Accessibility is a moral imperative, but it’s also a clear business accelerator. Globally, 1.3 billion people live with disabilities, and the World Economic Forum estimates the global spending power of people with disabilities and their families at $13 trillion. Furthermore, companies that prioritize accessibility are four times more likely to outperform their competitors in total shareholder returns. For businesses like Salesforce, the most effective strategy for building better products for humanity is efficiency — embracing what a software developer would call a shift-left mentality to move tasks to earlier stages in the development process rather than tacking them on later. In other words, build things right the first time, turning accessibility into a trust differentiator that accelerates growth.
However, the shift-left approach remains a persistent challenge. Shifting left isn’t just a technical adjustment; it’s a behavioral one. It demands significant resourcing, executive support, and a rewrite of established workflows. Most teams don’t overlook accessibility out of lack of care; they simply lack the framework to truly prioritize it. Now, we face a new AI frontier: the era of hyper-accelerated app development. As code generation hits warp speed, the window for manual intervention is closing. It’s more critical than ever to redefine how we intercept the development train before it leaves the station.
To address this challenge, Salesforce today is introducing a suite of AI-powered accessibility skills designed to bring guidance directly into the tools where development work happens. At the center of this suite is the Accessibility (A11Y) Agent — an AI-powered companion that helps builders translate product requirements into accessible code, review code against Web Content Accessibility Guidelines (WCAG) global standards, and remediate issues directly within their workflow.
Salesforce has consistently focused on building products that are accessible out of the box, which to a software developer or business leader means the platform is ready to use immediately without coding. But to a person with a disability, it means the platform is designed to be usable by everyone, regardless of physical or cognitive abilities.This includes screen-reader compatibility, keyboard-only navigation, and color contrast and text designed to be readable for people with low vision or color blindness. Our approach adheres to WCAG, is informed by inclusive user research, and is designed with accessibility baked into our Salesforce Lightning Design System (SLDS). Ultimately, we bring deep accessibility expertise to all functions throughout the product lifecycle.
The shift-left philosophy hasn’t changed, but we must change how we shift to meet the new velocity of code development. How do we scale to meet that 100x increase? By using AI to augment and dramatically accelerate how we include accessibility into our building process, both for engineers within Salesforce and for our customers and partners. This enables everyone building on Salesforce to benefit from our innovations and create accessible experiences from the start.
The A11Y Agent helps builders:
Think of it as having a dedicated accessibility expert influencing and reviewing your entire software development lifecycle in real time, right where you work.
Unlike general-purpose large language models (LLMs), the A11Y Agent is built with deep, curated knowledge specific to the Salesforce ecosystem. This encompasses Lightning Web Components (LWCs), SLDS, and WCAG implementation guidance — as well as Salesforce Enterprise Accessibility Standards, which provide consistent clarity on how Salesforce implements A11Y.
The A11Y Agent understands not just what to fix but why and how. It explains findings in context so developers build lasting accessibility knowledge over time while also accelerating issue resolution to drive productivity gains.
The agent follows a philosophy of assisted remediation. This is a collaborative process, not an autonomous fix. Builders bring the context, the agent surfaces the guidance, and the human reviews and validates the results. Accountability stays with the user, but the overall effort is significantly reduced. Most importantly, quality stays high.
The Accessibility Agent doesn’t take control. It’s built to work alongside you, surfacing critical accessibility suggestions. The final word always belongs to the developer. Our goal is to augment your developer experience, not replace it.
Nell Gawor, Software Engineering Lead, Salesforce
A11Y Agent in action: When a developer asks the A11Y Agent to review an LWC, it identifies specific issues, such as a missing label or missing accessible name, and then generates targeted fixes, including adding aria-label attributes and resolving keyboard accessibility gaps.
The A11Y Agent is available in any tool that supports Model Context Protocol (MCP). Developers can use it via the Salesforce DX MCP package — installable in any MCP-compatible environment — or directly within Agentforce Vibes, where the tools come pre-bundled and only need to be configured. By selecting a profile (such as the front-end developer profile), developers unlock the accessibility-specific tool set, keeping the interface focused and relevant to the task at hand.
Accessibility doesn’t begin when a developer opens their Integrated Development Environment (IDE). It begins the moment a designer sketches a component or a product manager writes an acceptance criterion. Yet for most teams, accessibility reviews don’t happen until code is already written — often too late to make meaningful changes without costly rework.
The A11Y Agent shifts this paradigm. Designers and product managers can use the agent to review component specifications, UX documentation, design mockups, and notes against accessibility best practices before a single line of code is written.
For UX designers, this means submitting a mockup of an interaction pattern or a component spec to the agent and receiving immediate feedback on potential accessibility gaps. Because the agent can process images, it can flag insufficient color contrast, unclear labeling conventions, or interaction models that may not work for keyboard or screen-reader users. Catching these issues at the design stage is dramatically cheaper and faster than discovering them in QA.
For product managers, the agent serves as an early gut check on feature requirements. Does the acceptance criteria account for keyboard navigation? Does the user story consider assistive-technology users? The A11Y Agent’s knowledge base can help surface these questions early, turning accessibility from a late-stage blocker into an upstream conversation.
With the advancements of AI, we’ree all becoming builders. By placing the A11Y Agent at the forefront of ideation, we ensure that prototypes are grounded in what it means to deliver an accessible product. That is truly shifting left.
For teams dealing with large codebases or backlogs of accessibility issues, we’re leveraging the A11Y Agent to address technical debt. The agent can examine a failure, recommend a fix, retest the fix, and once passed, generate a pull request (PR) that an engineer can review and approve, resulting in cleaner code. Unlike IDE-based assistance, the agent can now operate across features that contain debt, even when that feature isn’t actively being modified or iterated on. At Salesforce, the agent runs on top of our automated accessibility testing capabilities powered by sa11y and axe-core. As issues are surfaced, the agent analyzes them, recommends a fix, and then retests them before passing the fix on to the engineer for a human-in-the-loop review and check-in.
Tools like the A11Y Agent don’t solve every accessibility challenge today. However, as we continue to innovate with AI, we’re focused on tangible progress. These new AI-powered tools make it meaningfully easier for more people to build with accessibility that scales from the start.
This article may include references to services or features that are still in development and are unreleased. Customers should make their purchase decision based on fully released and available features.
此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。 原文来自 — 版权归原作者所有。