AI for healthcare clinics
AI for healthcare clinics
Documentation, intake, and the admin behind every visit, off your clinicians' plates. 5x ROI in 30 days, or we work free.
- Family practice
- Specialist clinics
- Allied health
- Home care agencies
- Walk-in and urgent care
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Teams we build for
- Hoyes Michalos
- Nurse Next Door
- Fedi
- UBC Sauder
- Merchant House Capital
- Picton Investments
- Campbell Froh May & Rice LLP
- Barnakl
- Hungerford
- Breez
Proof from the pattern
Our closest published work is a booking marketplace that lists clinics, not a clinic. The mechanics below are what we would build and measure with you.
We guarantee 5x ROI inside 30 days of deployment, in writing, measured against a baseline you sign before we build. If the system misses the bar, we keep working for free until it clears.
Charting follows clinicians home.
Notes finished at the kitchen table after clinic hours. Documentation burden is why good clinicians leave.
The front desk cannot keep up.
Phones, refills, reschedules and intake clipboards, all landing on the same two MOAs. The line out the door is the phone queue.
Billing leaks revenue quietly.
Fee codes picked from memory at 6pm and rejections nobody has time to rework. The money is earned. It is just never collected.
Prior auth still runs on a fax machine.
Special Authority requests and insurer forms assembled by hand, faxed, lost and re-faxed. The patient waits inside that loop.
No-shows hollow out the day sheet.
The 9:40 sits empty while the waitlist holds patients who would have taken it. Nobody has a hand free to make those calls.
The recall list is a guilt pile.
Bloodwork due, paps overdue, the follow-up the physician asked for in March. It gets worked in spare afternoons that never come.
Healthcare clinics, before and after
The manual path is dashed: Notes finished after hours, Front desk drowning in calls, Prior auth by fax and phone. The system path replaces it, and a person approves before anything ships: Draft notes ready to sign, Intake and recalls handled, Auth packets ready to send.
Before: by hand
- 01Notes finished after hourshuman
- 02Front desk drowning in callshuman
- 03Prior auth by fax and phonehuman
After: the system
- 01Draft notes ready to sign
- 02Intake and recalls handled
- 03Your approvalhuman
- 04Auth packets ready to send
The research
Canadian physicians spend about nine hours a week on administrative work and judge 47% of it unnecessary, 19.8 million hours a year (CMA and CFIB, 2026, 1,924 self-selected respondents).
Canadian Medical Association and Canadian Federation of Independent Business · 2026 · 1,924 practising physicians, residents and fellows, open-link voluntary survey fielded July 31 to August 21, 2025, published January 26, 2026
Built around your rules
| Regime | What it demands here | How the system complies |
|---|---|---|
| Provincial health privacy acts | PHIPA in Ontario, Alberta's HIA and the BC framework make the clinic the custodian of the record, answerable for every person and system that touches it. | The build operates as your service provider under your policies. The data-flow map is written with your privacy officer before anything ships, and access is scoped per role with audit logs. |
| PIPEDA and BC PIPA | Patient contact details, billing records and the recall list are personal information under federal and BC privacy law, collected and used within consent. | That information stays in clinic-controlled systems under no-training, no-retention API terms, and every outbound contact runs under consent rules the clinic sets. |
| College oversight of the record | The colleges of physicians and the allied-health colleges hold the clinician responsible for the chart and the care, regardless of what drafted either. | Every clinical-adjacent output is a draft with a named signer. Nothing writes to a chart or reaches a patient with clinical content until a clinician releases it. |
| Consent for ambient scribing | College and CMPA guidance expects patients to consent before an encounter is captured by an AI scribe. | Consent scripting is built into intake and the visit flow, capture starts only after a yes, and the note still leaves as the clinician's own signed record. |
| Data residency | Some clinics and some provinces require patient data to stay in Canada, and the requirement differs by act and by contract. | Residency is scoped per engagement and stated in writing. Where Canadian hosting is required, it is designed in before anything is built. |
| Safeguard and breach duties | PIPEDA and the provincial acts carry mandatory breach reporting with real timelines, and a clinic must be able to say who accessed what. | Access is logged at the database layer and the incident path is written down before go-live, so the report you may one day owe a commissioner can actually be written. |
The regimes that govern healthcare clinics, what each demands, and how the system complies
Regime
- Provincial health privacy acts
What it demands here
PHIPA in Ontario, Alberta's HIA and the BC framework make the clinic the custodian of the record, answerable for every person and system that touches it.
