RE: Breakout Proposed

Hi Janina,

My github handle is - abhina-kumar

Thanks,
Abhinav Kumar

-----Original Message-----
From: Janina Sajka <janina@a11y.nyc>
Sent: 22 October 2025 05:50
To: Kumar, Abhinav <abhina.kumar@sap.com>
Cc: 'Adapt tf' <public-adapt@w3.org>
Subject: Breakout Proposed

[You don't often get email from janina@a11y.nyc. Learn why this is important at https://aka.ms/LearnAboutSenderIdentification ]

Thanks Abhinav. Thanks also to Lionel for our drafting session earlier.
I've done my best to capture the essence!

https://github.com/w3c/tpac2025-breakouts/issues/62

PS: I will need your github handles to add you as Co-Facilitators.

Kumar, Abhinav writes:
>
> Hi Janina, Hi Lionel,
>
> As discussed during today's call, please find the first draft for TPAC 2025 - Breakout Session.
>
> ----
> Title: Semantic Web Navigation for Agents Using Well-known Destinations
>
> Session Description:
> The rise of agentic AI systems presents unprecedented opportunities for web accessibility. While these autonomous AI agents can serve as powerful assistant helping users navigate complex multi-site workflows, they face a fundamental barrier that directly impacts accessibility. The reality is that 70-80% of websites lack APIs for basic functions like finding contact information, accessing help resources, or managing user accounts, forcing agents to rely on brittle web scraping techniques that break whenever sites update their designs, causing unreliable experiences.
>
> This breakout session explores how WAI-Adapt's Well-known Destinations proposal can bridge this gap through Semantic Web Navigation Tools. A standardized set of tools that enable reliable discovery and navigation of well-known destinations in a webpage, using semantic identifiers rather than fragile HTML/CSS selectors. Instead of agents hunting for contact information using breakable selectors like `.contact-info`, they can reliably navigate to semantic destinations marked with rel="contact".
>
> These navigation tools provide three core capabilities:
>
>   *   Destination Discovery: Finding all available semantic destination on webpage
>   *   Reliable Navigation: Moving between pages using these well-known destinations
>   *   Content Retrieval: Extracting information from these pages for LLM processing.
>
> The proposal extends accessibility benefits in two directions: direct accessibility (helping users with disabilities navigate websites) and indirect accessibility (enabling AI agents to work reliably across any compliant website).
>
> Key discussion topics include:
>
>   *   How agentic AI systems can leverage semantic navigation for reliable web interaction
>   *   Integration patterns with AI frameworks like Model Context Protocol (MCP) or WebMCP
>   *   Balancing automated navigation with human-in-the-loop approaches for sensitive operations
>
> This session addresses the critical intersection of AI advancement and accessibility, ensuring that as agentic systems become more prevalent they enhance the accessible web experience.
>
> Session Goal
>
>   *   Validate the technical approach and gather community feedback on extending WAI-Adapt Well-known Destinations to support reliable Web Navigation for agents, with focus on accessibility benefits and standardization pathways.
>
>   *   Explore integration patterns with existing AI frameworks and protocols
>   *   Build consensus on balancing automated capabilities with human control for sensitive operations
>
> Agenda for the Meeting
> 1. Introduction and Problem Framing
>
>   *   Challenge of Web Navigation for Agents: 70-80% of sites lack APIs
>   *   How brittle web scraping fails agents
>   *   Well-known Destinations as semantic bridge between reliability and universality
>
> 2. Real-world Applications and Use Cases
>
>   *   Use Case 1 - Accessibility compliance monitoring across website portfolios
>
> 3. Technical Deep Dive: Semantic Web Navigation Tools Architecture
>
>   *   Core capabilities: destination discovery, semantic navigation, content retrieval
>
>   *   Integration patterns with Agents (Eg. MCP or WebMCP)
>   *   Balancing AI capabilities with human-in-the-loop approaches
>
> 4. Open Discussion and Community Feedback
> ----
>
> Thanks,
> Abhinav Kumar
>
>

--

Janina Sajka (she/her/hers)
Accessibility Consultant https://linkedin.com/in/jsajka

The World Wide Web Consortium (W3C), Web Accessibility Initiative (WAI)
Co-Chair, Accessible Platform Architectures     http://www.w3.org/wai/apa

Linux Foundation Fellow
https://www.linuxfoundation.org/board-of-directors-2/

Received on Wednesday, 22 October 2025 10:05:09 UTC