Hackathon evidence
Type C: Accessibility Audit + AI Redesign
Everything a judge needs, on one page: what was wrong, what the AI found and recommended, and what replaced it.
Before
The existing inaccessible experience
- Form HS-1042-A — 17 pages, 214 fields
- Grade 13.2 reading level, English only
- 37 unlabelled fields, broken reading order
- 15-minute session timeout, no save
- Accommodation guidance on page 14
AI audit + redesign
Barriers identified, recommendations generated
- Accessibility score: 42/100
- 18 barriers across six categories
- Each barrier scored, explained, and matched to affected people
- Plain-language rewrite generated question by question
- Workflow regenerated as an adaptive conversation
After
The redesigned Agentforce experience
- 8 questions, one at a time, visible progress
- Grade 6–8 plain language
- Read aloud, keyboard-only, larger text, high contrast
- Accommodations and language offered first
- Save and resume, and a person on request
Concrete transformations
- 17-page PDFGuided conversation214 fields reduced to 8 questions that matter
- Grade 13 languageGrade 6–8 plain languageReading level rewritten and re-scored
- All questions at onceOne question at a timeProgressive disclosure with 'Step 3 of 8'
- Hidden accommodationsProactively offered assistanceAccommodation offered in the first message
- English onlyMultilingualLanguage choice before the first question
- Visual-first navigationMultimodal accessible interactionRead aloud, keyboard, voice, larger text
- Generic workflowAdaptive experienceFlow changes with the support each person picks
Audit trail: the critical findings
- Criticalseverity
Reading level: Grade 13.2
Rewrite every question in short, direct sentences and explain each term in context.
- Criticalseverity
37 form fields have no programmatic label
Ask each question as labelled text in a conversation, with one clear input at a time.
- Criticalseverity
Accommodation request is buried on page 14
Offer accommodation options in the first message, before any question is asked.