How the system complies
The build operates as your service provider under your policies. The data-flow map is written with your privacy officer before anything ships, and access is scoped per role with audit logs.
- PIPEDA and BC PIPA
What it demands here
Patient contact details, billing records and the recall list are personal information under federal and BC privacy law, collected and used within consent.
How the system complies
That information stays in clinic-controlled systems under no-training, no-retention API terms, and every outbound contact runs under consent rules the clinic sets.
- College oversight of the record
What it demands here
The colleges of physicians and the allied-health colleges hold the clinician responsible for the chart and the care, regardless of what drafted either.
How the system complies
Every clinical-adjacent output is a draft with a named signer. Nothing writes to a chart or reaches a patient with clinical content until a clinician releases it.
- Consent for ambient scribing
What it demands here
College and CMPA guidance expects patients to consent before an encounter is captured by an AI scribe.
How the system complies
Consent scripting is built into intake and the visit flow, capture starts only after a yes, and the note still leaves as the clinician's own signed record.
- Data residency
What it demands here
Some clinics and some provinces require patient data to stay in Canada, and the requirement differs by act and by contract.
How the system complies
Residency is scoped per engagement and stated in writing. Where Canadian hosting is required, it is designed in before anything is built.
- Safeguard and breach duties
What it demands here
PIPEDA and the provincial acts carry mandatory breach reporting with real timelines, and a clinic must be able to say who accessed what.
How the system complies
Access is logged at the database layer and the incident path is written down before go-live, so the report you may one day owe a commissioner can actually be written.
Patient information is handled to respect PIPEDA and provincial health-information law, with residency scoped per engagement and a clinician signing every clinical document.
EMR, Practice mgmt, Payer portals and the system run inside a boundary labeled your accounts. The only path that crosses the boundary is the audited egress to the model API, under no-training terms. The boundary is what answers Provincial health privacy acts and PIPEDA and BC PIPA.
Inside your accounts
- 01EMR
- 02Practice mgmt
- 03Payer portals
- 04System in your cloud
- 05Audit log
Outside, through the audited port
- 01Model API
The numbers in healthcare clinics
Canadian physicians spend about nine hours a week on administrative work and judge 47% of it unnecessary, 19.8 million hours a year (CMA and CFIB, 2026, 1,924 self-selected respondents).
Canadian Medical Association and Canadian Federation of Independent Business · 2026 · 1,924 practising physicians, residents and fellows, open-link voluntary survey fielded July 31 to August 21, 2025, published January 26, 2026
One health system's ambient AI scribes saved physicians an estimated 15,791 hours of documentation time (AMA, 2025).
American Medical Association · 2025
The largest multi-centre study puts the real saving at 16 minutes of documentation time per 8 hours of patient care, and calls it modest against vendor claims (JAMA, 2026, 1,800 clinicians).
JAMA · 2026 · 1,800 clinicians across 5 academic medical centres, 2023 to 2025
AI auto-approves up to 90% of prior-authorization requests for covered members (Cohere Health, 2025, vendor-reported).
Cohere Health · 2025
Who this is built for
You leave with the charts closed
The encounter note is drafted before you open the chart to sign it, and the refill queue arrives as decisions, not paperwork. What the published evidence supports is minutes back per clinic day, not hours, and we will say that to your face. The part that matters is where the minutes land: inside clinic hours instead of at your kitchen table.
Recalls run without a spare afternoon
The recall list has always been worked in the gaps, and there are no gaps. Now due patients are surfaced from the chart, contacted under rules you set, and booked into real openings, while anything flagged clinical routes to a person. The list shrinks on a schedule instead of on guilt.
New patients arrive already charted
Intake forms, consent and history are done at home before the first visit, and the WorkSafe and MVA paperwork that eats your evenings starts from a draft built out of the assessment your therapist already wrote. Your clinicians treat for the full appointment. The clipboard retires.
The data map exists before the tool does
You have watched staff paste patient details into free tools because the sanctioned path was slower. Here the data-flow map is written with you before anything ships: what the system reads, where it runs, what crosses to a model under no-training terms, and what never leaves the clinic. Access is logged per role, so your next audit answer is a query, not a shrug.
