NHS training system transformed with AI development platform

At the National School of Healthcare Science, we design and manage training programmes for healthcare scientists across the NHS, from audiology to vascular science and everything in between. We train over 2000 people a year at master's and doctorate levels, preparing them to become registered clinical scientists and consultant clinical scientists.

For years, we managed all this data across fragmented spreadsheets and an Access database. The problems were predictable: data duplicated across systems, manual entry, and over-dependence on a few people to provide information to others. When COVID-19 hit, and we couldn't access the office, the limitations became stark.

The real pressure point was around trainees in difficulty. Our training support team was managing complex data about extensions, appeals, changes of circumstance, and pastoral support in spreadsheets. We were keeping records about records.

And all the while, we'd been waiting six years to be onboarded to an internal NHS system that still wasn't finished.

We tendered for discovery work, and Torchbox's proposal stood out for its thoroughness; they didn't just want to understand user needs, they wanted to evaluate different technical approaches.

I'd assumed we could buy something off the shelf. The discovery work proved me wrong. What Torchbox recommended was an AI development platform: OutSystems.

When we tendered for the alpha and beta work, one detail in Torchbox's approach really stood out: they would track all work against Government Digital Service standards. In the NHS, where it's hard to get money spent and approval to spend it, this evidence proved crucial for gaining certain necessary approvals.

But what really set Torchbox apart was their approach. No grand promises or marketing claims. Just pragmatism, recommending solutions but committing to test them. Evidence-based hypothesis testing. After being burnt before, this felt different.

We went from Alpha to live in 36 weeks. Record time for a service like this in the NHS.

Interactive prototypes accelerated the project. When people could see something working (HTML mock-ups, reports in Google tools), they understood workflows better and gave better feedback faster.

Getting user feedback can be challenging in a busy environment, so testing working functionality rather than commenting on sketches made all the difference.

The system has normalised so quickly that people can't remember a time before it.

For case managers, it's transformed their work. They can filter on their caseload, see the status of a case, and view the full trainee record (academic history, support case history) all in one place.

One standout feature is the demographic analysis tool. You can look at any grouping of trainees and immediately see demographic composition by sex, gender, ethnicity, disability, and sexual orientation. We can pull out trends that were impossible to see when the data was stuck in spreadsheets.

Since launch, we've done swift enhancements, plugging gaps and weaning people off the last legacy spreadsheets.

The training management system tells us how each trainee is progressing. Now we need the same for the training delivery itself.

Our accreditation team ensures training quality across all providers, currently using spreadsheets. We're extending the system to support their work too.

Lessons learned

• Don't assume you can buy off the shelf. Your context is probably more specific than you think.
• Track against standards from the start. In the public sector, this is your path through bureaucracy and your evidence base for funding approvals.
• Value honesty over promises. Evidence-based hypothesis testing beats marketing claims.
• Get people testing working functionality early. Nothing beats hands-on experience for generating good feedback.
• AI development platforms can deliver at pace without compromising quality when you have the right partner who understands both the technology and your context.

What made this work was the combination of rigour and honesty, the thoroughness of approach, and the willingness to understand our context while challenging us to work better.

By the numbers

• 36 weeks: Time from alpha to live service
• 18 sprints: Two-week development cycles
• 2000+ trainees and several thousand trainers: On multiple different programmes lasting from 1 to 5 years
• 30+ specialties: From audiology to vascular science