# Weekly Update: Accessibility and Inclusion in Agentic AI

This week brought important developments showing how agentic AI is becoming available to more people and organizations. The focus is shifting from experimental systems used only by specialists to practical tools that businesses of all sizes can use safely and responsibly.

## Government Standards Shape Accessibility

One of the most significant developments involves government leadership on accessibility. The U.S. government has updated its design mandates to require that accessibility, trust, and usability are non-negotiable principles—not just optional extras. This means that any AI system built for government work must be designed so that everyone can use it, regardless of their abilities or background. This standard is spreading to the private sector, with companies recognizing that when they build agentic AI systems, they need to think about accessibility from the very beginning. These government standards are setting the bar higher for everyone, making sure that advanced AI tools don't leave anyone behind.

## Democratization Through Open Models and Tools

A major breakthrough for making AI more accessible came from DeepSeek's democratization of powerful AI models. Previously, only large, wealthy companies could afford to build cutting-edge AI agents. DeepSeek changed that by creating efficient, powerful models that people and small companies could use. This sparked what happened on "DeepSeek Monday" in January, when a Chinese AI application reached the top of the US App Store. This showed that innovation in AI could come from anywhere in the world, not just Silicon Valley. Alongside this, companies are actively lowering technical barriers by creating workflow builders and platforms that let people without coding experience build custom agent systems.

## Open Frameworks Making AI Agent Building Easier

IBM's BeeAI Framework demonstrates how companies are focusing on making agent development accessible to developers. Initially, IBM tried making agents for regular business users, but they discovered what developers really wanted: an easy way to build agents without learning multiple complicated systems. The open-source framework now has over 3,000 GitHub stars and supports multiple AI models, meaning developers can choose the tools that work best for them. Similarly, Amazon's Bedrock Agent Framework includes features like session isolation and episodic memory, designed to help developers build agents more easily. These tools are framework-agnostic, meaning they work with agents built different ways, promoting flexibility and choice.

## Standards for Interoperability Benefit Everyone

The creation of the Agentic AI Foundation by the Linux Foundation represents a major step toward making different AI agent systems work together. Before this, companies had to build agents using incompatible tools. Now, through standards like the Model Context Protocol and Agent2Agent protocol, agents built by different companies can communicate and share resources. This interoperability benefits smaller organizations and developers who might not have resources to build everything from scratch. They can now combine tools from different sources, making it easier and cheaper to create solutions. Think of it like how websites from different companies can all display properly in your browser—standardization creates freedom, not restriction.

## Enterprise Accessibility: Professional Services Leading the Way

The Big Four accounting firms (Deloitte, EY, PwC, and KPMG) are making agentic AI accessible across their organizations. EY launched EY.ai, giving 80,000 tax employees access to 150 AI agents for everyday work tasks. PwC deployed 25,000 intelligent agents across client operations since launching their agent platform in March. This shows how agentic AI is moving from experimental labs into mainstream business operations where regular employees use it to do their jobs better. These systems are making work more accessible by automating tedious, repetitive tasks and letting humans focus on more creative and strategic work.

## Building Trustworthy Systems Everyone Can Rely On

As agentic AI becomes more widespread, companies and experts emphasize the need for clear governance and safety guardrails. The missing safeguards include auditability, decision logging, and strict boundaries on what agents can do without human approval. When companies build accessible AI systems, they must also build systems that people can trust and understand. Without these controls, even the most powerful AI tools become risky and unreliable. Industry leaders recognize that 2026 will mark the shift from experimental AI to trusted, accountable agentic systems. This means accessibility includes not just ease of use, but also transparency and safety—making sure the AI systems help everyone, not just their creators.