What stays human
- She works the exceptions, not the list
- Refills wait for a clinician's yes
- She reads the packet before it goes anywhere
The steps the day below leaves to a person, by design.
A clinic day, run from the front desk
The 9:40 cancellation fills itself
8:55A patient cancels at the desk. The system works the waitlist in the booking module, offers the slot under the clinic's contact rules, books the first yes, and pulls the chart forward. The day sheet stays full without anyone dialing.
She works the exceptions, not the list
10:10humanBetween check-ins she reviews what the morning surfaced: one recall flagged because the patient asked a clinical question, one intake form with an answer that needs a nurse's eyes. Two items, not forty. The rest ran under the rules the clinic wrote.
Intake walks in the door already done
11:30This afternoon's two new patients completed their Ocean forms at home. History, medications and consents are in the chart before the visit, and the answer that mentioned chest tightness was routed to the nurse the moment it was typed, not discovered at the counter.
Refills wait for a clinician's yes
13:15humanThe routine refill requests are prepped with the chart context attached. The physician approves most, declines one and books the patient in instead. No prescription moves on a machine's judgment. That line is architectural, not aspirational.
The Special Authority packet assembles from the chart
14:40A drug needs coverage approval. The request is drafted from the chart with the criteria the payer publishes, attachments gathered, ready to send. The fax-and-phone loop that used to take a week of remembering is now a review.
She reads the packet before it goes anywhere
14:55humanShe checks the packet against the patient in front of her, fixes a date, and sends it. Nothing reaches a payer without a person deciding it should.
Claims are scrubbed before they leave the building
17:05The day's billing is checked against fee codes and payer rules, mismatches flagged while the encounter is still fresh enough to fix, and yesterday's rejections queued with the reason attached. The desk goes home at five. So does the biller.
What we build for healthcare clinics
Before the visit
Patient intake before the visit.
Forms, histories and consents collected ahead of time, urgent flags routed to staff immediately.
After-hours phone coverage.
Scheduling and routine requests handled around the clock. Anything clinical escalates to a person.
Waitlist backfill for cancellations.
A cancelled slot is offered to the waitlist under your contact rules and booked before it costs the day anything.
In the room
Visit documentation support.
Notes drafted from the encounter for clinician review and sign-off. The clinician edits and owns every chart entry.
Recall and refill workflows.
Due patients surfaced and contacted, routine refill requests prepped for clinical approval.
After the visit
Prior-authorization prep.
Authorization packets assembled from the chart against the payer's own criteria. Staff reviews and submits.
Billing and claims scrubbing.
Claims checked before submission, rejections queued with the reason attached and prepped for rework.
The economics
Before and after economics
Line
- Admin hours in scope
Before
At the page defaults, 6 staff spending 10 hours a week on intake, recalls, prior auth and billing rework is 60 hours, $4,200 a week at $70 loaded cost
After
The calculator below prices your own clinic, and the 5x ROI guarantee is measured against a baseline you sign before we build
- After-hours charting
Before
Notes finished at the kitchen table after the last patient leaves
After
Draft notes wait inside clinic hours for a clinician's signature, and your saving gets measured, not quoted from a vendor deck
- The recall list
Before
Worked when someone finds a spare afternoon, which is never
After
Due patients surfaced and contacted on a schedule, clinical flags routed to a person, qualitative by design
- Rejected claims
Before
Pile up unworked because the desk is on the phones
After
Scrubbed before submission, rejections queued with the reason attached, measured against your own baseline
The only dollar figures in this table are arithmetic on this page's calculator defaults, 6 staff at 10 manual hours a week each at a $70 loaded hourly cost, which is 60 hours or $4,200 a week in scope. No clinic outcome numbers exist for this page yet. Our closest published case is a booking marketplace that lists clinics, and it measured pages published, not clinic hours, so no row borrows from it. Every other row is qualitative and the guarantee is measured against a baseline you sign before anything is built.
Where the data comes from
EMR
OSCAR Pro, TELUS PS Suite, Med Access or Accuro holds the chart, the medication list, the labs and the encounter history. It is where draft notes wait for a signature, where the recall queries read from, and where intake lands once a person accepts it. Where the vendor exposes an API the build uses it, and where it does not, the build works the admin around the chart and says so up front.
