Key points
- Organisations of two hundred and fifty staff or more, and public bodies, probably use AI already, sanctioned or not. The challenge is scale and control: proper buying routes, thorough supplier checks and real board oversight.
- A government survey found fifty-two per cent of large businesses using AI in twenty twenty-six. The Office for National Statistics, measuring differently, found forty-nine per cent of large businesses reporting use of at least one AI technology that June.
- Pilots often stall, each stuck in one department, never reaching everyday work. A small central team that sets standards and spreads successes usually beats free rein or an outright ban. One typical first project: map the AI tools staff already use, and move the useful ones onto company-approved accounts.
- Public bodies usually buy through approved routes such as G-Cloud, a government cloud services framework. Its description doesn't mention AI, so ask suppliers which category covers their offer. Frameworks speed buying but don't do the evaluating: you still need a clear brief, uniform questions and a way of scoring.
- Put identical security questions to every AI supplier, such as where your data is sent and who can see it, and keep written answers. Your own security team should decide, with legal and information governance advisers checking anything involving personal information.
- Boards can oversee AI without knowing how models operate, but should hear where it is used, what value it brings, any problems and who is responsible. A brief update at regular intervals is a good first step.
- To go further, read the full guide on AI Consultancy Finder.
Transcript
Organisations of two hundred and fifty staff or more, and public bodies, probably use AI already, sanctioned or not. The challenge is scale and control: proper buying routes, thorough supplier checks and real board oversight. A government survey found fifty-two per cent of large businesses using AI in twenty twenty-six.
The Office for National Statistics, measuring differently, found forty-nine per cent of large businesses reporting use of at least one AI technology that June. Pilots often stall, each stuck in one department, never reaching everyday work. A small central team that sets standards and spreads successes usually beats free rein or an outright ban.
One typical first project: map the AI tools staff already use, and move the useful ones onto company-approved accounts. Public bodies usually buy through approved routes such as G-Cloud, a government cloud services framework. Its description doesn't mention AI, so ask suppliers which category covers their offer.
Frameworks speed buying but don't do the evaluating: you still need a clear brief, uniform questions and a way of scoring. Put identical security questions to every AI supplier, such as where your data is sent and who can see it, and keep written answers.
Your own security team should decide, with legal and information governance advisers checking anything involving personal information. Boards can oversee AI without knowing how models operate, but should hear where it is used, what value it brings, any problems and who is responsible.
A brief update at regular intervals is a good first step. To go further, read the full guide on AI Consultancy Finder.