Insights

How we think about AI in public services

Our perspective on building AI that organisations can trust, and an open view of where each of our products stands.

Our perspective

RSS: /insights/feed.xml

AI should show its working

In public services, an answer is only useful if it can be checked. We design systems that point to the document, recording or rule behind every output, so staff can verify before they act.

Rules for safety, AI for language

Language models are good at understanding and drafting. Decisions about safety, eligibility and approval belong to clear rules and accountable people. Our products keep those responsibilities separate.

Language access is a service issue

When people can use the language they are most comfortable in, including Welsh, services become fairer and more efficient. Speech technology can make that practical at scale.

Only claim what we can evidence

We describe only capabilities we can evidence. Organisations deserve to know exactly what they are getting.

Where products stand

Business Evidence Wallet

In development. Running in a private environment; not yet publicly launched.

View product page →