Where it stops. Nothing writes to a chart without a clinician releasing it. Patient information stays in clinic-controlled systems under the privacy act that applies to you, model calls run under no-training, no-retention terms, and access is scoped per role with an audit log your privacy officer can actually read.
Practice mgmt
The booking module of the EMR, or Jane on the allied side, Juvonno, or AlayaCare where the visits happen in living rooms. This is the day sheet, the waitlist, the provider schedules and the billing ledger, and it is where cancellations get backfilled and no-show patterns become visible instead of anecdotal.
Where it stops. The system fills openings and preps claims. It does not decide fees, write off balances or change a provider's template. Patient contact runs under consent rules the clinic sets, and every outbound message is one a person could have sent from the desk.
Payer portals
Teleplan in BC, the OHIP claim lanes in Ontario, WorkSafeBC, TELUS eClaims for extended health, and the Special Authority forms that still travel by fax. This is where claims, prior auths and third-party forms actually go, and it is the loop the build shrinks hardest, because it is pure assembly wrapped in waiting.
Where it stops. Prepared, never fired. Packets and claims are assembled and checked by the system and sent by a person. Portal credentials stay the clinic's own, browser automation reads and fills but does not submit, and no payer ever hears from an unsupervised machine on your behalf.
How the system is built for healthcare clinics
See the full capability mapRetrieval
Clinic protocols, payer rules and intake documents indexed so front desk and clinicians get the same answer. When the source does not cover the question, the system says so rather than inventing a policy.
- pgvector
- Full-text BM25
- Abstention on low confidence
Agents and orchestration
Agents handle recall, intake chase and claim preparation on a schedule. Anything touching a clinical record or a patient message is a proposal a person releases. Autonomy is capped low here on purpose.
- agent-worker
- Autonomy guard
- Proposal queue
Evaluation
Extraction and drafting are graded against records your team has already completed. Published research on ambient documentation reports a wide spread of time savings, so we measure yours instead of quoting a vendor's.
- Eval graders
- Baseline measurement
- quality-worker
Models
Document and speech models for intake forms, faxes and dictation. Frontier models for drafting. Ordinary classifiers for triage and coding suggestions, which are faster and easier to audit.
- Document and speech extraction
- Frontier models, one gateway, routed per task
- Classifiers
Data boundary
Health information stays in your accounts, scoped per clinic and per role at the database layer, under no-training API terms. Built to respect PHIPA and PIPEDA obligations from the first line.
- Supabase row-level security
- Per-clinic scoping
- Access audit logs
EMR, Practice mgmt, Payer portals feed a hybrid index. The agent runtime works from that index, and every consequential action passes a human approval before it reaches Workflow builder, Approval queue, Ask your brain.
Your systems
- 01EMR
- 02Practice mgmt
- 03Payer portals
The system
- 01Hybrid index
- 02Agent runtime
- 03Your approvalhuman
Where your team works
- 01Workflow builder
- 02Approval queue
- 03Ask your brain
The objections
Our college has guidance on AI scribes, and the CMPA has opinions.
Good, because the build follows both. The clinician stays responsible for every word in the record, the patient consents before any encounter is captured, and consent scripting is part of the intake flow, not an afterthought. Every clinical document is a draft with a named signer. Nothing in this system asks your college for forgiveness.
Half our patients are in their eighties. They will not fill in forms on an app.
They do not have to. The phone line answers calls, not just texts, the paper form that still arrives gets scanned and read into the chart, and the front desk keeps working exactly as it does for anyone who wants a human. The automation serves the patients who take it and buys the desk time for the patients who will not.
We are two MOAs and a part-time biller. We cannot run an IT project.
You are not being asked to. Engagements are fixed scope, quoted after a free audit, and we run the build. If the audit says the numbers cannot work at your panel size, we tell you that and nothing gets built. Small clinics are the ones with no slack to gamble, and the scope is sized so you do not have to.
An AI on the phones will miss the chest-pain call.
It is built to fail the other way. Anything clinical, urgent or unclear escalates to a person immediately, and the escalation rules are written with your clinicians before the line goes live. The after-hours agent books, reschedules and answers office questions. It never gives medical advice, and it treats doubt as an escalation, not a guess.
