AI collaboration: A data leader’s advice for CIOs

AI's impact is significantly influencing most parts of enterprise businesses in 2026, leading to confusion over who has ultimate ownership -- IT, or business operations. Where does this leave CIOs?

In this Q&A, Bharath Krishnamachari -- senior director of data and AI at foodpanda, foodora and Yemeksepeti -- speaks to TechTarget about lessons he’s learned from building and leading an AI center of excellence for Delivery Hero brands.

Editor's note: This Q&A has been edited for clarity and conciseness.

AI is touching so many parts of the enterprise, leading to confusion over who has ultimate ownership. What’s your take on this and how are you approaching it at foodpanda, foodora and Yemeksepeti?

Bharath Krishnamachari: I'll give you a sense of how we've been approaching it and where we're heading, and what I’m seeing in the industry more broadly.

We've been approaching it with a two-pronged strategy. One prong is a bottom-up approach that asks: “How do we get everyone in the organization enabled with AI?” Nothing beats on-the-job learning, letting people safely get their hands dirty, and playing with the technology in the functional areas. Everyone has access to Gemini, and we’ve been encouraging folks to use it. This can empower people to identify opportunities for automation or things that just couldn't be done before.

The second prong is more top-down. We want to reimagine workflows, and you’re not going to do that bottom-up. You need to work with executive sponsorship at the C-level. Through this, we've created an AI center of excellence. We've pulled some engineering capacity, and have got a product manager, and data and prompt modelers. We've come up with a framework for how we take requests from different functional areas for AI initiatives, and then we act on them with top-down executive buy-in.

It sounds like the C-level is really getting behind AI at foodpanda, foodora and Yemeksepeti.

Krishnamachari: It is absolutely key for the C-level. The CEO of foodora just recently told me he's happy to give more than 20% of his time per week to this topic. That's how seriously he thinks about it.

In terms of getting resources, that is not an issue. We are working on some pretty big opportunity areas right now. We've already realized some value in some of those workstreams, and we're going to scale this. I'm hoping for more of a snowball effect.

Let’s talk about AI collaboration between CIOs and other functions. What have you seen that’s worked?

Krishnamachari: In this day and age, one thing that AI has done is reduce time-to-value expectations enormously. Nobody's willing to wait a quarter to roll something out. When they have an idea, they want to get it in front of stakeholders pretty rapidly. You do have to think about the security risks, too.

I think that the way you must think about [cross-function AI] is: “How can I make this a platform?” ‘Platformizing’ means that different teams have a mechanism in which they can at the very least build. You have to have those hard conversations about what you're willing to cede in terms of access so folks can get platformed, but if you're partnering with the likes of Google, for example, you already have the advantage of tools that are safe, secure, and integrated into the Google Workspace. Making sure everybody knows about this means they will actually use it and drive adoption.

It's no longer a time to lean back. You're going to leave your organization behind if you're not leaning in.

How about shadow AI. What's your company's policy on that, and how are you managing it?

Krishnamachari: Shadow AI is a risk. The one way we're addressing it, is that you have no need to use shadow AI, because you get access to pretty much the state-of-the-art tech. Most people in a tech and product role already have access to Google Workspace and Gemini. So, there's very little incentive for you to go outside of that.

For folks who are not in the tech and product umbrella, shadow AI doesn't limit itself to coding tools. It could be meeting note takers, productivity tools or tools highly relevant to a function. We've basically been reiterating to our employees that you have to go through the approved IT list. You should not be installing solutions on your own and should ensure you follow the standard IT procurement process.

We could take a very strict approach: just block certain things. But I don't know how far that's going to go, to be honest. I'm sure there are companies that have tried to clamp down on it 100% -- I don't know how successfully that is working for them. We're trying to take an approach that really trusts employees, which has always been the case. Whether it's AI or other software, there's always been tools that folks like, and it's on them to make sure they don't use unapproved software. We do have our IT security teams constantly checking for that as well.

What about regulatory and governance -- what’s your advice here?  

Krishnamachari: I'd be lying if I said we have it 100% figured out. At the end of the day, AI or no AI, who should have access to what data?

An example is something like a human resources system. Maybe certain folks can get access to the org hierarchy and so on. Anything beyond that -- say, if you're starting to go into performance information, then you must get legal involved.

As such, we will have what we call a ‘chapter’ -- like the Atlassian Spotify model -- where you have squads and domains to make sure that people, including forward-deployed folks, understand what is and what is not kosher from a risk management and security standpoint.

That's why we don't want to just let it be, "Here's your tools, here's your tokens, go out and do whatever you please." We are making sure that we have controls in place. Anything that really has to go into production -- Workday, Jira or a data infrastructure -- we want to know what it is that you're doing and clear the use case.

Harriet Jamieson is a senior manager on the IT Strategy team at TechTarget.