The task comes before the sector.

A sector label is not a research brief. Summarising a policy document, extracting a field and answering a question against a controlled source collection each need different evidence and assessment criteria.

We focus on limited questions where the contribution of an AI approach can be examined. The domain determines what counts as a correct answer, which omissions matter and what human review remains essential.

The examples below are potential research applications. They are not claims of completed deployments, sector accreditation or demonstrated outcomes.

Data access and representation

The material available for testing may differ from the intended operating environment. An experiment should explain those differences and how they limit the result. Permission to use the data must be established before it enters the research workflow.

Domain review

Specialist language and exceptions make evaluation more demanding. A domain reviewer helps identify errors that a generic similarity score might miss.

Accountable decisions

A technical assessment contributes evidence. It does not replace the organisation's legal, security, governance or professional review. We keep those boundaries visible when describing the next step.

Internal laboratory research is described separately in Sentinel. For the engagement options that apply across contexts, see research services.