From 9pm Notes to Home by 6: A Physiotherapist's AI Workflow

For years, the standard private physio day ended like this: last patient at 5:30pm, home by 6:30pm, dinner, then an hour of notes between 8:30pm and 9:30pm. Every day. Forever.

Most physios I talk to in 2026 have stopped doing that. Not because their caseload is smaller — if anything, it's bigger — but because their notes workflow has changed.

Here's what a realistic day looks like with AI doing the heavy lifting on documentation. One clinic, one physio, 10 patients.

7:45am — Open laptop

Morning coffee. Open the clinical system. Review today's 10 patients. The AI summary on the caseload dashboard shows:

"3 new patients. 7 follow-ups. 1 potential discharge (patient #6 — pain trending ↓ over last 3 sessions, last session reported 1/10 at worst). Patient #4 flagged for red-flag re-screen per last session's note."

Useful. Ten seconds to read, versus five minutes clicking through ten notes.

8:30am — Patient 1 (new)

New assessment. 45 minutes. I type nothing during the session. I used to write as I went, and I'd miss half of what patients said.

Now: I talk, I assess, I listen. At the end, I have 90 seconds to tap out bullet-point objective findings on my tablet — ROM numbers, positive tests, what I palpated. Subjective was noted in three or four lines while they were speaking.

Push "generate SOAP." The AI takes my bullets and produces a clean, professional SOAP note in 20 seconds. I read it. Edit one line (it misinterpreted "resisted abduction 4/5" — I fix it). Sign.

Time on notes for patient 1: under 3 minutes. Used to be 12.

9:30am — Patient 2 (follow-up)

30-minute follow-up. At the end, I record a 45-second voice note on the tablet:

"Session 4. Subjective improvement — pain down to 3/10 from 6. Sleep better. Objective — ROM flexion 110 up from 95, still limited end-range. Treated grade 3 PA glides L4-5, added hip hinge drill with band, home ex progressed to single-leg RDL 3x10. Patient responding well. Next session focus on gait retraining on gym treadmill. Plan to discharge in 2-3 sessions."

The AI transcribes, structures it into S-O-A-P, pulls in the exercise progression detail into the treatment log. I read it. Approve. Done.

Under 2 minutes.

10:00am — Patient 3 (new, complex)

Chronic low back pain, 8 years, seen four other physios. Complex subjective with psychosocial elements. I take longer subjective notes during the session itself — because I want to get the nuances right.

After the session, the AI structures my bullets into a proper SOAP note AND flags: "Consider yellow flag screening — fear-avoidance beliefs referenced in subjective. STarT Back tool recommended."

Good prompt. I hadn't formally STarT-Back'd yet. I do it. Document it.

11:00am — 12:30pm — Patients 4, 5, 6

Three back-to-back follow-ups. Same voice-note-at-the-end workflow. Total note-writing time across the three: 6 minutes.

12:30pm — Lunch

Actual lunch. Not eating a sandwich while typing notes.

While I eat, I look at the "outstanding admin" panel. Two patients need outcome letters to their GPs. The AI has already drafted both based on the last session's note. I read them, tweak a sentence each, send.

Two GP letters in five minutes. Used to be 30.

1:30pm — 4:30pm — Patients 7, 8, 9, 10

Four afternoon patients. Same workflow. Patient 10 is the potential discharge. I spend the last 10 minutes of the session doing a proper outcome measure (LEFS, this one) and reviewing progress.

The AI pulls together a discharge summary from the last 6 sessions — pain trajectory, ROM progression, outcome measure scores at start and end. I review, edit, finalise. Discharge letter to GP drafted automatically.

Full discharge documentation: 8 minutes.

5:00pm — Admin tidy

Everything is signed. Everything is filed. I check the end-of-day view:

• 10 sessions documented, 10 signed
• 2 GP letters sent
• 1 discharge letter sent
• 0 notes in "draft" status

Close the laptop. Walk out the door.

5:30pm — Home

Not 9pm. Not 10pm. 5:30pm.

What changed, precisely

Before AI: 10 patients × 12 min notes = 120 minutes. Plus 30–60 min of GP letters and admin = roughly 2.5 hours of after-work documentation per day.

Now: 10 patients × 2–3 min notes = 25 minutes. GP letters and admin: 15 minutes. Total 40 minutes — all done inside clinic hours.

That's roughly 10 hours a week reclaimed. Across a year, that's 500 hours. That's a part-time second income, or time with your kids, or just not resenting your job.

The catch (being honest)

This isn't magic. The workflow only works if:

1. You trust the tool enough to use it for voice notes, but not so much you stop reading every SOAP note it generates. Always read. Always edit. Always sign.
2. You build the habit of clear dictation or bullet-point input. "Um, so they had a bit of pain, I think on the right side, maybe 6 out of 10 but it could be worse on Tuesdays" produces a worse note than bullet points. Garbage in, garbage out.
3. You don't use a consumer AI (like ChatGPT) for this. Patient data goes in a purpose-built, GDPR-compliant, clinical-AI tool. Full stop.
4. You don't skip the clinical reasoning. The AI writes the words. You still think about the patient.

The question I get asked most

"Doesn't it feel cold? Isn't it like the patient isn't getting your full attention?"

Opposite. When I stopped typing during sessions, my clinical relationships got better. I make eye contact the whole session. I notice the grimace when they mention their dad's cancer. I pick up on the throwaway comment about sleep.

The AI handles the typing. I handle the physio. The patient gets a better session AND a better note. That's the actual pitch.

If you want to try this

Oris is one option — built specifically for this workflow, voice-note-to-SOAP in under 20 seconds, UK-hosted, GDPR-compliant, free on the Starter plan. Other clinical AI tools exist; test a few. The important thing is: if you're still doing notes at 9pm in 2026, the tools exist to stop you having to.

Your evenings are worth getting back.