One of the most striking commitments in the Policing White Paper is the £115 million allocated over three years to accelerate the “rapid and responsible” adoption of AI and automation across all 43 forces, led by a new National Centre for AI in Policing – Police.AI. Its remit includes giving operationally independent Chief Constables “the evidence and resources they need to ensure the AI they are using has proven benefits, is backed by robust scientific evidence and is maximising accuracy while minimising bias.” (HM Government, 2026: para. 290).
That is a demanding standard, and it is worth being clear about what it requires. Evidence of proven benefit and reduced bias cannot be generated by the AI tools themselves, but depends entirely on the quality, structure and completeness of the operational data those tools are trained on and evaluated against. The White Paper acknowledges that current data quality across forces is inconsistent and fragmented. A national AI centre cannot retroactively repair that gap for every force; it can only work with the data discipline already in place locally.
This is the overlooked dependency behind every AI adoption programme in policing: responsible AI is downstream of responsible Organisational Learning.
A force that captures observations, incidents and Lessons in a structured, consistent, audit‑ready format is already building the evidence base Police.AI will need to demonstrate accuracy and minimise bias. A force still relying on fragmented local records is not simply behind on AI – it actually lacks the foundation Police.AI’s own standard assumes exists.
While ISARR’s OLLM platform was not designed as an AI readiness tool, its effect is precisely that. It enables structured, guided capture of operational Learning; embeds Lessons rather than leaving them as static documentation; and makes data filterable and analysable in the way any credible evidence base for AI evaluation requires.
Getting this foundation right now – ahead of national AI rollout – is the difference between adopting Police.AI’s tools with genuine evidence behind them, and adopting them on faith.
Other articles in this series may be accessed below as they are published:
Ending Fragmented Priorities
43 to 1: Why Data Consolidation in Lessons Management Can’t Wait for a National Standard
No content found
No content found
No content found
References
HM Government (2026) From Local to National: A New Model for Policing, CP 1489. London: Home Office, paras. 283, 289–290.
