How it works
Scope. Fix. Verify. Review. Report.
The system that writes a fix never grades it. Verification runs separately within Suretto, with explicit test requirements and human decisions where judgement is needed.
- 01
Agree the journeys and test scope
We map your site or app from the sitemap, the router and real navigation, then exercise interactive states: menus, dialogs, filters, carts, multi-step forms, validation errors. We agree which journeys, states, components and environments to assess, and record exclusions before starting.
Output: an inventory of templates, components and states, with the WCAG criteria that apply to each.
- 02
Fix in your code
Defects are repaired at the source: semantic structure, names and roles for custom controls, label associations, focus management, contrast tokens, motion preferences. AI drafts the change within rules we wrote; an engineer reviews it; it reaches you as a pull request.
Output: small pull requests, one defect class at a time, each citing the criterion it addresses.
- 03
Verify independently
A separate verification engine tests each fix three ways: automated rules on the rendered page, recorded runs with NVDA, JAWS and VoiceOver, and keyboard-only runs. Contrast is calculated from browser-derived foreground and background colours, with image analysis supporting complex visual cases.
Output: scoped results with recordings and data attached, including failed, blocked, untested and review-needed items.
- 04
Review exceptions
Anything that fails, or needs human judgement (is this alt text meaningful? is this error message clear?), goes to a reviewer with the evidence in front of them. Your own team can be that reviewer, or ours can.
Output: a decision and reasoning on every exception, kept in the report.
- 05
Issue evidence and agree follow-up checks
The evidence report is issued, and a conformance report (VPAT/ACR) when you need one. Ongoing monitoring repeats the automated checks agreed in your plan. Human reassessment, new features and ACR updates are separately scoped; the report shows what was tested and when.
Output: a dated report, a conformance document, and regression alerts.
Why it gets better
Every reviewed exception teaches the engine.
Each decision a reviewer makes is stored with the evidence it was made on: the criterion, the defect pattern and the ruling, without your source code. Those rulings refine the rules our fixer works within and the checks our verifier runs, so the share of work that needs a person falls with every project. That is how the price of a verified page comes down over time while the standard of proof stays the same.
We do not train any model on your code. How AI handles your code.
Timeline
Agree the timeline before kickoff.
| Week | What happens | What we need from you |
|---|---|---|
| 1 | Inventory, automated and assistive-technology runs on the agreed templates | Staging or production URL, test accounts, read access to the repository |
| 2 | Fix pull requests, verification, exception review, report issued | Batched decisions and code review; effort estimated during scoping |
| 3+ | Monitoring at the agreed frequency; separately scoped reassessment | A deploy webhook or a scheduled crawl |
See it on your own product.
Start with one agreed user journey at no cost through early 2027. We agree the defects, deliverables and your review effort before the pilot begins.