01 the business
Physiotherapy and sports injury clinic
3 sites, 25 practitioners
New patients send online forms, emailed referral letters and photos of paper sheets. Reception retypes them into the practice system. Missing referrals or consent forms come to light on the day, and appointments start late or get moved.
02 before
The problem
Front-desk staff retype details from online forms, emailed scans and paper sheets into the practice system. Missing referrals, ID or signed consent come to light on the day, and the appointment starts late or gets moved. The same questions get asked twice.
- Details are typed twice, with typos.
- Missing documents are found at check-in.
- Patients are asked the same questions again in the room.
03 the build
What the AI does
- Step 1: Picks up new intake forms from Jotform or Typeform and documents sent to the shared inbox.
- Step 2: Reads names, contact details, history and reason for visit from forms and scanned files.
- Step 3: Checks the file against your checklist for that service, such as referral letter, ID or signed consent.
- Step 4: Emails the client a friendly request for anything missing, with a secure upload link.
- Step 5: Creates or updates the record in Cliniko, Clio or Karbon and attaches the documents.
paused: waiting for a person to approve
How it was put together
- Read new forms and documents from the intake tools and shared inbox.
- Extracted details and checked the file against a checklist for each service.
- Requested anything missing with a secure upload link.
- Prepared the record in the practice system for reception to confirm.
04 human in the loop
Where a person stays in charge
Front-desk staff confirm each new record before it is created and review any field the system was unsure about. Clinical or legal judgement on the intake stays with the practitioner.
- Reception confirms every new record before it is created.
- Health data stays in approved systems, with access logs.
- Clinical judgement about the intake stays with the practitioner.
05 rollout
From kickoff to live in 5 weeks
- Weeks 1–2Map forms, checklists and data protection requirements.
- Weeks 3–4Process real intakes in parallel with reception.
- Week 5Switch to review-and-confirm for all new patients.
06 after
What changes
Files are complete before the client walks in. Reception spends less time retyping and more time with the people in front of them.
Assume 400 new patients a month and 12 minutes of reception time each to retype and chase, falling to 3 minutes to confirm.
400 × 12 minutes = 80 hours a month today. 400 × 3 minutes = 20 hours. 60 hours a month go back to the front desk.
What to measure
- Share of first appointments with a complete file.
- Reception time per new patient.
- Appointments delayed by missing documents.
- Data entry errors found later.
07 evidence
What the research says
About 75% of generative AI's potential value sits in four areas: customer operations, marketing and sales, software engineering, and R&D.
The global average cost of a data breach reached $4.99 million, a record high.
IBM, Cost of a Data Breach Report (opens in a new tab) · 2026
08 lessons
Mistakes to avoid
- Skipping the privacy review. Health data needs care and the right agreements.
- Over-chasing. One friendly reminder and one follow-up is usually enough.
- Letting the system decide clinical suitability.
09 questions
Common questions
- Does it work for law and accounting firms?
- Yes. The checklist changes to ID, engagement letters and client documents, and the record goes to Clio or Karbon.
- Can it read handwriting?
- Often, from clear photos. Anything it is unsure of is flagged for reception.