A conversation with Arnaud le Roij, Head of Sales at Cuez, ahead of IBC2026.
Arnaud* talks to broadcasters every day. Ahead of IBC2026, we asked him what he is actually hearing from newsrooms and what we should expect to see on the show floor.
Key takeaways
- Interoperability, not AI, is the unsolved problem. Without it, AI doesn't actually do anything for you.
- AI moved from a strategy slide to a built-in expectation. Nobody pays a premium for it anymore.
- Trust, not fear of AI going rogue, is the real hesitation. Synthetic news is now polished enough to fool journalists too.
- MCP enablement only works with interoperability in place first. Clean metadata, open APIs, and one persistent story ID decide if AI actually works.
- Openness is becoming the real differentiator. The longer broadcasters wait to decide, the more expensive that decision gets.
- The biggest blocker isn't the technology. It's getting people to actually change how they work.
*Arnaud le Roij is Head of Sales at Cuez, with a background in commercial strategy at Toyota Belgium and Just Eat Takeaway. He stays close to customers, helping them see why change is worth it, not just what the technology does.
What's the biggest shift in what broadcasters are prioritizing at IBC this year compared to last?
— Last year everyone was talking about AI and automation, and the DPP research in Paris showed the same. That's still true. What surprises me is how often interoperability gets treated as solved, because in almost every conversation I have, it isn't. I can't tell you what broadcasters will prioritise at IBC, that's their call. I can tell you what I keep running into. It's usually one of two things.
The first is "make this talk to what we already have." That's an attempt at interoperability, and I understand it completely. You've got multiple systems that can't speak to each other, they're paid off, they work, everyone knows them. Nobody wants to throw that away. So the ask goes to whichever vendor claims to be open and agnostic: come and be the glue for our setup. Which often means bending over backwards to connect things because sometimes to connect to half of what that same solution could have replaced outright.
The second is a tender for something new that has to work in exactly the same way, over the same protocols we've had for fifteen years. Also reasonable. Also not interoperability.
I sat with a broadcaster recently who was asking all the right questions about data. Genuinely good vision, clear on what the model needed to look like. It was impossible to build, because none of the surrounding systems was open enough to let it happen. Not a knowledge problem, not a budget problem, just the estate wouldn't allow it.
Which is why my AI conversations are mostly not about AI. They're about API connectivity, MCP enablement, working across platforms. You can buy the most capable system on the market, but if the things around it can't do the same, it's a Ferrari with a permanent 30 km/h limiter (19mph for the Americans). You'd have been just as fast on a bike.
How is broadcasters' approach to AI different in 2026 vs. 2024?
— In 2024, AI was a strategy slide. Every broadcaster needed a position on it. By 2026, it stopped being its own budget line and became an expectation inside tools that were being bought anyway. Nobody's paying an AI premium now.
What's actually running and what's still in pilot is still split, though. Reversible things like tagging, subtitling, archive search, compliance logging — anything where a person sees the output before the audience does, is already present in newsrooms. Anything writing to air unsupervised is still a pilot, and has been for two years.
The next question, and the one I'd be asking any vendor, is whether you can connect your own LLM. If you can't, you're using something that has no idea how your organisation works and only ever sees what's inside that one tool. There's a ceiling on how useful that can be, and you hit it fast.
The hesitation was never really about AI going rogue. It's about trust. That's exactly why it matters more now, not less.
What's one AI trend you'd call overhyped right now, and one you'd call underrated?
— Synthetic presenters for daily news. Not because the quality is bad — it's improved fast. Because nobody has answered the boring question: when one gets a fact wrong, who owns the correction, and how quickly can you pull it? Until there's an answer, it stays a stand demo. It's a long way from a live gallery in the middle of a normal week.
Underrated is everything that happens before air. Clean metadata, open APIs, a rundown a machine can read. And specifically: one story ID that survives across every system it touches. Almost nobody budgets for that, and it decides whether any of the AI works in your setup or not.
I sell newsroom solutions and automation, so weigh that as you like. But the broadcaster I'm thinking of wasn't blocked by a model. They were blocked by a data layer that didn't exist.
What's one prediction for broadcast tech in the next three years?
— The buying question is shifting from what does this tool do to what does it connect to. Being open becomes the real differentiator.
I'm careful not to overstate the pace, though. Looking at how little actually changed over the last year, I doubt broadcasters will move much faster. What I am confident about is the cost of waiting: the longer a decision gets postponed, the more expensive it becomes, until tools have to get dramatically cheaper before change is even possible.
One live example: everyone is jumping on MCP enablement right now, and rightly so, but without interoperable solutions around it in place first, an MCP layer is a fancy add-on that only gets you so far.
What's the most surprising thing a customer has told you in the last six months?
— That the technology is rarely the blocker; the change management is. This statement itself is not the surprise, as this is a reality we encounter daily. The surprise was the level of awareness, since the majority have not expressed this and tie it back to the technology.
I see the same pattern, broadcasters ask for new technology to replicate the old system exactly. Part of it is mindset. Some broadcasters still run systems over 30 years old. The instinct is still to stick with what's known or ask for something that is new, but does exactly what the old system does (and preferably with the same interface).
That's understandable, because change is challenging, maybe even scary sometimes. But doing the same thing over and over and waiting only deepens the water. The change gets bigger, so it feels scarier, so it hits harder, and resistance grows, until the decision gets so pressing that you nearly make it in a hurry or when it's already too late and has impacted the entire organization and people in it, with an immediate ROI attached. Deciding earlier is cheaper. And seriously considering change every 3 to 5 years is not insanity. And I mean seriously, not just because you need to renew your contract and do a benchmarking exercise and launch your next tender that will be written on the body of your current provider. It's healthy in a world where everything changes so fast.
What's your advice for broadcasters walking the floor this year at IBC?
— Book fewer meetings and make them longer. 6 real conversations beat 20 badge scans.
At a stand, don't ask for the tour. Ask what it connects to, have them show it rather than describe it, don't focus on feature lists but focus on workflows and ask what happens on day 400, not day 1. If you take one question around the halls, take this one: can another system read this out and write back into it, without you building it for me first? Watch how quickly that turns into a scoping conversation.
And visit one hall that has nothing to do with your job. Most good ideas are borrowed from somewhere adjacent.
Ecaterina Caprari
Content Marketeer