Healthcare administration in Vancouver is drowning in paperwork. I've worked with three clinics in the past eight months, and the story is always the same: front-desk staff spend 40% of their day on data entry, appointment confirmations get missed because there's no good follow-up system, and patient intake forms still live in filing cabinets or badly designed PDFs that need manual transcription.
The irony is that most of this work follows strict, repeatable patterns. Insurance forms ask the same questions. Appointment reminders use the same language. Consent documents have the same structure. It's exactly the kind of work Claude Code was built to handle — and the clinics I've helped automate these workflows have seen immediate, measurable relief.
Here's how Claude Code is being used in Vancouver healthcare right now, what's working, and what the legal and privacy constraints actually look like in practice.
Patient Intake Automation That Actually Works
The first thing I automate for any healthcare client is patient intake. Not the clinical assessment — that still requires a licensed professional — but the administrative data collection that happens before a patient ever sees a doctor.
Most clinics still use paper forms or basic online forms that dump responses into an email inbox. Someone then has to manually transfer that information into the practice management system. It's a 5–10 minute task per patient, and when you're seeing 30+ patients a day, that's hours of work that could be eliminated.
The solution I've built with Claude Code takes a different approach:
- Patient fills out a web form (custom-built, hosted on the clinic's domain, encrypted in transit)
- Claude Code validates the responses in real-time — flags missing fields, checks date formats, ensures BC health numbers match the expected pattern
- Once submitted, the data is structured into the exact format the practice management system expects
- A CSV or JSON export gets generated automatically, ready to import with one click
For one clinic in Kitsilano, this cut intake processing time from an average of 8 minutes per patient to under 90 seconds. Over a month, that's 20+ hours of administrative time returned to patient care.
Appointment Workflows and Follow-Up Sequences
Missed appointments are expensive. The average no-show costs a Vancouver clinic around $200 in lost revenue, and most clinics see a 10–15% no-show rate. The usual fix is manual phone calls or text reminders, which work but require someone to remember to send them.
I've built automated appointment reminder sequences using Claude Code that integrate with most scheduling systems. The workflow looks like this:
- When an appointment is booked, Claude Code logs it in a local database
- Three days before the appointment, it sends a confirmation email or SMS (depending on patient preference)
- If the patient doesn't confirm within 24 hours, a second reminder goes out
- One day before, a final reminder with directions and pre-appointment instructions
The messages are personalized with the patient's name, the provider's name, and the appointment type. They don't sound robotic because Claude Code generates natural language that matches the clinic's tone.
One family practice I worked with reduced their no-show rate from 12% to under 4% in two months after implementing this. That's roughly $8,000 in recovered revenue per month for a mid-sized practice.
Compliance Documentation and PHIA Requirements
This is where a lot of people get nervous about using AI in healthcare, and for good reason. In British Columbia, patient health information is protected under the Personal Health Information Access and Protection of Privacy Act (PHIA). Any system that handles this data has to meet strict security and consent requirements.
Here's what that means in practice for Claude Code implementations:
- No patient data is sent to Anthropic's servers — all processing happens locally or on the clinic's own infrastructure
- Any data storage uses encrypted databases with access controls limited to authorized staff
- Audit logs track every time patient information is accessed or modified
- Consent forms are generated dynamically but reviewed by a human before being presented to a patient
The biggest use case I've seen for compliance is generating customized consent forms. Different procedures require different consent language, and manually editing templates for each case is time-consuming and error-prone.
I built a tool that takes a procedure code and patient information, then generates a consent form with the correct legal language, risks, and patient-specific details pre-filled. A clinician reviews it, makes any necessary edits, and then it's ready to print or send electronically. What used to take 15 minutes now takes 3.
This isn't about replacing clinical judgment. It's about making sure the administrative scaffolding around that judgment doesn't slow things down.
Referral Letter and Medical Documentation
Referral letters are another high-volume, high-friction task. A GP might write 10–20 referrals a week, each one needing patient history, test results, current medications, and a summary of why the referral is being made. The format is standardized, but the content is unique to each patient.
Claude Code can draft referral letters from structured input. The doctor provides bullet points — "42F, Type 2 diabetes, HbA1c 8.2, needs endocrinology consult for insulin initiation" — and Claude Code turns that into a properly formatted letter with all the standard sections filled in.
The doctor still reviews every word before it goes out. But instead of starting from a blank page or copying and pasting from old letters, they're editing a 90% complete draft. That's the difference between 10 minutes per letter and 3 minutes per letter.
For specialists who receive these referrals, I've also built intake parsers that extract key information from unstructured referral letters and populate their own intake forms. It's a two-sided efficiency gain.
What Claude Code Can't Do in Healthcare
I want to be very clear about the boundaries here, because there's a lot of hype around AI in medicine that glosses over the hard limits.
Claude Code is not a diagnostic tool. It doesn't interpret lab results, suggest treatment plans, or make clinical decisions. Any healthcare provider who tries to use it that way is both violating best practices and potentially breaking the law.
It also can't replace the human judgment required for patient interaction. A good intake coordinator knows when a patient sounds confused or distressed and escalates appropriately. An automated system doesn't pick up on tone or context the same way.
And it's not a magic fix for bad processes. If your clinic's scheduling system is a mess, automating it will just make the mess faster. You still need to design the workflow properly before you automate it.
What Claude Code does well is handle the structured, repeatable administrative work that eats up time but doesn't require medical expertise. That's a big category, and it's where most of the efficiency gains live.
Getting Started If You Run a Clinic
If you're a clinic administrator or healthcare provider in Vancouver and you're curious about this, here's how I'd recommend starting:
- Pick one high-volume, low-complexity task — appointment reminders or patient intake are the easiest wins
- Map out your current workflow in detail, including every manual step and handoff
- Identify which steps are purely data transfer or formatting — those are the automation candidates
- Run a small pilot with 20–30 patients to validate that the quality is consistent
Don't try to automate everything at once. The clinics that see the best results are the ones that start with one workflow, prove it works, and then expand from there.
If you want to see how this might apply to your specific practice, I walk through real examples on this getting-started guide. And if you're wondering whether your use case is feasible under PHIA, the FAQ page covers most of the common compliance questions I get asked.
For more on how I approach automation projects in general, here's a breakdown of what I've built across different industries, including healthcare. And if you're comparing Claude Code to other automation tools, this comparison with Zapier might be helpful — healthcare workflows often need more customization than Zapier can handle.
The tools exist. The question is whether you're ready to give your staff their time back.