SBA VetCert SDVOSB / VOSB approved 2026-05-18 Local-first deployment patterns Model substitution and human approval controls
DOL RFI OAA-26-00311 Capability Page

Governed AI Support for Section 508 Accessibility Workflows

Article 11 AI combines local-first OCR and accessibility triage with signed governance patterns for AI-assisted remediation. The current Accessibility Radar prototype scans web assets for a practical WCAG subset, triages PDFs for accessibility risk, emits code-level remediation guidance, and produces dashboard artifacts for human review.

Operational Prototype

What Exists Today

Accessibility Radar v0.1 is a working local prototype, not a completed enterprise compliance suite. It is designed to show DOL and prime partners the concrete direction: AI-assisted accessibility work that is auditable, local-first, and governed by human approval.

Automated WCAG Subset Scan

HTML scanner checks common issues such as document language, page title, image alt attributes, labels, headings, landmarks, duplicate IDs, link names, button names, iframes, tables, and mouse-only interaction risk.

working prototypecode-level suggestions

PDF Accessibility Triage

PDF triage checks title metadata, language, tagged-PDF signals, structure tree signals, text-layer extraction, likely scanned pages, file size, page counts, and SHA-256 report provenance.

working prototypetag writeback roadmap

Local OCR Path

S2_LEX provides a local-first PDF extraction pipeline using native PDF text first and Tesseract OCR fallback for scanned pages. Document content does not need to leave the controlled environment.

Tesseract availableair-gap compatible

Dashboard Artifacts

Reports are emitted as JSON, CSV, Markdown, and HTML dashboard output, suitable for review queues, leadership summaries, and audit handoff packages.

working prototype

CI/CD Templates

GitHub Actions and Jenkins starter templates are included so accessibility triage can run in developer workflows and fail builds above configurable severity thresholds.

templates ready

AI Governance Layer

Article 11's broader SPIRALMESH and proof APIs support model substitution, human-gated AI actions, signed receipts, and auditable authority boundaries for AI-assisted remediation.

Article 11 core
Workflow Architecture

From Flat Content To Auditable Remediation Work

1. IngestHTML, PDF, and document evidence are scanned locally; PDFs can be paired with the Tesseract-backed S2_LEX OCR path.
2. DetectAutomated rules flag common WCAG and PDF accessibility risks with line-level or asset-level evidence.
3. SuggestThe system emits code-level remediation recommendations and PDF remediation guidance for human review.
4. GovernAI-assisted remediation can be routed through human approval, model substitution controls, and signed audit receipts.
5. ReportDashboards and machine-readable reports support QA queues, trend reporting, and audit handoffs.
DOL Requirement Fit

Capability Matrix

This matrix is intentionally candid. It separates working prototype capability, Article 11 core infrastructure, near-term roadmap, and partner lanes.

DOL Requirement Area Current Status Article 11 Position
Automated accessibility scanning and testing Prototype live Accessibility Radar performs automated checks for a practical WCAG subset across HTML pages and emits severity, WCAG mapping, evidence, and remediation guidance.
PDF testing, OCR, and remediation support Prototype live PDF triage and Tesseract-backed OCR paths exist. Automated tagged-PDF writeback and full PDF/UA remediation are roadmap or partner functions.
Code-level remediation suggestions Prototype live Findings include code-level recommendations for alt text, labels, headings, landmarks, names, iframes, duplicate IDs, tables, and keyboard interaction risk.
Dashboards and exportable reports Prototype live Reports emit JSON, CSV, Markdown, and HTML dashboards. Trend dashboards and multi-project analytics are roadmap.
Jenkins and GitHub Actions integration Templates ready Starter CI templates exist and can fail builds above configurable severity thresholds.
AI model flexibility and government-approved model substitution Article 11 core SPIRALMESH supports multi-runtime routing, model substitution posture, and evidence-bearing model output. This is Article 11's strongest fit.
AI-enabled actions with approvals, rollback, and audit logging Article 11 core Article 11's governance layer is built around human-gated effects, signed receipts, authority boundaries, and auditable chain-of-custody patterns.
Guided manual testing workflows Roadmap Current reports support human review; guided persona-specific manual testing flows are planned for the 30-90 day build path.
50-seat licensed platform, RBAC, collaboration, issue queues Roadmap Article 11 can build a small multi-user platform, but it is not currently a production 50-seat SaaS with full RBAC and support workflows.
Drupal, Appian, Salesforce, SharePoint, Microsoft 365 connectors Partner/roadmap Connectors are feasible but not present today. Article 11 should team with a prime/OEM or build these in phased delivery.
FedRAMP, ATO, SSP package Roadmap FedRAMP is not held today. Article 11 can support security documentation and local/on-prem deployment patterns while pursuing a formal authorization roadmap.
Evidence Artifacts

Sample Reports

These artifacts were generated by Accessibility Radar v0.1 before this page was published. They are examples of report structure and workflow output, not Section 508 compliance certifications.

Article 11 Federal Page Scan HTML WCAG subset scan of the federal contracting page. Result: one minor landmark finding.
VetCert PDF Triage PDF accessibility triage against two SBA VetCert evidence PDFs. Result: title, language, and tagging signal findings.
Public DOL pages blocked this unauthenticated prototype scanner with HTTP 403 during a quick command-line test. Article 11 will not represent blocked scans as DOL findings. Demonstrations should be run against authorized pages, provided sample assets, or agreed test environments.
Security And Privacy

Built For Sensitive Federal Workflows

Local-First Content Handling

Accessibility Radar and S2_LEX can operate without sending source documents to third-party model APIs. This supports sensitive document triage, air-gapped pilots, and on-premise evaluation paths.

Human Authority Preserved

AI output is treated as recommendation and evidence, not final compliance determination. Human reviewers approve remediation decisions and remain accountable for publication actions.

Model Substitution

Article 11's multi-runtime architecture is designed to route work to government-approved models rather than locking agencies into one vendor default.

Traceable Evidence

Reports include issue evidence, remediation suggestions, timestamps, and SHA-256 provenance. Signed chain publication is available through Article 11 governance patterns when explicitly authorized.

Federal Contact

RFI, Demo, Or Teaming Inquiry

Article 11 AI, Inc. is an SBA VetCert SDVOSB/VOSB small business. For DOL RFI OAA-26-00311 or prime teaming discussions, contact Steve Sonza.

Point Of Contact

Steve Sonza, Founder
steve@article11.ai
Phone available in RFI submission and on request.