Skip to content
Sector programmes · Public sector

AI for the public sector,
worthy of public trust.

Independent, standards-led support for departments, agencies, councils, and arm's-length bodies putting AI to work: grounded in government guidance, the Algorithmic Transparency Recording Standard, and public law duties, free of vendor incentives.

— State of play —

The ambition is set.
Delivery is the work.

Government has placed AI at the centre of its plans for public service productivity, from AI tooling for civil servants to automation in casework and correspondence. Most public bodies are piloting or planning something. Far fewer have the data, skills, and assurance in place to take a pilot into a service that citizens depend on.

The rules of the road are clearer than they were. The Algorithmic Transparency Recording Standard is mandatory for central departments, equality and data protection duties apply in full, and decisions made with AI still have to satisfy the standards of public law. Turning national ambition into lawful, transparent, working services is where most organisations need help.

37%
of government bodies had deployed AI by 2024
70%
were piloting or planning AI use cases
£45bn
a year in unrealised savings from full digitisation of public services
— Use cases —

Where AI is already
earning its keep.

The strongest public sector use cases keep a caseworker accountable and give time back to frontline service, not the back office.

Casework and triage

Applications, claims, and case queues prioritised by urgency and complexity, so people with the greatest need are seen first and officials make the decisions.

Citizen services

Assistants that resolve routine queries around the clock, route the rest to the right team, and stay accessible to people who cannot or will not use them.

Correspondence and FOI

Drafting, summarisation, and redaction support that clears ministerial correspondence and information requests faster, with a full audit trail.

Policy and consultation

Consultation responses analysed at scale and evidence synthesised across sources, so policy teams spend their time on judgement rather than collation.

Fraud and error

Anomaly detection across payments and procurement that protects public money, designed and tested against bias before it touches a single case.

Transparency and assurance

ATRS records, DPIAs, and equality impact assessments done properly and published proactively, before a journalist or a judicial review asks for them.

— Sector considerations —

What we watch
most closely.

01
Transparency

Publish before you are asked

The Algorithmic Transparency Recording Standard is mandatory for central government, and the direction of travel is clear for everyone else. A tool you would not be comfortable describing in public is a tool you should not be running.

1
ATRS record per algorithmic tool
02
Lawfulness

Public law still applies

Decisions assisted by AI must still be rational, fair, and within statutory powers, and a court will expect a human to have genuinely exercised judgement. Automation that hollows out discretion invites judicial review.

0
decisions delegated wholly to a model
03
Equality

No one gets designed out

The public sector equality duty applies to algorithmic services exactly as it does to human ones, and digital exclusion is real. Every AI-assisted service keeps a route for people who need a person.

100%
of services keep a non-digital route
04
Procurement

Bought to be left

Public bodies keep systems for a long time and answer for every pound. Total cost, data portability, and exit routes belong in the evaluation from day one, not in a renegotiation five years in.

5yrs
typical life of a public technology contract
05
Workforce

Capability before contracts

Outsourcing AI without building internal skill creates dependency that outlasts the contract. Training comes first, so your own people can specify, challenge, and run what is built for them.

2x
adoption after structured staff training

Set your organisation's
AI position.

A short call with the public sector team. Plain answers, a clear next step, and no software to sell.