Industry Insights: How AI is finding a place in everyday media workflows

Artificial intelligence is moving beyond experimentation and into the daily mechanics of broadcast and media operations.

In this first installment of our three-part Industry Insights roundtable, vendors examine where AI is now delivering practical value across production, post, editorial and distribution workflows.

The discussion focuses on the use cases that are proving durable in real environments, from transcription, localization and metadata enrichment to video indexing, archive search, clipping and encoding optimization. It also looks at where teams are seeing measurable gains in speed and operational efficiency, and how organizations are evaluating whether those gains hold up at scale.

Rather than revisiting AI as a theoretical promise, this conversation centers on where it is becoming embedded in systems, workflows and daily decision-making across the media supply chain.

Key takeaways from this Industry Insights roundtable

• AI in daily workflows: Many broadcast organizations are moving AI from pilot programs into operational workflows, particularly for transcription, captioning, localization and metadata enrichment.
• Automation of repetitive tasks: AI is delivering its most consistent value by reducing manual logging, tagging, search and review tasks across production, post-production and archive workflows.
• Integration drives adoption: Deployments tend to succeed when AI capabilities are embedded directly within MAM, newsroom and orchestration systems rather than introduced as standalone tools.
• Operational metrics matter: Teams are measuring AI impact through practical benchmarks such as time-to-content, search success rates, compute savings and reductions in manual processing steps.
• Rapid time to value: Contributors suggest that clearly defined AI deployments can begin delivering measurable workflow improvements within weeks or months when data and infrastructure are well organized.

Where has AI moved from pilot to production inside broadcast and media workflows? 

Yang Cai, CEO and president, VisualOn: AI has moved into production most visibly in video processing workflows, including content-aware encoding optimization, quality assessment, and automated metadata generation at scale. These systems are now embedded in real-world VOD and live workflows where consistency and throughput matter more than experimentation.

Michael Chan, VP, delivery operations in corporate, Accedo: Media companies are using AI for metadata and asset management, for generation of time coded transcripts, object detection, identification of key scenes within video, and for tagging content at scale. These are all labor-intensive manual processes, but with AI, broadcasters and media companies can process these tasks at speed. This represents a fundamental shift in how media, editorial and newsrooms operate and function.

Josh Rosen, VP, production and content, North America, Gravity Media Group: AI has most clearly moved from pilot to production in transcription, captioning and localization, automated logging, and metadata extraction that improves search and retrieval. It’s also becoming real in areas like rotoscoping and masking assistance, along with other “first-pass” organizational tasks that were previously manual. These deployments work because they reduce friction without asking AI to make creative decisions.

Clara Aler, head of marketing, Knox Media Hub: What I am seeing as fully in production and pretty standard industry-wide are localization workflows and QC (language detection, speech-to-text, translation, subtitle generation, automated QC flags, etc.). There’s a clear ROI: faster localization workflows, easier to deliver content globally, and the human review process is infinitely faster compared to entirely manual translation. AI has a long and successful history in this domain, making it a a reliable area for good results.

Advertisement

Scott Goldman, general manager, U.S., Qibb: AI is now embedded in day-to-day broadcast operations, particularly for automated metadata enrichment, content discovery, and real‑time clipping. Qibb customers are already running these capabilities directly in live production and editorial workflows, not as isolated pilots. The shift has been driven by orchestration platforms that make AI outputs predictable, repeatable, and operationally reliable.

Jacob Arends, senior product manager, playback and AI, Bitmovin: AI is being used to analyze video content to detect things like scene transitions, objects, characters, location and mood, in order to label scenes with keywords and generate metadata that is immediately usable for personalization, targeted advertising, and discovery. AI use in observability is another area that is developing rapidly, with video service providers using AI driven tools to identify and resolve quality, monetization, and performance issues in real-time. And with MCP (model context protocol) servers, users can interrogate data and query complex metrics in everyday language.

Ken Kobayashi, business manager, remote cameras, Sony Electronics: In broadcast and media workflows, secondary cameras using AI tracking features are creating additional footage and providing more options. This setup is being used to capture corporate events, where the main camera with a manned operator widely frames people on the stage, and, at the same time, a camera with AI tracking can be used to specifically track the presenter to keep them in the IMAG PinP window. Together, this creates a more engaging, dynamic and relevant presentation.

Charlie Dunn, executive vice president, products, Telestream: Today, we are selling, and customers are deploying the use of AI in two beachhead locations. Within Vantage workflows, we are generating captioning and/or subtitling using AI, and if needed, we can create translations into 120 different languages of those captions and subtitles into an IMF or similar package. Compared to sending content out for captioning or translation, the savings are immense. The second location is to create time-based referenced metadata as well as a summary for incoming news feeds.

