Trust, not technology, is the barrier to scaling AI
TL;DR:
- Public-sector bodies are growing more confident with AI, with more pilots moving into live production, according to Belfast-based Kainos.
- Its head of responsible AI, Theresa Yurkewich Hoffmann, says the biggest barrier to scaling is a lack of trust — in outputs, processes and the people using them.
- Successful programmes start with low-risk “lighthouse” use cases and build governance in from the outset.
Interest is shifting towards more ambitious deployments, Hoffmann says — agentic casework, digital twins for policy-making, and document sifting — as organisations grow comfortable moving beyond pilots. Yet many pilots that succeed in controlled conditions “fail to scale because organisations can’t build confidence in the outputs, the processes behind them, or the people using them”.
The barriers are human, not just technical
The concerns Hoffmann describes are as much cultural as computational: bias and accuracy worries, unclear accountability between humans and AI agents, and “emotional barriers” such as fear of replacement. Her prescription is unglamorous but consistent — pick a high-value, low-complexity scenario, embed responsible-AI principles from the start, and turn it into a repeatable blueprint. She also cautions that sovereign AI is widely misunderstood as a question of where data is hosted, when it is really “about control — assurance, supply chain, and the broader ecosystem of AI”.
Looking forward
The diagnosis chimes with hard evidence of what happens when trust and governance are treated as afterthoughts. Research has estimated that UK businesses waste £67bn a year on failed AI projects, while workplace AI use is doubling but the gains cluster at the top. Hoffmann’s argument is that the organisations pulling ahead are not those with the best models, but those that made their AI defensible before scaling it — a distinction UK public-sector leaders, under pressure to show results, cannot afford to skip.