@northwindclinic
Northwind Clinic
AI intake assistant that cut patient admin time by 60%.
- Client
- Northwind Clinic
- Year
- 2024
- Services
- AI Automations
- Duration
- 4 months
- Industry
- Healthcare
Challenge
Staff at Northwind Clinic spent more time on intake paperwork than with patients. Details were collected on paper and by phone, then retyped into the records system by hand.
Solution
We built an AI intake assistant that talks with patients before their visit, collects history and reason for visit, and produces a structured summary for staff to review.
Outcome
Patient admin time per visit fell by 60%, returning clinical hours to patient care every week.
01 — Challenge
The challenge
Staff at Northwind Clinic spent more time on intake paperwork than with patients. Details were collected on paper and by phone, then retyped into the records system by hand.
Transcription errors were common, and missing information often surfaced only when the patient arrived, delaying appointments.
Any solution had to respect strict privacy rules and keep clinicians firmly in control of what entered a patient record.
02 — Solution
Our approach The solution
We built an AI intake assistant that talks with patients before their visit, collects history and reason for visit, and produces a structured summary for staff to review.
Language models handle the conversation and extraction, while deterministic code validates every field against the records schema. Nothing is written to a record until a staff member approves it.
- Conversational intake by web link or text message
- Structured extraction validated against clinic schemas
- One-click staff review and approval
- Audit log of every change and access
We ran the assistant in shadow mode for four weeks against manual intake, measuring accuracy before it handled live patients.
Results
Admin time saved
Patient admin time per visit
Intake accuracy
Patient details captured correctly at intake
03 — Outcome
The outcome
Patient admin time per visit fell by 60%, returning clinical hours to patient care every week.
Intake accuracy reached 99% on reviewed fields, and missing-information delays at the front desk largely disappeared.
The clinic now extends the same pattern to referrals and follow-up reminders.
Our front desk used to retype everything. Now they review a clean summary and approve it, and patients walk in with the paperwork already done.
Dr. Helena Marsh
Clinical Director · Northwind Clinic
Tech stack
- Python
- LLM APIs
- Postgres
Next project
Greenhome
Greenhome