How UK.gov's new AI tool is trained on "synthetic emails"

One of the UK government’s latest AI projects is using OpenAI LLMs to create “synthetic emails” for training an email-sorting system, overcoming a common data blocker for public sector projects.

The government’s AI and Data Readiness team shared details of “Project Kestrel” in a gov.uk blog post recently, describing how the “end-to-end prototype” used LLMs to create a dataset it could train its central Cuckoo system on.

Senior Product Manager Faye Greasley said the Government Digital Service team created “Tiger Heron,” a synthetic email generator that produced emails appearing similar to the day-to-day communications processed by the public sector without using live data.

She said: “Access to a safe, realistic data set is one of the most common blockers to data innovation in the public sector. Tiger Heron may offer a practical model for getting past it and one that could apply well beyond email.”

Tigers, Herons, and Cuckoos