How to Get Started with AX
While we’re still identifying concrete best practices, here are some ways to apply what we already know in the real world:
-
Identify human engagement areas: Determine the key areas where users will want AI agents to act on their behalf. These might include customer service, content creation, workflow automation, or data management. Understanding where humans want to engage with your service will help you prioritize AX assessment and improvement efforts.
-
Assess agent readiness: Evaluate whether your website is structured for AI interaction. Are APIs well-documented and machine-readable? Can AI agents easily understand the system’s functionality and execute necessary actions without excessive human intervention? Can agents access your documentation or is it behind a paywall or some other gating that prevents them from supporting your customers?
-
Optimize documentation for AI: Provide structured, machine-readable documentation that AI agents can parse and understand. This can include offering OpenAPI specifications, embedding structured metadata, and ensuring that documentation is regularly updated and accessible agents. Generating tool-specific instructions for popular tools in your industry (like
.cursorrulesfor developer tools) take this one step further at helping agents local to the user’s computer achieve better results with your site or service. -
Simplify access controls for agents: Ensure AI agents can easily obtain the necessary credentials, permissions, and data to function effectively. This may involve creating secure, token-based authentication systems or designing clear role-based access controls tailored for AI agents. Clear access controls will also aid your support, security, and compliance teams understand who/when actions were taken.
-
Use agents to test your product: There’s no better way to understand the capabilities of agents using your service or platform than to use tools (the same tools your end users use) that put agents to task working with your offering. This will organically help you understand the end user’s experience and it will reflect back at you how effective your agent experience is. Do this while your assessing agent readiness, after you’ve made improvements, and as new techniques come out.