Case study · 2017–2020
The national digital triage service for England. When people have an urgent medical concern, 111 online walks them through clinically safe questions and points them toward the right care: A&E, their GP, a clinical callback, or how to self-care at home. I joined in private beta after the service's GDS assessment came back “not met” on accessibility, helped get it back to standard, and stayed through rollout to a live service handling tens of millions of triage sessions a year. Highlights of my time there included the first integration with the NHS App and building an online emergency prescriptions pathway.
111 online is for when something’s wrong and you’re not sure what to do. It needs to be clinically safe, quick to load, and easy to use when someone’s stressed or not thinking clearly. It has to work for everyone, everywhere, including people on low-end devices, poor signal, or using assistive tech.
Underneath the service is NHS Pathways, the same clinical triage used on the 111 phone line. I joined in 2017 after the team's GDS beta assessment came back “not met” on accessibility. My first job was helping get it back to standard: we passed beta reassessment six months later, and the service went live nationally a year and a half after that. My focus across all of it was to make sure the technical choices we made around it never compromised the trust a user is placing in us when they answer honestly about their symptoms.
One milestone from that era I still think about: shipping the very first integration between 111 online and the NHS App, when the NHS App was brand new. The NHS App is now how most people in England access NHS services.
People arriving on the site are often frightened. Forcing a consent modal on them, so a third-party tracker could watch them, was the wrong trade. We stripped Analytics, removed the cookie banner, rewrote the privacy policy, and built a first-party, non-personally-identifying event system on our existing audit database with a PowerBI dashboard on top.
A research survey shipped by a third-party tool passed an automated WCAG check and still failed screen-reader users badly: duplicated labels on every checkbox, missing focus states. From then on, every new tool on the site got a manual run-through with a screen reader before it went near real users. Automated checkers are a floor, not a ceiling.
Health software doesn't forgive late failure. We kept stories small enough to ship behind a flag and measure, worked in a genuinely multi-disciplinary team (clinical safety, research, product, engineering all in the same room), and used retrospectives to look at the behaviour behind outcomes, not just the outcomes. The Service Standard assessments were a check on that, not a hoop to jump through.
Starting when the service had around a thousand users and leaving with it at tens of millions taught me something most side-projects can't: the architectural and ethical choices you make at small scale have to still hold at national scale. The decisions I'm proudest of only look more important in hindsight.
The best work I did on 111 wasn't a feature. It was the group of small, unfashionable decisions that said: this service belongs to the people typing into it, not to us. Strip the tracker. Test the survey with a screen reader. Don't let an analytics dashboard be worth more than a user's trust.
The lesson that stays with me: when you notice a quiet voice saying “we could just delete this”, listen. Doing the easy thing is the default everywhere. Choosing the harder, more honest thing is almost always defensible in hindsight, and it compounds. I wrote about the feeling of making that kind of call in How it feels to make the right decision.