AI for high-volume case and correspondence work
FOI requests, case triage, policy-grounded Q&A. The work that piles up faster than headcount, where careful automation buys real time back.
Public sector teams drown in correspondence: FOIs, casework, policy enquiries, freedom of information requests. AI can't make policy decisions — but it can summarise, route, draft, and search your own policy library faster than anyone could by hand.
The jobs we'd look at first.
Case triage and routing
Inbound cases classified against your routing rules, with the reasoning attached and ambiguous cases escalated rather than guessed.
FOI handling and drafts
First-draft FOI responses built from your previous releases and disclosed information, ready for review before they go out.
Policy-grounded Q&A
Citizen-facing and internal Q&A grounded strictly in your published policy library, with citations and no off-policy answers.
Casework summarisation
Long case files summarised into a structured brief so caseworkers walk into a meeting reading one page, not the whole bundle.
The projects we'd turn down.
Worth being explicit. These are the asks we'd say no to in this sector — and saying no is part of the work.
We would not build anything that decides on a person's entitlement, benefit, or status. Those decisions are reserved for humans, with the audit trail to match.
We would not deploy a citizen-facing chatbot that gives free-text policy answers. Grounded, cited responses or nothing.
We would not skip the data-protection impact assessment. If we can't do the assessment, we can't do the build.
A first call, no slides.
Tell us the workflow. We'll tell you, plainly, whether it's worth a Diagnostic and what we'd expect to find.