If your organisation is planning to deploy AI agents to accelerate its business processes, you must start at the very foundation: making your data ready for AI consumption. Agentic AI scales purely on data strength, and a weak foundation will lead to failure.
According to Niels Zeilemaker, Global CTO at Xebia, building an advanced AI agent is meaningless if the underlying data architecture is flawed. “If you don’t think about the foundation, you can build the best agent, but it will never be able to find the correct data; it might misinterpret it or connect fields that should never be connected,” Zeilemaker explains. “These mistakes are not necessarily the fault of the AI agent. It’s the fault of your data foundation.”
The Importance of Advanced Data Cataloguing
One critical area that enterprises must rethink is data cataloguing. While data cataloguing is not a new concept, the rules change entirely when transitioning from human workers to autonomous AI agents.
- The Human Fallback: In traditional setups, if data documentation is unclear, a human employee can simply pick up the phone, walk to a colleague, and ask for clarification.
- The AI Reality: AI agents do not have a “back door” or a colleague to consult. They rely 100% on what is written in the data catalogue. If the description is wrong or incomplete, the agent will completely fail to perform.
Accelerating Data Migration with Xebia Axis
To solve this issue, Xebia has developed Xebia Axis: Agentic Data Foundation (ADF). This solution extends traditional data platforms to host AI agents securely, utilizing them for both customer-facing use cases and internal operations.
By combining purpose-built AI agents with expert engineering, Xebia is able to compress traditional 12- to 24-month data migration timelines into fixed-price, milestone-bound engagements, helping enterprises modernise their infrastructure faster than ever.
Eliminating the Risks of “Vibe Coding”
Another major challenge for enterprises today is maintaining control over AI-generated code. Zeilemaker points to the trend of “vibe coding”—where anyone can prompt an LLM to build an application, but no one dares to actually push it into live production due to quality and security fears.
To address this, the company offers Xebia ACE: AI-Native Software Engineering, a framework that embeds AI across the entire Software Development Lifecycle (SDLC). When implemented correctly, this framework can:
- Accelerate software delivery by up to 40%
- Cut legacy transformation costs by up to 70%
“Xebia ACE gives you a very nice framework to use LLMs in coding without the risk of losing control or governance in the process,” says Zeilemaker.
The Future: AI as a Senior Team Member
As AI-driven code generation increases, maintaining cybersecurity remains paramount. Zeilemaker notes that the industry is moving towards integrating advanced LLMs as “automated reviewers” for code before it goes live.
By adding a highly sophisticated language model to the process, enterprises essentially gain a very senior, automated team member capable of performing deep third-party security reviews. Ultimately, whether an organisation is just assessing its data readiness or is fully prepared to build, setting the correct data foundation remains the absolute key to true digital transformation
