Xebia: AI agents need data catalogues, ADF and ACE

Niels Zeilemaker, Xebia’s global CTO, warned at AI & Big Data Expo and TechEx Global North America that AI agents will fail unless organisations make their data discoverable and well documented. He recommended using Agentic Data Foundation (ADF) and Xebia ACE to prepare data for AI and accelerate migration work.

Xebia describes ADF as an extension of a data platform that can host purpose-built agents for customer-facing services and internal processes. The company says combining agent design with engineering can shorten migration projects that it estimates often take 12 to 24 months into fixed-price, milestone-based engagements.

Zeilemaker focused on data cataloguing as a practical requirement for agents. He noted that unlike human staff, agents cannot call a colleague for clarification and must rely on catalogue entries. He warned: “Agents have to rely on the data catalogue, what’s written there, and if the description is wrong, the agents will not perform.”

Xebia ACE is presented as an AI-native software engineering framework that embeds AI across the software development lifecycle. The firm says ACE can speed delivery by up to 40% and cut legacy transformation costs by up to 70% while allowing organisations to keep existing governance and quality controls. Zeilemaker used the term “vibe coding” to describe apps created with low-code or LLM-assisted tools that are often not ready for production, and he argued ACE helps keep quality while using LLMs.

The company is also integrating LLM-assisted coding into migration workflows. Xebia says it has used LLMs to speed parts of traditional migrations and is now embedding those capabilities into the data platform so agents and platform context can help automate and validate migration tasks. The firm said several clients are co-developing solutions to shorten timelines and reduce risk.

On security and governance, Zeilemaker said large volumes of generated code could introduce vulnerabilities and pointed to recent work on an automated pull-request reviewer as an example of how LLMs might be used in review processes. He added: “There will be very lengthy pull request reviews… and then you add a very senior team member in the form of an LLM to your process, which does a sort of third-party review.”

Xebia said it participates in industry events to share practices and that ADF and ACE are intended for organisations assessing data readiness and moving AI agents from pilots into operational systems.