If your goal is to integrate AI agents into your organization to speed up operations, you need to begin with the fundamentals—making your data accessible and usable for AI. According to Niels Zeilemaker, Global CTO at Xebia, the effectiveness of agentic AI depends heavily on the strength of your data foundation.
“Without addressing this, you might build a highly capable agent, but it won’t be able to locate accurate data, may misinterpret information, or incorrectly merge unrelated data fields,” Zeilemaker explains. “These errors aren’t always the agent’s fault—they often stem from a foundation that hasn’t been properly prepared for AI agents.”
Zeilemaker highlights data cataloguing as a critical area requiring special attention. Although this practice isn’t new, its importance significantly increases when implementing AI agents. “When creating a data catalog for an organization staffed by humans, there’s always a backup option,” he notes. “If documentation appears unclear, employees can contact colleagues directly to understand how to work with specific datasets.
“AI agents lack this direct communication option. They must rely solely on the information in the data catalog, meaning inaccurate descriptions will lead to poor agent performance.”
Xebia specializes in helping organizations transform AI strategies into production-ready solutions that deliver rapid, meaningful business transformation. The company prioritizes people-first principles and uncompromising quality, but Zeilemaker emphasizes that knowledge sharing stands as perhaps their most valuable approach—evident through participation in events like TechEx Global North America.
“Knowledge sharing is crucial for us, enabling us to stay ahead of industry trends and adapt quickly to market changes because we’re constantly eager to explore new developments and exchange insights about effective approaches,” Zeilemaker explains. “Through this commitment to collaborative learning and innovation, we establish ourselves as authorities in specific domains.”
Data and AI expertise clearly falls within this sphere. At AI & Big Data Expo, Zeilemaker presented attendees with practical guidance on building robust AI foundations and integrating scattered data environments. He provided straightforward insights into how combining specialized AI agents with expert engineering can compress typical 12-24 month timelines into fixed-price, milestone-driven projects.
Central to this approach is what Xebia designates as the Agentic Data Framework (ADF)—a methodology that enhances data platforms to support AI agents, leveraging them for both customer-facing applications and internal business processes. While organizations have shown strong interest in transitioning from outdated to modern platforms, Xebia observes growing client demand for faster, more reliable migration approaches. Zeilemaker indicates this involves collaborative solution development between consultants and clients.
“AI agents must depend on the data catalog’s descriptions, and any inaccuracy in those descriptions will cause agents to fail”
“Having completed traditional migrations and some accelerated through LLM-assisted coding, we’re now incorporating AI capabilities directly into data platforms, utilizing the additional context they provide to further expedite migration processes,” he states.
This accumulated knowledge shaped Xebia Axis: Agentic Data Foundation—their comprehensive solution for helping enterprises prepare data for AI consumption more efficiently than alternatives.
Xebia’s additional key capability is Xebia ACE: AI-Native Software Engineering, a comprehensive framework integrating AI throughout an entire organization’s software development lifecycle (SDLC). Proper implementation can boost delivery speed by up to 40% while reducing legacy transformation costs by up to 70%.
Zeilemaker observes that Xebia ACE proves particularly valuable for larger enterprises wishing to maintain specific governance standards or traditional workflows within their SDLC. Beyond immediate benefits, there’s a broader significance here. He cites vibe coding as an example: “Consider vibe coding—anyone can develop applications this way, but few organizations will actually deploy these creations in production environments,” he explains. “By implementing ACE, you retain the productivity benefits of LLM acceleration while ensuring output quality matches historical standards.
“For organizations considering AI integration in software development, Xebia ACE provides an excellent framework to leverage with minimal risk, avoiding the problems of unmonitored LLM implementation where control and governance may be compromised,” Zeilemaker adds.
For enterprise clients, maintaining control is paramount. With substantial volumes of AI-generated code, the development lifecycle could introduce security vulnerabilities. Zeilemaker acknowledges this remains an industry-wide challenge requiring resolution, though he notes considerable interest in Anthropic’s recent introduction of a pull request review feature.
“This development represents an engaging trend likely to gain further adoption,” he remarks. “Organizations conducting extensive code reviews before major production releases can now incorporate an LLM functioning as an experienced reviewer providing independent evaluation.
“I anticipate we’ll witness more innovations following this pattern in coming years.”
Regardless of where organizations stand in their AI journey—whether evaluating data readiness or preparing to implement solutions—Xebia assists in establishing solid foundations and building transformative capabilities upon them.
Photograph credit: fabio via Unsplash
Interested in expanding your AI and big data expertise through industry leader insights? Explore upcoming events hosted in Amsterdam, California, and London. These comprehensive TechEx events feature co-located leading technology conferences—click here for additional details.
AI News receives support from TechForge Media. Discover additional upcoming enterprise technology events and webinars by clicking here.
