Accelerating Accessibility Remediation with Agentforce Conversation Client and AI

This article details how the Agentforce Conversation Client (ACC) team engineered a scaled, AI-driven workflow to tackle complex accessibility remediation challenges across Salesforce.

ACC Team Mission: Building an Accessible Conversational UI

The Agentforce Conversation Client (ACC) team's mission is to develop a scalable conversational AI UI platform. Accessibility is a core architectural requirement, not an afterthought. WCAG standards, semantic HTML, ARIA, and accessible interaction patterns are integrated into the ACC component architecture from inception. Given ACC's role as a foundational platform for numerous Salesforce products, ensuring accessibility at the source is critical.

Operational Constraints in Large-Scale Accessibility Remediation

ACC's distributed nature across multiple Salesforce clouds created operational challenges. Independent accessibility audits by various teams, often on differing release schedules, led to a surge of hundreds of remediation findings simultaneously accumulating in the ACC backlog. The team faced the dual challenge of investigating, prioritizing, remediating, and validating these issues while simultaneously delivering roadmap features under strict M1 delivery timelines. This meant accessibility work directly competed with active platform development.

While initial efforts involved distributing ownership, grouping issues, and coordinating with experts, recurring patterns highlighted the unsustainability of purely manual remediation. Traditional processes were not designed to absorb continuous, high-volume accessibility audit spikes while maintaining platform velocity.

Limitations of Traditional Accessibility Remediation

Traditional workflows are characterized by manual interpretation and fragmented tooling. Developers typically:

1. Run accessibility scans (e.g., axe-core).
2. Interpret scan outputs.
3. Map violations to WCAG guidelines.
4. Determine remediation strategies.
5. Apply fixes.
6. Manually validate results.

At ACC's scale, this process became unsustainable. Accessibility violations are complex, requiring deep understanding of WCAG requirements, affected DOM structures, impact severity, and framework-specific remediation patterns. The manual, repetitive nature of this process was time-intensive and prone to inconsistency due to varying developer expertise. This could lead to fixes addressing symptoms rather than root causes, or introducing new regressions.

To mitigate this overhead, the ACC team standardized accessibility reasoning within their MCP workflow, incorporating structured guidance, WCAG-aware rules, prioritized scoring, and consistent validation. This built upon the foundation of the ADK a11y agent, transforming remediation from manual debugging into a structured, scalable engineering process.

Building an MCP-Based Accessibility Remediation Platform

While existing tools like ADK a11y agent and AI scanners provide robust detection, the core opportunity lay in automating the remediation itself. The ACC team developed a Conversation UI-centric MCP-based remediation workflow, focusing on the ACC domain and complemented by a shared execution layer for analysis, automation, and AI-assisted remediation.

This platform injects deterministic accessibility context directly into the remediation pipeline using WCAG rules, axe-core analysis, structured scoring, and prioritized recommendations. The resulting pipeline:

• Accessibility Issue Detected
• ADK/ACC MCP Scan Executed
• AI Prioritization Applied
• AI-Generated Remediation Produced
• Validation Performed
• Pull Request Generated

This AI-driven pipeline reduced approximately 80% of the ACC accessibility remediation backlog, significantly improving scalability across the platform.

Engineering Framework-Aware Automated Code Remediation

The primary engineering challenge was achieving precision. WCAG guidelines are nuanced and highly implementation-dependent. Generic accessibility fixes often fail in large UI systems. To address this, WCAG guidance was embedded directly into the MCP workflow, allowing the remediation engine to reason against authoritative constraints alongside axe-core analysis using deterministic rules and framework-aware patterns.

Lightning Web Components (LWC) and ACC's specific architectural patterns added complexity. Many existing tools assume generic HTML. The ACC team calibrated MCP outputs specifically for LWC rendering flows, state management, and ACC component behavior to ensure generated fixes aligned with the platform's architecture.

Validation was crucial. Structured before-and-after comparison workflows were implemented to measure actual compliance score improvements. Engineers continued to review generated remediations prior to merge to ensure framework behavior, accessibility compliance, and platform stability.

Scaling AI-Driven Accessibility Remediation Across Salesforce

As the MCP-based workflow expands beyond ACC, several challenges emerge:

• Framework Diversity: ACC is primarily LWC-based. Other Salesforce teams utilize different architectures and patterns. Remediation rules must be adaptable to these diverse environments.
• Extensibility: The approach is designed to complement Salesforce's ADK a11y infrastructure. Teams can extend the MCP workflow with framework-specific automated remediation built into the existing shared scanning foundation.
• Trust: AI-generated remediation for nuanced areas like accessibility requires developer trust. This is built through deterministic outputs, measurable validation, transparent scoring, and clear escalation paths for edge cases.

The long-term vision is to establish scalable AI-driven accessibility remediation infrastructure in partnership with the ADK/Accessibility Engineering organization, operationalizing this across conversational AI surfaces throughout Salesforce.

Key Takeaways

• Agentforce Conversation Client (ACC) team implemented an AI-driven MCP workflow for accessibility remediation.
• This approach achieved a 5x acceleration in remediation speed and reduced the backlog by 80%.
• Key engineering challenges included achieving precision with WCAG guidelines and framework-specific LWC patterns.
• Scalability across diverse Salesforce frameworks and building developer trust are critical for broader adoption.
• The goal is to create a standardized, AI-powered accessibility remediation infrastructure for Salesforce.