NAB Show Preview: Artificial intelligence moves from hype to measurable impact

The AI conversation arriving at NAB Show sounds different from the one that dominated the exhibit floor a year or two ago. The demos are still there, but the questions have changed. Organizations are no longer asking whether AI works – they are asking where it works, how to manage it across existing infrastructure, and what happens when it goes wrong.

“The conversation around AI shifted from experimentation to operational value and impact. Media organizations are now prioritizing AI that reduces manual effort within existing workflows — tasks such as captioning, metadata extraction, content analysis, and compliance validation rather than flashy generative AI. The real lesson is that AI is only useful when it’s integrated seamlessly into the media supply chain. When it’s embedded directly into ingest, processing, QC, and delivery workflows, AI becomes a force multiplier, delivering measurable efficiency rather than adding another system to manage,” said Benjamin Desbois, chief growth and strategy officer at Telestream.

That framing – AI as embedded infrastructure rather than standalone capability – is expected to define much of what is shown and discussed at NAB Show 2026.

From point solutions to orchestrated workflows

One of the operational challenges emerging from AI adoption in broadcast is that most AI functions are developed and deployed by different vendors. Speech-to-captions, content moderation, metadata extraction, highlights generation – each tends to arrive as a separate tool, creating a new kind of integration burden for operations teams.

“There is a growing appetite from our customers to integrate AI capabilities along their video pipeline when processing live content. Speech to captions, voice translation, inappropriate live content detection, sports highlights extraction and scene-level metadata extraction are typical AI functions customers are testing through proofs of concept. They are improving every month, but they sit in silos because each AI function is developed by a different vendor. Addressing the operational challenge of orchestrating multiple AI functions applied simultaneously on the same linear channel or live events is what matters now,” said Eric Gallier, vice president of video customer solutions at Harmonic.

The architecture question, how to build AI into production infrastructure in a way that is configurable, redundant and secure, is where the practical work is happening.

“The difference between AI hype and real-world use cases comes down to measurable value creation. At NAB Show 2026, we expect more conversations on exactly where AI is most beneficial in a workflow and how it will solve a challenge. No company can claim to solve every problem alone, which is why our architecture is designed to let customers deploy the most effective AI for their own requirements,” said Gwendal Simon, senior director of technology for video networks at Synamedia.

“Real business value is achieved by moving beyond isolated point solutions and instead orchestrating intelligence throughout the entire technology stack. By leveraging AI to analyze vision, audio, and transcripts, media companies generate the rich metadata needed to simultaneously enhance search, discovery, and monetization. This model-agnostic approach ensures that workflows remain flexible, allowing broadcasters to seamlessly adapt to new AI innovations while efficiently distributing content across both traditional and creator ecosystems,” said Juan Martin, chief technology officer at Quickplay.

Advertisement

The metadata foundation

Underlying effective AI deployment is the quality of the data and structure it operates on. Organizations discovering the limits of AI in their workflows often find the problem is not the AI model itself but the environment it is being asked to work in.

“AI is moving from experimentation into everyday operations, but its impact depends on the foundation beneath it. In 2025, Iconik customers ran more than 11 million AI-powered jobs across transcription, visual analysis, and metadata enrichment. The teams seeing meaningful results are not adding AI to disorganized environments. They are building the metadata, governance, and workflow structure that allows AI to perform reliably inside real production and media operations,” said Kathleen Barrett, chief executive officer of Backlight.

“AI is starting to deliver its biggest value behind the scenes in broadcast operations. As media libraries grow, intelligent discovery tools help teams surface relevant footage and repurpose assets faster without relying solely on manual tagging. That kind of contextual search is becoming essential for newsrooms trying to move quickly while getting more value out of their archives,” said Sam Peterson, chief operating officer at Bitcentral.

The realism about pace and readiness is notable.

“AI is cautiously making its way from lab experimentation to media technology business operations, but quite a bit slower than is often touted. Smart automation and optimization are the real low-hanging fruit at this point. There is a ready-to-roll opportunity here to improve both audience engagement and catalogue monetization by effectively closing the loop and connecting content strategy, bundling, curation and scheduling in a self-optimizing way,” said Ivan Verbesselt, chief strategy and marketing officer at Mediagenix.

The infrastructure argument

Beyond workflow integration, one vendor has pointed to a more fundamental architectural question about how AI functions within broadcast infrastructure at the data level — an argument that goes deeper than orchestration.

“The meaningful shift now is toward hierarchical data structure codecs like SMPTE VC-6 with levels-of-quality and region-of-interest capabilities, where AI models subscribe only to the resolution and image regions they need rather than decoding full frames. That turns the retrieve-all, decode-all, discard-most loop into selective, parallel inference — the difference between AI bolted onto broadcast infrastructure and AI as a native operating principle embedded in the data layer itself,” said Fabio Murra, senior vice president of product and marketing at V-Nova.

SMPTE VC-6 is a video compression standard that encodes image data in a hierarchical structure, allowing different levels of detail to be accessed independently. In the context of AI processing, this means a model analyzing video for metadata or object detection can access only the resolution and spatial region it needs rather than decoding and then discarding most of a full frame — a more efficient approach at scale.

Agentic AI as the next phase

The current wave of AI deployment in media is largely task-based – a model performs a specific function, returns a result and passes it along. The emerging conversation is about agentic AI: systems that can observe conditions, make decisions and trigger actions autonomously across a workflow without human instruction at each step.

