So you've learned to build. This page is about using those skills in your working life without ending up in front of a panel. Most AI courses skip this part entirely. I do clinical safety for a living, so here it is, plainly.
Three lights. Green means build freely. Amber means build with care. Red means stop, because you've crossed into territory that a weekend build cannot carry. Knowing where the lines are is what lets you build with confidence, and the good news is that the green zone is enormous.
🟢 Green: build freely
No patient data and no clinical decision-making means no regulator is interested in your project. This covers more than people expect: personal professional tools, teaching materials, and your own admin.
A logbook for your appraisal and revalidation. Quiz banks and flashcards for your students. A skills checklist for preceptorship sign-off, built around roles rather than named people. A personal planner for your own rota (your rota; your colleagues' shifts are their information, not yours). A professional website. Every one of these makes your working life better, none of them touches a patient, and you can start any of them tonight.
🟠 Amber: build with care
Amber is clinical content with no patient data and no decision support. Patient information leaflets. Teaching cases, fully made up. Guideline summaries for your own revision. An audit dashboard running on synthetic data.
The risk here is accuracy and currency, not regulation. AI writes confident clinical prose that is sometimes wrong or out of date, and your professional registration stands behind anything you publish or hand to a patient. So the rule is the one you already know from the course: check every claim against the source before you trust it. That checking is the real work, and it is exactly the skill the course drills.
🔴 Red: stop
Patient data never goes in. Not into a chat, not into a prompt, not into an app you host. "Anonymised-ish" doesn't count either: small numbers re-identify people easily, and a rare condition plus an age plus a rough location is often enough to name someone. If you want realistic data to build with, generate synthetic data. It works brilliantly and it is what my own demo dashboards use.
Watch where the data goes. Most AI tools process what you type on servers in the United States. Your employer's policy, and UK data protection law, may bar even legitimate work data from leaving the UK or Europe. Check before you paste. This is also where UK and US colleagues differ: the UK works to UK GDPR and the NHS Data Security and Protection Toolkit, the US to HIPAA. Same instinct, different rulebooks.
The medical device line. Anything that informs or drives the care of a real patient (calculates a dose, scores a risk, suggests a diagnosis, triages, monitors) is software as a medical device under UK medical device regulations, regardless of how small it is or how good your intentions are. That means UKCA marking, a quality management system, clinical evaluation, and a legal manufacturer standing behind it. A weekend build cannot carry that, and it isn't meant to. The skill is recognising when an idea crosses the line: crossing it is a different, much bigger project, with formal clinical safety standards (DCB0129 for the manufacturer, DCB0160 for the organisation deploying it) and a real Clinical Safety Officer involved.
Deployment escalates everything. "Works on my computer, for me" is a safe and respectable stopping point, and the course teaches it as one. The moment colleagues start using your tool in a care pathway, your organisation picks up formal safety obligations (that DCB0160 standard again) even if the tool itself is not a device. Sharing something at work is a decision, not a default.
What to build: the clinical project gallery
Real project ideas with their traffic light. Everything green or amber here is fair game for your first builds, and the course covers the skills each one needs.
Swipe the table sideways to read why each one is a good build →
| Project | Light | Why it's a good build |
|---|---|---|
| Professional website or portfolio | 🟢 Green | Every clinician is told to have one. This site is the live worked example. |
| Appraisal and revalidation logbook (CPD hours, reflective notes, zero patient identifiers) | 🟢 Green | A universal pain point for doctors and nurses alike. Try the live demo, built the way the course teaches. |
| Teaching materials: quiz banks, flashcards, OSCE and skills checklists | 🟢 Green | The ideal first build for anyone who teaches. Instant use, zero risk. |
| Preceptorship or competency sign-off tracker (roles, not names) | 🟢 Green | Nurse educators especially: this is a rota of skills, not of people. |
| Personal rota and on-call planner | 🟢 Green | Your own rota only. Colleagues' shifts are their personal information. |
| Audit dashboard on synthetic data | 🟠 Amber | Teaches the synthetic-data habit that keeps you safe for life. My grocery dashboard models the exact pattern. |
| Journal club summariser or guideline revision cards (own use) | 🟠 Amber | The build is easy; checking the AI's accuracy against the source is the real skill it teaches. |
| Patient information leaflet drafter | 🟠 Amber | Teaches accountability: you own every word the patient reads, not the AI. |
| 🔴 Red | Shown deliberately as the "this is a medical device" examples. Not built, and that's the lesson. |
A note for US readers
The traffic lights translate directly; the rulebooks differ. Where a UK clinician thinks UK GDPR and the NHS Data Security and Protection Toolkit, you think HIPAA. Where the UK medical device line is UK MDR with UKCA marking, yours is FDA regulation of software as a medical device. The instinct to learn is identical: patient data never goes into consumer AI tools, and anything that informs a real patient's care is a regulated product, not a side project.
Want this taught live?
The workshop covers this safety layer alongside the building itself: you make something real for your working life, and learn where the lines are as you go, from a doctor who does clinical safety for a living.
See the workshop