Xebia: On building the data foundation for AI agents – and then accelerating
If your remit is to assist your organisation add AI agents to speed up its processes, you must begin at the foundation – and meaning making your data obtainable for AI consumption. Agentic AI scales on data power, as Niels Zeilemaker, world CTO at Xebia, explains.
“If you don’t take into consideration that, you’ll be able to construct the finest agent, however it’s going to by no means be capable of discover the right data; possibly it’s going to misread the data, possibly it’s going to be part of completely different fields collectively in your data which ought to by no means be related,” explains Zeilemaker. “And these errors usually are not essentially the fault of the agent. It’s the fault of your foundation, which isn’t prepared for AI agents.”
One space to significantly think about, Zeilemaker notes, is data cataloguing. It’s not a brand new idea, however the sport modifications for agents. “If you’re establishing a data catalogue for an organisation solely consisting of people, there’s at all times a fallback,” he says. “If there’s one thing probably not properly documented, you’ll be able to choose up the cellphone, stroll to a colleague, and have a kind of again door, in ‘how ought to I work with this specific set of data?’
“Agents don’t have such a again door. They need to depend on the data catalogue, what’s written there, and if the description is flawed, the agents won’t carry out.”
Xebia’s focus is to assist organisations flip AI technique into production-ready options which drive actual transformation quicker. The firm’s core values embrace being individuals first and high quality with out compromise, however maybe the most necessary, as Zeilemaker sees it, is sharing information – equivalent to at occasions like TechEx Global North America, at which Xebia participated.
“I feel sharing information is essential for us, and it additionally permits us to be a bit forward of the curve, undertake rapidly to new modifications in the market, as a result of everyone has this eagerness to search out out new issues, and to share what works, what doesn’t work,” says Zeilemaker. “By pushing quite a bit into this sharing information and innovation, we attempt to additionally choose a few domains the place we wish to be the authority.”
Data and AI is evidently a kind of areas. At AI & Big Data Expo, Zeilemaker advised attendees easy methods to construct this AI foundation and unify their fragmented data landscapes. It was an trustworthy account of how combining purpose-built AI agents with professional engineering compresses a 12- to 24-month timeline right into a fixed-price, milestone-bound engagement.
The overarching thread for that is what Xebia calls Agentic Data Foundation (ADF), which extends the data platform to host agents, and then make use of them each in customer-facing use instances and inner processes. While there has at all times been an enormous urge for food in migrating from legacy to trendy platforms, Xebia is seeing extra prospects asking for an strategy to extra rapidly – and reliably – migrate into data platforms. Zeilemaker says that is the place advisor and buyer are co-developing the resolution.
“After doing migrations the old school manner, and accelerating some with LLM coding, we at the moment are integrating this into the data platform, making use of the further context it could possibly present to speed up migrations even additional,” he says.
That gathered expertise is what formed Xebia Axis: Agentic Data Foundation, Xebia’s reply to serving to enterprises make their data AI-ready quicker than any various.
Another weapon Xebia has in its arsenal is Xebia ACE: AI-Native Software Engineering, a framework which embeds AI across an organisation’s entire software development lifecycle (SDLC). Done proper, supply will be accelerated by as much as 40%, whereas legacy transformation prices are minimize by as much as 70%.
Zeilemaker notes that Xebia ACE is especially helpful for bigger enterprises who ‘possibly nonetheless wish to keep on with a specific governance or manner of working whereas doing SDLC’. Yet there’s a larger image right here. Zeilemaker makes use of vibe coding for instance. “If you consider vibe coding, everyone can create an app, however no person is daring to truly push these apps into manufacturing,” he says. “If you undertake ACE, you continue to get lots of the advantages of the acceleration of LLMs, however you’re nonetheless having the similar high quality finish outcomes as you’re used to in the previous.
“If you’re trying to make the swap to utilizing LLMs in coding, Xebia ACE offers you a really good framework to make use of, with out the threat, or any drawbacks of doing darkish manufacturing facility LLM and hoping for the finest – and shedding a little bit of management or governance in the course of,” provides Zeilemaker.
For enterprises, that management is vital. With a lot code being generated, the AI-driven SDLC may grow to be a safety weak spot by vulnerabilities. Zeilemaker argues it’s one thing the trade nonetheless wants to determine to a level, however notes with curiosity the recent move by Anthropic to launch a pull request reviewer.
“It’s an fascinating one, which we’ll most likely see extra of,” he says. “There can be very prolonged pull request evaluations, which you apply everytime you go and attempt to do a brand new manufacturing launch. And then you add a really senior group member in the type of an LLM to your course of, which does a kind of third-party evaluation.
“I feel that’s an fascinating angle with what we’re going to see extra of in the future.”
Ultimately, wherever organisations are of their journey, from assessing their data readiness to being able to construct, Xebia is ready to assist get the foundations proper – and create the transformations on prime of it.
Want to study extra about AI and huge data from trade leaders? Check out happening in Amsterdam, California, and London. The complete occasion is a part of TechEx and is co-located with different main expertise occasions, click on here for extra info.
AI News is powered by TechForge Media. Explore different upcoming enterprise expertise occasions and webinars here.
The publish Xebia: On building the data foundation for AI agents – and then accelerating appeared first on AI News.
