Embracing scepticism, earning trust: lessons from Sport England’s Data & AI Lab

By Cyprian Kucaj

A recent study by researchers at King’s College London found that ‘half of the public (48%) would rather avoid AI, 41% are afraid of it and only 24% think it’s positive for humanity’. 

It should come as no surprise then, that the idea of artificial intelligence in the workplace often generates a mixture of excitement and scepticism. While some are keen to explore its potential, others worry about its reliability, its trustworthiness, and the impact it might have on their careers. 

AI scepticism is an asset, not an obstacle

This isn’t irrational. Scepticism is a normal psychological response to uncertainty and perceived risk, and AI can often feel like a black box: its decision-making process closed off, its outputs potentially flawed. In large organisations, data can become fragmented and inconsistent, causing uncertainty about AI to rise and reinforcing doubt in new digital tools. However, approaching AI critically and with an acknowledgement of its limitations is an essential step in responsible, effective implementation; staff scepticism shouldn’t be seen as something to overcome, but as an important part of the process.

This was the rationale which informed our work with Sport England, where we developed a Data & AI Lab and data strategy, focusing on building AI solutions in partnership with staff rather than for them. 

The challenge for Sport England

Sport England is an arms-length government body which supports grassroots sport and physical activity. Its strategy commits it to helping historically marginalised communities into sport. With little interoperable or real-time data, slow and duplicative reporting, and clunky legacy systems, Sport England was struggling to understand its impact and justify its decisions to government.

In the belief that AI tools should enhance what people do, not do it for them, the Data & AI Lab was set up in order to produce prototypes; this helped people across Sport England to see what was actually achievable and built upon the work already started by some teams internally. Ultimately, the approach was successful in generating a cultural shift within the organisation – “It opened my mind to what we could do with the tools that we’ve got. It’s really exciting” – but what is it about this incremental, transparent style of deployment which makes it effective?

Trust Calibration Theory in practice

One reason might be found in the Trust Calibration Theory, in which trust is adjusted based on the perceived reliability of systems (like AI), and scepticism arises when users have insufficient evidence of reliability (for instance, where past experiences have included errors or failures).  The interaction between a person and a system is a fluid one, constantly readjusting itself based on new outputs. 

For example, staff at Sport England are responsible for processing lengthy applications for their grants. By designing a prototype that used the Copilot model, we were able to help staff process these applications faster. Upon exposure to prototypes such as these, in which AI supports expert staff, people began to move from a position of doubt to one of curiosity. As a result, they were much more willing to invest time and effort into exploring the potential of AI with us.

Earning trust over time, and why it works 

What this highlights is that trust is earned through repeated positive experiences and feedback loops. At Sport England, nobody was told to trust AI; instead, trust was earned incrementally, through testing, observing and developing effective solutions in partnership with staff. This approach produced real results: teams began tackling shared problems together, and senior leadership committed to a full digital transformation programme. Designing systems and processes that enable trust to be calibrated over time is perhaps not the quickest or most straightforward approach, but as we saw on this project, it is one that works.

Grid image: “Last Mile” by Norman Meyer on Unsplash