Free tool
A working diagnostic for the first five Service Standard points.
Reread for AI‑enabled services. One question per point, written in your own words. The model reads and drafts the report. About fifteen minutes.
01What it covers
The first five Service Standard points, reread.
One question per point, pulling on the assurance posture of an AI‑enabled service. The remaining nine points are written for you to apply yourself.
- 01
Understand users and their needs
How the AI‑enabled parts of your service understand users, and how the people in the loop are accounted for.
- 02
Solve a whole problem for users
Whether the AI‑enabled element solves the user's problem in full, or hands off to a part of the service that's still manual or absent.
- 03
Provide a joined‑up experience across all channels
How the AI‑enabled channels read alongside the non‑AI ones, and where users notice the seam.
- 04
Make the service simple to use
Whether the AI‑enabled interaction is intelligible to a first‑time user, and what the recovery path looks like when the model is wrong.
- 05
Make sure everyone can use the service
How your service performs against assistive technology, low‑literacy users, and the populations most likely to be misclassified by the model.
02What you get out
Verdicts in the margin. Prose in the spine.
Each point gets a one‑line verdict in italic in the margin, plus a paragraph of reasoning in the spine. Built to be readable, reviewable, and easy to argue with.
Sample point, from the report
Point 04. Make the service simple to use.
The AI‑enabled triage step gives a confident answer most of the time, and the team has tested it well in the dominant happy path. The recovery path when the model is wrong is less clear: users see a generic error and are returned to the start of the journey rather than to the step where the model intervened.
The team's research repository shows two rounds of usability testing on the happy path and none on the recovery path. A run of recovery‑path testing, with users from populations most likely to be misclassified, would close the gap.
The example above is one point from a five‑point report. The full report covers all five, with a one‑page summary at the front and a methodology note at the back. It's yours to keep, share with your team, or take into a Service Standard assessment as background.
03The practitioner
Built by Chris Leo.
Eighteen years in UCD, eight in UK central government. Cross‑government and HMRC Service Standard assessor. The diagnostic builds on Cambridge's HCI for AI Systems methodology and on the Service Standard as published by GDS.
Free to use, free to share. The tool is intentionally light: no account, no analytics on your answers, no follow‑up other than the report itself and an opt‑in to the newsletter.
Drawn from
Cambridge HCI for AI Systems·MRS Diploma·BSc Psychology
Start the diagnostic
Free, anonymous, about fifteen minutes.
Five Service Standard points, one question each, written in your own words. Your answers stay private until you ask for the report. Pause and pick up from where you left off on the same browser. No account needed.