We don't know which AI uses we can approve safely, or which policies and laws apply.
AI Governance Baseline, then an AI approval route.
Public sector
Using AI safely in public services.
Residents, suppliers and staff are already using AI. The questions are about control: which AI uses to approve, how to keep personal data under control and how to give staff a safe way to use it.
Where organisations often start
Four questions, and where each leads.
We are worried about personal data being processed outside the UK.
AI Governance Baseline: where each tool sends data, set out for your Data Protection Officer to decide.
AI-written letters and requests are adding to our workload.
AI Opportunity Assessment for that service area.
Staff have been told not to use AI, or are using their own tools.
AI approval route, then staff training on safe use.
How I help
Four services, one route to safe AI use.
Charged by the day, day rate on request. Days are estimated and agreed in writing before work starts.
01
AI Governance Baseline
What AI is in use, including staff's own tools and AI features already switched on in existing systems; what each supplier's own terms say about where data is stored and processed; the gaps; and a one-page view for senior leadership.
02
AI Opportunity Assessment
For one service area, such as correspondence: map the work, sort each step by the data it involves, and identify where AI could safely help and where it should stay with people.
Ends with a decision paper: the recommended first use, its safeguards, owner and measures.
03
AI approval route
Decision rights, an approval process for new AI uses, an AI register and the policy updates, each signed off by your own specialists. Tested on a real use so it is used rather than filed.
04
Staff training on safe AI use
Practical sessions on the uses your organisation has approved: what can go into a tool, checking every draft before it leaves, and keeping a named person accountable.
One route for every rule
Each AI use checked once, by the people who own the rules.
AI touches rules your organisation already manages. The risk is that each team checks AI on its own, or nobody does. I bring them into one approval route, and approved uses then feed straight into staff training and adoption.
Data protection
Lawful basis, impact assessments, processing outside the UK, automated decisions about people.
Data Protection OfficerFreedom of information
AI outputs and the prompts used can be held information within scope of a request.
Information governanceRecords management
Where AI drafts and chats are kept, and for how long.
Records managerPublic Sector Equality Duty
Whether an AI use affects some groups differently, and a record of how that was considered.
Equalities leadInformation security
Tool settings, access and supplier security.
IT security and the SIROComplaint handling
Quality and timescales when AI helps draft replies.
Complaints leadService-specific rules
Extra protection for children and vulnerable adults, and safeguarding in social care.
Service director and safeguarding leadProcurement and transparency
Supplier and data processing terms, and publishing how AI tools are used.
Procurement and the policy leadWhat I do not do: legal or data protection advice, security testing, or building and configuring software. I do not sell or resell any AI product. Each rule area is decided by your own specialists; I make sure it is asked, assigned and tracked.
Free guide
Using AI safely in public services.
The two-page overview of this approach, including the policies that typically need an AI section.
AI for speed.People for decisions.
A 30-minute conversation about where AI sits in your organisation today.