' does not mean what most practices fear. A six-question scorecard tells you whether the records behind your first workflow are ready, and what to fix if they are not.
Every AI tool your clinic runs sits on a model you do not own, and it can change under you without ever throwing an error. A short scorecard shows how exposed you are, and the fix costs far less than the migration you are picturing.
Your staff already made the call on AI. Not you, not a partners' meeting, one paste into a personal chat window at a time. In a clinic each of those pastes is an unlogged disclosure of personal health information. A short exposure check shows where you stand, and the fix is smaller than a policy document.
Doing nothing feels free. It is a purchase your practice re-makes every month, on a line item that never shows up in the overhead. A live meter that sizes the bleed from your own numbers, honestly, then points at the one leak to stop first.
In one month, Microsoft, Google, and OpenAI all launched businesses built to deploy AI, not to train bigger models. That pivot is a tell, and it changes where a practice should point its own effort next.
A live ROI calculator that models both sides: time saved and what the AI costs. See net value, payback period, and the one number most clinic business cases get wrong, which is how much saved time actually becomes appointments.
An interactive quiz. Judge five AI answers as grounded or hallucinated, learn why a clinician cannot reliably tell from the text alone, and what actually catches it before it reaches a chart.
A practice does not adopt AI in one leap. A staged ladder from first quick win to deeper automation, where each rung earns the trust and the budget for the next, and where the top rung stays empty for anything clinical.
The hardest part of AI in a practice is not the model, it is choosing where to begin. A simple way to pick a first workflow that is high-value, low-risk, and proves itself fast, and the three candidates a clinic should never start with.
A scribe that demos beautifully can still be wrong in ways nobody sees until it is in a chart. Evals are how a practice measures whether a tool actually works, before it goes live and every month after. A non-technical guide to doing it honestly.
Clinic software fails at the front desk far more often than it fails in the architecture. Here are the five real reasons a working tool sits unused in a practice, and a practical way to close the gap between paid for and actually used.
Most AI advice assumes a blank slate. A clinic is the opposite: an EMR, a decade of records, and staff with habits. The honest choice is three-way, and for a practice one of the three is almost always wrong.
The top reason a practice stalls on AI is not cost, it is not knowing where the chart goes. Here is how to add intelligence while personal health information stays in your environment and in Canada, plus the seven questions to ask any vendor before you sign.
Stop searching, start asking. How a grounded assistant reads your own protocols, letters and policies and answers in plain language with the source attached, instead of handing your staff a folder to scroll through between patients.
The clipboard in the waiting room was a workaround for software that could not understand a sentence. Generative, conversational interfaces let a patient describe what is wrong and let the system assemble the intake, the chart entry, and the screen your staff actually needs.
A pilot is not a demo, and in healthcare most pilots never become anything else. Here is what a real proof of value measures in a practice, how to scope it to one workflow and one metric, and what a fair price looks like before you commit.
One drafts beside your staff, the other runs the process end to end. A simple way to tell which a clinic workflow needs, and how to sequence from copilot to agent without losing a clinician's trust.
Most clinic AI projects fail on adoption, not models. Use this checklist to choose a partner who ships something your staff actually use, signs as your agent under PHIPA, and owns the outcome with you.
You cannot safely bolt AI onto a system nobody dares to touch, and in a clinic the thing you dare not break is a patient record. Modernize the foundation first, with tests that lock in current behavior, then make it intelligent.
Not every clinic workflow should have AI in it. A simple framework for deciding where intelligence earns its place, where a plain rule still wins, and where the answer is a firm no.
The conference demo dazzles, the practice signs, and six months later nobody opens it. The gap is rarely the model, it is trust, control, and design. Here is how to cross it before you sign.
Nothing in the archive matches that filter.
The 14 places a practice loses hours inside software it already pays for, and what each becomes when a step is removed. Every number in it is sourced.
You get the occasional post by email too. Your data stays yours, never shared or sold.
A free, no-obligation consult. One clinic workflow gets mapped and ranked by what it pays back, and patient information never leaves Canada.
The founder is on every engagement. Specialists join for the stage that needs them, never an account manager.