Public sector AI fails when governance is treated as paperwork
Every public sector AI programme eventually collides with the same constraint: legitimacy.
When an AI leader from a major supplier steps into a governance role for national health bodies, the story is not “who got the seat”. The story is whether governance is treated as an enabling discipline or as a box-ticking exercise that happens after procurement.
Healthcare is where data systems carry moral weight. People can forgive slow service; they do not forgive systems that feel unaccountable. That is why “governance” cannot just mean policies and committees. It has to mean observable controls.
For example: clear lines of accountability for model decisions, auditable data access, and a simple answer to the public question, “What happens when this system is wrong?” If the only answer is a slide deck, trust erodes.
There is also a practical risk that gets missed in the headlines. AI capability is not the limiting factor. The limiting factor is organisational readiness: data quality, workflow design, staff training, and the ability to monitor systems in production.
The contrarian point is that stronger governance can increase speed. When controls are real and transparent, teams can ship with confidence instead of pausing every time a scandal-risk headline appears.
Views wanted:
- What would “auditable AI” look like in a healthcare setting, in plain English?
- How should conflicts of interest be managed when vendors are close to governance?
- Which decisions should never be automated, regardless of accuracy claims?
- What minimum transparency should the public expect about data use and model behaviour?
- How do you balance faster delivery with the need to protect trust?