“Agentic AI is transforming media operations from reactive to proactive by helping to orchestrate entire content workflows autonomously,” said Steph Lone, global leader, solutions architecture, media and entertainment, games and sports, Amazon Web Services. “It’s quickly becoming an operational partner with massive potential. However, for the industry to have full confidence in AI-driven decisions, stronger guardrails around data governance, transparency, content provenance, and responsible automation practices will be essential.”

“AI vision models and emerging agentic tools are capable of turning passive video cameras into intelligent teammates that automate actions and scale multi-camera productions. The emphasis must remain centered on enhancement, not replacement,” said Claudia Barbiero, director of global marketing at PTZOptics.

Advertisement

“AI agents can be used to understand video content at scene level, which opens the door to more precise ad placement and smarter monetization using contextual advertising. AI search agents allow users to search for clips, objects, characters or key moments across an entire content library using only natural language, and AI clipping agents can identify, generate, and compile clips for vertical output and multi-platform distribution,” said Ian Baglow, co-CEO of Bitmovin.

In subscriber management, agentic applications are also beginning to move from concept to deployment.

“One area gaining momentum is agentic revenue orchestration — systems that can independently analyze subscriber behavior and trigger operational actions. This could involve identifying churn risk and automatically presenting a retention offer, adjusting subscription configurations based on engagement patterns, or recommending targeted upsell opportunities,” said Sahil Dhar Hakim, chief business officer at Evergent.

Audio and the broader workflow

AI applications in audio have followed a similar trajectory to the rest of the production workflow — from experimental to practical. Translation, transcription, normalization and metadata generation are the areas seeing the most active deployment.

“From an audio perspective, we’re seeing growing demand from broadcasters for AI tools that analyze content, support metadata generation, assist with translation and transcription, and help identify issues such as dialogue clarity or audio balance. The conversation is shifting from possibility to governance and reliability, with organizations focusing on where AI can meaningfully improve efficiency without disrupting creative control,” said Costa Nikols, executive team strategy advisor for media and entertainment at Telos Alliance.

Sony has positioned AI’s role in production explicitly as a support function rather than a replacement for human decision-making.

“Broadcasters and content creators are integrating AI directly into their operational workflows, treating the technology as a partner that complements and augments human talent instead of as a reactive tool. This is how Sony views AI’s role in the production workflow: freeing creators from burdensome, time-intensive tasks so they can focus on their storytelling, or helping to support camera operation in locations where it’s unsafe or impractical to position humans,” said Kento Sayama, deputy head of the media segment for imaging solutions at Sony Electronics.

Governance, compliance and trust

As AI becomes a standard component of production and distribution workflows rather than an experimental addition, the governance questions that were easy to defer during the proof-of-concept phase are becoming operational realities.

“AI is no longer just a demo — it’s embedded in daily production, newsroom, post, and operations. Yet many teams still operate in what feels like the AI Wild West, experimenting without clear governance, oversight, or regard for emerging regulations like the EU AI Act’s high-risk obligations. With AI tagging, localization, and generative content now part of everyday workflows, ignoring compliance risk isn’t just careless — it could be perilous. The companies that succeed are those that bring structure, accountability, and transparency, turning AI from a risky experiment into a reliable driver of efficiency, creativity, and operational insight,” said Julian Wright, founder of Blue Lucy.

The EU AI Act, which came into force in 2024 and is being phased in through 2027, establishes risk-based obligations for AI systems used in the European Union.

Advertisement

“For the industry to have full confidence in AI-driven decisions, stronger guardrails around data governance, transparency, content provenance, and responsible automation practices will be essential,” said Steph Lone, global leader for solutions architecture in media, entertainment, games and sports at Amazon Web Services.

For media organizations operating across markets, high-risk classifications, which can apply to AI systems used in content moderation, automated decision-making and biometric identification, carry specific requirements around transparency, human oversight and documentation.

The trust question extends beyond regulatory compliance into the content itself. As AI-generated media becomes more prevalent, the ability to verify what is real and what has been synthetically created or altered has become a practical concern for broadcasters and platforms.

“Real-time compliance editing and AI dubbing systems capable of conveying the emotional tone of sports commentary in multiple languages are enabling content owners to scale globally with greater speed. As AI-generated media proliferates, the growing importance of digital watermarking standards to verify authenticity will also make trust and governance key themes at the show. One of the key challenges is ensuring that standards protect broadcasters from manipulated content by addressing emerging threats such as subtle video alterations and audio manipulation that could seriously damage a broadcaster’s reputation,” said Mathieu Boivin, lead architect for cloud native at Viaccess-Orca.

The consent and traceability question is particularly acute in AI dubbing, where synthesized voices raise questions about ownership and attribution that the industry has not yet resolved through standards.

“The industry still lacks clear standards for consent, ownership, or tracing a generated voice back to its source. Every company in this space should be able to say the same because regulators will ask,” said Anton Dvorkovich, chief executive and founder of Dubformer.

AI’s emerging role in detecting synthetic and manipulated content mirrors the broader dynamic playing out across the industry: the same technology creating new production efficiencies is also creating new verification challenges. How broadcasters, platforms and standards bodies respond to both simultaneously is likely to be one of the more consequential conversations at NAB Show 2026.

NAB Show 2026 opens April 18, with exhibits running April 19-22 at the Las Vegas Convention Center. Make sure to check out the latest NAB Show News in our dedicated section or visit the NAB Show website to register for the show.