Scribe vendors promised us hours a day. Why should we believe you?
You should not believe anyone's demo. The largest multi-centre study, 1,800 clinicians across 5 academic medical centres, measured 16 minutes of documentation time saved per 8 hours of patient care and called it modest against vendor claims (JAMA, 2026). We quote that study on our own page, and we measure your baseline before and after instead of quoting anyone, including ourselves.
You will automate away the MOAs we are trying hard to keep.
The clinics we talk to are short-staffed, not over-staffed. The build takes the phone queue, the chase and the rework off the same two people who currently absorb all of it, which is the difference between an MOA who stays and one who burns out by spring. Nobody's job here is the fax machine.
How we build it for healthcare clinics
Fix the front desk before the clinical work.
Intake, recall and scheduling carry no clinical risk and consume most of the staffing gap. That is where the first build goes and where the baseline gets measured.
Bring documentation in with a measured baseline.
Published multi-centre evidence puts ambient scribe savings at 16 minutes per 8 hours of patient care, which is real but well below what vendors imply. We record your own baseline before and after rather than quoting anyone's number.
Add claims and coding support last.
Revenue-cycle work touches payer rules that change. It goes in once the eval harness is running and someone owns the alerts.
What we will not automate
Diagnosis, triage decisions and anything written to a clinical record without a clinician releasing it. Autonomy is capped lower here than anywhere else we work.
What that means in practice
Vision, documents and speech
AI for eyes and ears: reading documents, watching camera feeds, transcribing calls.
Where we stop. Bespoke computer vision is justified by volume and latency, not by novelty. Below that bar, a vision-language model on demand is cheaper to run and easier to maintain, and we will tell you which side of the bar you are on before anything is built.
Security, privacy and governance
Keeping your data yours, and your AI safe to put in front of customers.
Where we stop. We do not claim certifications we do not hold, and we will not ship a customer-facing agent without a human gate and an eval suite. If a vendor cannot offer no-training terms, it does not get into the stack.
Agents and orchestration
AI that does the work instead of just answering: looks things up, calls your systems, completes multi-step tasks, and knows when to hand off to a human.
Where we stop. Multi-agent swarms are oversold; most jobs need one well-guarded loop. If a cron job and a script solve it, that is what we build, because 90 percent per-step accuracy compounds to 59 percent over five chained steps and no framework changes that arithmetic.
Where your team works
Tour the platformWorkflow builder
Describe the workflow. Watch it assemble.
Say what should happen in plain language and the builder assembles the automation on a canvas you can read, run, and change.
Approval queue
The AI proposes. You approve. Nothing sends itself.
Every consequential action arrives as a proposal with the risk, reversibility, and expiry spelled out before you say yes. Control stays in the room.
Ask your brain
Ask a question. Get the answer and where it came from.
Every answer cites the document behind it, so you can check the source yourself. When the brain does not have the answer, it says so instead of guessing.
Task board
People and agents, working the same board.
Every piece of work lives on one shared board, whether a person or an agent owns it. Handoffs between the two are explicit, so nothing falls in the gap.
By team
The same system, seen from the desk that runs it.
- Customer supportPatient questions repeat and the privacy rules are strict.
- OperationsIntake documents, scheduling and compliance records are onboarding and reporting problems.
- HR and peopleCredential and certification tracking runs on the same watch and chase pattern.
- Legal and compliancePrivacy-regulated operations where attestation and vendor paperwork carry real exposure.
Run your numbers.
Your operations
Savings use the low end of our hours-reclaimed range. The math is conservative on purpose.
The math
Calculated at the low end of every range.
Every first build is covered in writing: 5x ROI in 30 days. Or we work for free.
The hard questions
The systems behind this
Operations hub→
Your numbers, one place, no Friday scramble.
Client onboarding→
Documents chased until they arrive.
Custom agents→
For the jobs only your business has.
Knowledge brain→
Cited answers from your own documents.
Front desk→
Every channel caught, nothing dropped.
Voice agent→
Calls answered, booked and filed.
Free · 3-5 days
Know your number in five days.
We map your operations, find the highest-ROI automations, and hand you a ranked plan with the payback math attached. Yours to keep, whoever builds it.
Prefer to talk first? Book 15 minutes with James. No pitch deck.