Mediagenix’s Ivan Verbesselt on why the media industry needs to think in real time

The fashion industry, Ivan Verbesselt argues, has something to teach broadcasters.

Companies that make physical goods – fabric sourced across continents, supply chains that span months – have figured out how to read early market signals and scale production in near real time. Media companies, which deal in digital content with no manufacturing lag, have not.

“It hurts me to see that the media industry wouldn’t be able to do it,” said Verbesselt, chief strategy officer at Mediagenix, speaking at the 2026 NAB Show. “It’s digital in and out.”

That gap, between the signals audience behavior generates and the speed at which media companies act on them, is the core of what Mediagenix is calling the real-time media enterprise. It is a framing that sounds straightforward but, Verbesselt argued, runs into some genuinely hard problems the moment you try to put it into practice.

What real time actually means

The concept starts with a familiar frustration: broadcasters know what audiences watch, but struggle to translate that knowledge into faster decisions about what to program, how to bundle content and what to surface next.

Part of the problem, Verbesselt said, is that media companies treat “doing more of what works” as a simpler directive than it actually is.

“Is the success coming from the content itself, or from the way it was scheduled, the context it was scheduled in, the format, the channel it was promoted on?” he said. “Doing more of what works is not so trivial as it seems.”

That question, what actually caused a piece of content to perform, requires what Verbesselt calls semantic intelligence: a deep, machine-readable understanding of what content actually is, not just what it is labeled.

Advertisement

Knowing that a show performed well is only useful if you can identify which attributes drove that performance and find or commission content that shares them. A recommendation engine that tells a programmer to air more shows like a medical drama cannot distinguish between whether the audience was drawn to the medical setting, the procedural format or a specific character archetype.

“What is House’? Is it ‘Gray’s Anatomy’? Is it ‘The Good Doctor’? Or has it nothing to do with medicine — is it maybe a bit of a strange character?” Verbesselt said. “That understanding — being able to swap out content in real time — that’s the key.”

Two speeds, one flywheel

Verbesselt described the content lifecycle as operating at two distinct tempos that need to be connected rather than run in parallel.

The first is strategic: decisions about what content to commission, acquire or drop, made on a cycle of months or years. This layer, he said, governs roughly 5% of a media company’s content at any given time. It is important but inherently slow.

The second is tactical: the real-time decisions about how to schedule, bundle and surface the content that already exists. This is where Verbesselt said the biggest untapped value lies, because it operates on the library a media company already owns.

“Nothing is easier than monetizing the content you already have,” he said. “Only, it’s not easy. It seems to be very difficult.”

The difficulty is partly technical and partly organizational.

Most content management systems are built for batch processing, schedules are set days or weeks in advance, home screens are curated manually and updated infrequently, FAST channel lineups are assembled statically. Adapting any of those to respond to live audience signals requires both the infrastructure to ingest those signals and the tooling to act on them without requiring manual intervention at every step.

Verbesselt compared well-executed linear programming to museum curation, a sequence of experiences designed to hold attention across time, and argued that the same logic applies to continuous channels. The challenge is to do that curation at a speed and scale that match the volume of channels now being operated.

“There’s more and more FAST channels that need to be created,” he said. “We went from automation to optimization. And there’s no end, I think, to the kind of KPIs we’re going to be wanting to optimize against — because they’re contradicting each other.”

Cost and engagement are the obvious tension: the most expensive content is not always the content that retains a specific audience segment, and the most engaging content for one demographic may be the wrong choice for another. Building systems that can navigate those tradeoffs in something approaching real time is the technical problem Mediagenix said it is working to solve.

Linear as a model, not a legacy

One of Verbesselt’s more pointed observations at the show was about what media companies overlook when they treat linear as a declining format.

Advertisement

TikTok, he argued, is a linear experience. So is the Instagram feed, and so is YouTube’s autoplay. The most engagement-effective digital platforms are all, in some functional sense, continuous streams that remove the decision of what to watch next from the viewer.

“The eternal scroll is the quintessential hyper-personalized linear experience,” he said. “The most addictive engagement with media is linear.”

The difference between those platforms and traditional linear broadcasting is not the format — it is the speed at which the playlist adapts.

TikTok updates its recommendations in milliseconds based on how long a viewer pauses on a given video. A traditional linear schedule is set days in advance. Closing that gap, Verbesselt argued, is not about abandoning linear but about making it responsive.

He noted that Netflix, often cited as the benchmark for streaming personalization, still falls short of what the concept implies in practice. Home screens continue to prioritize high-investment originals regardless of individual viewing patterns. Thumbnails are personalized; the content hierarchy largely is not.

“There’s a hero, and the whole first three pages of scrolling through doesn’t feel very personalized,” he said.

On artificial intelligence, Verbesselt was measured in a way that cut against the prevailing tone at the show.

Generative AI, he said, has genuine applications in the content lifecycle, metadata enrichment, natural language search, editorial briefing tools, but is not the right tool for the scheduling and optimization problems that are Mediagenix’s core focus.

“We don’t blindly throw generative AI at all of this,” he said. “It’s horses for courses. It shines really well when you do metadata enrichment, AI search. But when it comes to the actual automation of a channel, more traditional AI optimization technologies will have generative AI for lunch.”

That framing — generative AI as one instrument among several, chosen on the basis of what the task actually requires — was a departure from the broader NAB Show narrative in which AI is presented as a single, unified capability applicable across all problems equally.

Advertisement

Verbesselt said the decision about which technologies to apply at which point in the content lifecycle is, itself, a strategic responsibility. “It’s really our responsibility to make the decision about what makes sense at which point in time,” he said.