Building Quantifiable Confidence with AI: Evidence from our AI Hackathon
How a one-day hackathon helped our team boost their confidence in using large language models (LLMs) by 17%.
At The PSC, we’re committed to helping the UK public sector navigate the opportunities and challenges of artificial intelligence. But to advise confidently, we need to be hands-on ourselves. At one of our recent ‘Away Days’ – set in the beautiful surrounds of De Vere Horsley Estate in Surrey – we ran an AI Hackathon designed to help our team explore how large language models (LLMs) such as ChatGPT and Copilot could be used more effectively in our day-to-day work.
What we found before we started
Our survey showed that AI tools are already firmly embedded in the way we work:
- 43% of colleagues were using them daily.
- 38% used them weekly.
- Only one person said they had tried them but didn’t plan to continue.
LLMs were being applied across a wide range of tasks, from research and writing to generating Excel formulas.
What changed during the hackathon
The hackathon gave us a chance to experiment together, test ideas, and learn from each other. One of the most striking outcomes was the increase in confidence. Before the session, the average self-rated confidence in using LLMs was 6.5/10; afterwards, it had risen to 7.6/10 — a 17% increase.
Even more telling, the proportion of people who rated their confidence at 7 or above jumped from 51% beforehand to 89% afterwards. That shift came not from theory, but from hands-on problem solving and seeing peers approach challenges in different ways.
New directions and use cases
The hackathon also opened up more ambitious thinking about AI, with team members identifying 56 different use cases to deploy in future projects. Beforehand, most uses were individual productivity aids: drafting, editing, research, and quick fixes in Excel. After the session, people were thinking bigger:
- Using LLMs for deeper synthesis of data and evidence.
- Enlisting LLM to undertake critical review of their work, and provide suggestions for feedback.
- Deploying AI for specialist tasks like contract review, coding prototypes of digital services, and structured red-teaming of outputs.
- Embedding AI into team workflows, from meeting reviews to training design.
The mindset shifted from “what can I do more quickly on my own?” to “how can we reimagine our ways of working together?”
Why this matters
For us, the hackathon was about building confidence, sparking new ideas, and strengthening peer learning. Of course, the meme-themed stickers added to the excitement, but the real buzz came from seeing colleagues push beyond their comfort zones and come away with new skills.
With the right format, an AI Hackathon can significantly advance an organisation’s confidence with AI in just one day, helping staff move beyond cautious trial-and-error to collective innovation.
At The PSC, we’ll continue exploring how to embed AI responsibly and strategically across our work. And we’re excited about the potential to help public sector teams do the same.
If you’re interested in running an AI hackathon in your organisation, please get in touch with Antonio.Weiss@thepsc.co.uk.
Latest News & Insights.
The Next 20 Live: Making AI work in public services - from opportunity to impact
AI is already shaping public services – but its value depends on how it is…
Improving Cultures of Care to Transform Productivity and Safety
Our newly published white paper, grounded in evidence from 163 inpatient wards,…
The Next 20: What we’ve learned so far – public value design, innovation, and the next steps
After eight episodes of The Next 20, we reflect on the themes shaping public…