72% → 41%
AI use by designers peaks in Discover and falls in Deliver. Nobody trusts it to finish.
Source: Jakob Nielsen, survey of 500+ designers, October 2026.
AI builds the product.People own the evidence.
Agents and generative design get you to a demo fast. Passing assessment still needs a practitioner's name on the work. AgentOps gives UK public services independent UX assurance: an expert review against the Service Standard, the GOV.UK Design System and WCAG 2.2 AA, with the evidence pack ready before your panel meets.
Ideas are cheap now. Finishing is not.
Generative tools produce something plausible every time. The errors hide in details only a practitioner checks, and they surface late, at assessment, when they cost the most time to fix.
14 points
Every service is assessed against the Service Standard. AI-generated work gets no exemption.
Source: GOV.UK Service Manual
Day one
Accessibility and AI transparency belong at the start, not bolted on at audit.
Source: WCAG 2.2 AA · Algorithmic Transparency Recording Standard
How it works
AI speeds up the checking. A person makes every call, and signs off at every gate.
01
Scoping call
We agree the service, the phase and your assessment date, and plan the review around your sprints.
02
Review
We work through your prototypes and agent workflows in your tools. Your team keeps building.
03
Evidence pack
Every finding has a severity, a source and a fix. Themes trace back to user quotes.
04
Assessment
You walk in with evidence mapped to the points your assessor scores, and a record your governance board can read.
What we check
Each check maps to the Service Standard, so the pack reads like your assessor's own form, and every team and programme gets the same format.
Design system compliance
GOV.UK Design System components and patterns, used the way they were designed.
Points 4 · 13
Accessibility
WCAG 2.2 AA, treated as a blocker from day one, tested with assistive technology.
Point 5
AI transparency
Users can see what the agent did, how sure it is and where answers came from, and confirm before anything changes.
Points 4 · 9 · 11
Evidence tracing
Every theme links to a user quote. Every decision links to a source.
Point 1
Journey coverage
Edge cases, error states and hand-offs across the whole journey and every channel.
Points 2 · 3
Plain language
British English, short sentences and GOV.UK style, so everyone can understand it.
Points 4 · 5
People, not a platform
A tool can flag a contrast ratio. It cannot tell your assessor why a journey works. That takes someone who has sat through assessments.
- A named, accountable lead on every engagement. Not a support ticket.
- A human at every gate. Nothing enters the pack until a person has reviewed it.
- No new software to approve. We work in your tools, under your data rules.
- SC cleared. Work on sensitive services can start without months of vetting delay.
Working with us
Timed around your assessment, not added on top of your sprint. Best started four to eight weeks before an alpha, beta or live assessment.
Readiness review
One service, one assessment. Fixed scope and fixed fee, quoted after the scoping call.
Retained assurance
Monthly cover for each release and assessment, across one service or a whole programme.
What it takes from your team
- Access to prototypes, agent workflows and research notes, in your own tools
- One named contact, for about an hour a week
- Your assessment date, so we can plan backwards from it
No extra headcount, no new software to approve, and no handover of your work.
Built by someone who has sat on the other side of the assessment
- HM Revenue and Customs
- Home Office
- UK Export Finance
- Department for Energy Security and Net Zero
- Ofgem
- NHS
- The Scottish Government
Past projects, not endorsements.
Ideation is free. Delivery is not.
Tell us which service you're assessing next. We'll map the review to it.