Ping YOUniverse: How AI lets identity management go ‘headless’

AUSTIN, TEXAS — Agentic AI will relieve administrators of much of the burden of running identity-management systems, permitting the systems to go fully "headless" and run without a graphical user interface, Ping Identity VP of Product Management Kate Atkinson said at the Ping YOUniverse conference here today (Sept. 2).

"Ping has been headless for years, but what's new is AI-first headless identity," Atkinson said in front of a slide proclaiming that "The Age of Human-Only Software Is Over."

"Agents can now configure and manage the Ping Identity Platform so you don't have to," she added.

Please click here to watch a recording of Atkinson's keynote address at Ping YOUniverse.

Instead of clicking buttons and filling out fields on a platform-management dashboard, Atkinson said, administrators will be able to use a command-line interface (CLI) to communicate with an AI agent using natural language.

She provided examples of commands and questions: "Set up passwordless with an iOS application," "Tell me why Bob can't log in," and "Has my configuration changed in production?"

Many companies will build their own AI interfaces and use their own agents to run the Ping Identity platform. Training agents can get expensive due to token costs, but Ping is making it simpler and cheaper, Atkinson said.

Instead of providing documentation about its platform as HTML or CSS files, the company is providing context in the form of easy-to-read files formatted in Markdown, a simple markup language that can convey document formatting in plain text.

"In our own testing, we see 70% less token usage by just training them on Markdown files," Atkinson said.

Guidance is provided by skills, transmitted as a reference library that AI agents can use as a resource. Guardrails are provided by MCP servers that direct agents to reliable, self-describing tools.

"Without guardrails, a wandering agent can create a production vulnerability," she pointed out. "The agent needs context. It can't guess."

With this framework, Atkinson explained, developers building out their own implementations of the Ping Identity platform can stop using the human-centric method in which developers must manually look up documentation and then use GUIs and APIs to interface with the platform.

Instead, the developers can use a CLI to communicate with a coding agent like Anthropic's Claude Code or OpenAI's Codex. The agents in turn can refer to documents for context and skills for guidance, then interact with the Ping platform using tools provided by the MCP servers.

All of this is free for existing clients of the Ping Identity platform, Atkinson said, and further features and tools will be released in the next six months.

She capped off her presentation with another slide: "When AI makes identity look easy, Ping Identity is the platform you want underneath."