Agentic retrieval is a hot topic for a reason: Every company wants to automate work, but they often just are not ready. It’s also extremely easy — and tempting — to spin up an agent, point it at a mountain of data, and spend heavily on tokens for a lackluster result.
To unleash agents that can accomplish work, you’ll need a different approach. You can’t throw a more intelligent AI model at a failing agentic workflow without first sorting out how to provide it with better data. It won’t do much good.
So how should companies prepare their information for quick, accurate agentic use? That’s what The New Stack will be digging into live with Vespa.ai’s Jenny Morris on Wednesday, December 2.
You have successfully registered for the webinar.
You might layer a tailored model into your agentic workflow to help translate between an agentic request and the dataset it’s querying. While this idea has merit, it can increase latency by adding another entire step to a search; who wants to go slower, or consume more tokens?
Vespa reckons that the way forward is to help agents handle complex, and often multi-threaded contemporaneous searches themselves. How? By providing agents with a map (markdown file) that covers how search tools work; without clearly marked fields, ranking profiles, and other tools are not expressly explained to the agent, it may decide to rip a generic query, leading to slower response times and less pertinent information.
Surprised that a markdown file could be so powerful? Don’t be. Vespa is following in Anthropic’s footsteps, donating agent skills to the Linux Foundation. Words are incredibly powerful.
In this webinar, we’ll aim to better understand how to move the corporate world from agent-curious to agentically enabled. So, we’re sitting down with Morris to learn more about what her company calls an “agent-ready schema.” Most critically, we’ll explore how hard it is to fix at least part of the context gap.
Morris will dig into all of this live in “Teaching AI Agents to Search Like Experts” on Wednesday, December 2, 2026, at 9:30 a.m. Pacific, with time to answer your questions throughout. Register now to join us.
What you’ll learn:
- Why model capability cannot fix retrieval errors, and how those errors compound across multi-step tasks
- Why you should check what an agent actually retrieves before changing the model
- Why human-facing search tooling is insufficient for agents
- How ‘Agent-Ready Schema’ and ‘Schema-to-Skill’ help agents construct queries using the right fields, filters, and ranking strategies
The post Here’s how to teach AI agents to search like human experts appeared first on The New Stack.
