AI, Data Architecture, and the Rise of Agentic

2026-02-03

As AI continues to drive change across the IT landscape, Field CTO Jon Collins sat down with analyst and data management expert William McKnight on why AI is raising the bar for data maturity, what “agentic” really means in practice, and why data architects aren’t going anywhere.

Jon: How do you see the AI market right now? Is it a bubble?

William: Tech leaders are making circular investments that are driving up the market. Does it have value at the level of investment we’re seeing? Today, probably not, but they say they’re investing for the long haul.

Nonetheless, AI still has tremendous value for enterprises, which can leverage the AI being developed. It’s not nearly as expensive for an enterprise to adopt AI and reap the benefits.

What AI and data-related challenges are you seeing as a primary focus?

Data architecture — because enterprises are trying to add AI to what they have. Nothing is taxing the data infrastructure quite like AI. Sometimes they learn that they don’t have a sufficient architecture to fully benefit from AI. Most larger enterprises don’t really have a single architecture — they have varying levels of data maturity across the enterprise. There’s a lot of redundant data being stored, for example.

I’m not saying the initial architecture has to be perfect. Enterprises can build on what we have and get some AI going, which creates the motivation to do more and develop data maturity. Initial initiatives are putting AI band-aids on top of what they have, which don’t give maximum value. You need a certain level of data maturity in order to get value out of your AI.

What about the architectural direction: “More decentralised… data mesh… lakehouse.”

The architecture style that I see organizations wanting more of, either by name or not, is the data mesh: decentralized components across the organization, creating data products for the rest of the organization and beyond, and sharing those products via APIs internally.

For the foreseeable future, it’s going to be backed by the data lakehouse – the best of both worlds between data warehouse and data lake. Also, there’s a data fabric over the top of all of this. The mesh is how the data is laid out architecturally, and the fabric is the delivery mechanism that lets you connect to data wherever it is.

Plus, MCP will  be adopted as part of the architecture. With MCP, you’re connecting lots of data, giving the agents good context.

Data architecture is a balancing act between what would be ideal — a centralized data warehouse that performs perfectly — and the realities of how initiatives happen in the organization.

That’s why you still need good data architects: they’re not going to be AI’d away in 2026. You still need expertise at the top of this. The data architect can make the right trade-offs between centralization and decentralization which  can create a good data mesh.

So what’s now and next in 12 to 18 months?

Not surprisingly, more AI, and a clamour to turn everything into agentic AI – we‘re seeing the rise of agentic marketplaces. That’s in the budget and the plan for many companies I talk to.

Whether it is realized is a different question. Enterprises are unlikely to get everything they want from agentic in 2026, but they will see ROI and stay committed. Agentic will be with us for quite a while, so it’s worth committing to, even at the low level of maturity that it’s at today.

But beware of the agentic that you adopt, if you’re not creating it. Is it truly AI, or AI-washed? Agentic AI is supposed to use AI, not BI and analytics. Some of the stuff you can see on an agentic AI marketplace is closer to good old analytics. Nothing wrong with that, but the possibilities are so much higher with agentic approaches.

What’s the broader impact, for example, on applications teams or leadership?

So many application teams have been managing the data they need for their application over the years, and not doing a good job of it; there’s no leverage for the enterprise. Data should be managed by data professionals and provided as a core service to the enterprise.

Meanwhile, I’m seeing continued pressure on cost and TCO, including for AI. It’s gone beyond exploratory and “we’ve gotta do it, so let’s do it,” to “How is this driving ROI?” You have to have the ROI skills to show that. So in 2026, leadership is more necessary than ever – someone who can see the big and small picture at the same time and drive efforts towards the enterprise’s goals.

So, business-focused technology leadership. Because AI accelerates everything, it either accelerates you over the cliff or it accelerates you up the mountain.

Yes, more than ever.

Thanks, William!