Sector
AI consultancy for the public sector
Transparency duties, public scrutiny and procurement rules set a high evidence bar. We build to the standards your assurance team already uses.
The constraint
Why public sector AI projects stall
Public bodies face scrutiny no business does. Decisions can be challenged, information requested, and algorithmic tools that affect the public may need publishing under the Algorithmic Transparency Recording Standard.
A reasonable bar, and a system not built for it will fail it. ATRS asks about data, model, oversight and impact, which has to be recorded during the build.
Then procurement. A compliant process, framework requirements, and supplier evidence that is increasingly AI-specific. That is why AI governance evidence matters commercially well before it matters legally.
Where it works
Where AI earns its place
Ordered roughly by how quickly they get approved. The first is usually the right first project: it is the least contentious.
Correspondence and casework triage
Classify, route and summarise correspondence and casework, flagging urgency and vulnerability for human decision.
FOI and information request support
Find material across systems and draft first-pass responses for officer review. High volume, well defined, measurable.
Policy and guidance retrieval
Frontline staff query policy and procedure with citations. Residents get the same answer whoever they reach.
Document and form processing
Turn submitted forms and evidence into structured data. Less rekeying, fewer errors.
The gate
What assurance will ask for
We produce it during delivery. Cheaper, and more likely to be accurate.
- ATRS record drafted during the build
- AI management self-assessment, completed accurately
- DPIA and equality impact assessment covering differential outcomes
- Documented human decision points, resourced realistically
- A route for the public to challenge an AI-assisted decision
- Supplier evidence in the form your procurement process requires
Illustrative scenario. A composite of how an engagement typically runs, not a specific client.
Illustrative: a local authority service
Officers spend too long routing and summarising correspondence. It arrives through several channels in inconsistent formats, and the backlog generates complaints of its own.
The correspondence contains personal and occasionally special category data, so processing stays in the authority’s own tenancy, with retention aligned to the existing records schedule.
The system classifies and summarises. It does not decide. That boundary is documented: "prioritised for an officer" is a workload tool; "decided by a system" is a decision open to challenge.
The ATRS record and equality impact assessment are drafted during the build, while the people who made the design decisions can still explain them.
Questions from this sector
Do we have to publish an ATRS record?
The Algorithmic Transparency Recording Standard applies to algorithmic tools used by public sector organisations that affect decisions about members of the public, with scope guidance from the Central Digital and Data Office. Draft the record during the build whatever you decide about publication. It surfaces oversight and data gaps while they are cheap to fix.
Is DSIT’s AI Management Essentials a procurement requirement?
No. DSIT consulted on it but said in February 2026 that it would not publish it, so it is not a procurement requirement. Panels still ask suppliers about AI governance. Public bodies: expect to ask for an AI inventory, a risk register and DPIAs. Suppliers: expect to be asked for them.
Can you work through a framework?
Yes. Tell us your route to market and we will work with your procurement colleagues on the compliant path, including framework options and direct award thresholds where they apply.
Next step
Start with a 30-minute call
No deck. Tell us the problem. We will say whether AI is the right tool, what it would take and roughly what it would cost.