Derek Barrilleaux, CEO, Projective: Major vendors have done a good job of adding AI tools that individual users can leverage — Adobe’s tooling is a great example, and many people are using LLMs for shortcuts. What’s lagging is a systematic approach at the organizational level. For instance, individual tools offer transcription capabilities that users can trigger independently, but at an organizational level, do you really want all your users initiating transcription individually? A more systematic, organizational approach requires change and thoughtful execution, and that’s where things seem to be breaking down.

Sam Peterson, COO, Bitcentral: AI has moved into full production in areas where reliability and scale matter most, particularly transcription, metadata enrichment, and content segmentation on live and recorded feeds. These capabilities now operate alongside and directly connected to MAM and newsroom systems, reducing search time and making archived and breaking content immediately usable. We’re also seeing AI reliably support automated clip creation and story segmentation as part of existing editorial workflows.

Rich Welsh, president, SMPTE: One area that has made very solid use of AI is 3D modelling in VFX. Machine learning has been used in solving 3D scans into models for many years, and more recently the emergence of Gaussian Splat has significantly changed 3D workflows. With GSplat models improving significantly as AI evolves and GPU speed improves, we’re now seeing simplification of capture and flexibility in use.

Danny Hollingsworth, director, post production product marketing, Avid: Speech to text and dialog indexing have moved into everyday production use, particularly for search, logging, and editorial decision making. Tools are enhancing collaboration and driving efficiency across media and broadcast workflows by transcribing and translating clips and interviews within seconds, identifying and renaming speakers and search within the text, rapid dialogue search and synchronization. Tools such as AI-powered analytics that provide facial detection and automated transcription, are further moving AI from pilot to production.

Steph Lone, global leader, solutions architecture, media and entertainment, games, and sports, Amazon Web Services: Agentic AI is fundamentally changing the way content moves across the media supply chain, from creation to audience delivery. It’s orchestrating entire content workflows autonomously, handling tasks like video understanding, metadata extraction, rights management, distribution and more, without constant human intervention. These systems continuously monitor operations, identify bottlenecks, surface archived assets when relevant, verify metadata accuracy, and handle rights checks, effectively making AI an operational partner.

Phil Petitpont, co-founder and CEO, Moments Lab: AI has moved from pilot projects into everyday newsroom workflows, where it now helps with footage review and archive search. It’s being used both upstream and downstream in production: upstream to build new shows from unscripted footage and unlock more value from archives, and downstream to recycle content for new audiences by quickly finding key moments for social clips and short-form video.

Zoe Liu, CEO and founder, Visionular: At Visionular, AI has moved firmly into production in three core areas: AI-driven encoding optimization (including automatic region-of-interest detection, adaptive bitrate allocation, compute-efficient RDO-aware block partition decision), video enhancement (including super-resolution, frame rate up conversion, adaptive tone-mapping), and no-reference visual quality assessment. These are no longer experimental capabilities. They are embedded directly into encoding pipelines and operate at scale across live, VOD and real-time communication workflows.

Miguel Coutinho, head, NDI: As of today, AI is predominantly being used as an embedded capability in both software hardware products in broadcast and media workflows. For example, AI models are used for auto-metadata tagging, camera tracking, automatic source switching, picture correction, or for placing graphics in a live recording.

Advertisement

Geoff Stedman, chief marketing officer, SDVI: We’ve seen AI tools used in production to analyze content, identify potential problem spots, and direct an operator to precisely the point in the content that needs manual verification. The other area where AI is now consistently adding value is localization, particularly for captioning and subtitling. If captions for a particular language already exist, creating a translated version for operators to validate provides huge efficiency gains.

Simon Wocka, regional sales manager, Pixitmedia: AI is now genuinely being used as part of daily operations for content indexing, automated clipping, and metadata enrichment. We’re seeing this built directly into content management and storage systems rather than sitting off to the side as separate tools.

Which use cases have delivered clear time or cost savings for teams on a daily basis?

Yang Cai, CEO and president, VisualOn: Automated content analysis for encoding decisions has reduced manual tuning and repeat processing, cutting both compute cost and turnaround time. AI-driven quality evaluation also minimizes subjective review cycles while maintaining predictable output quality.

Michael Chan, VP, delivery operations in corporate, Accedo: AI processing of tasks that used to be manual such as identifying key scenes, tagging, and transcription, is saving editorial and newsroom teams time and money every day. Then you have the use of AI-research and analytical tools, as well as real time editorial triggers with AI assisted outputs for summary, live blogging features, and live transcription. All of this is saving teams a huge amount of time and effort. Graphic production is another example.

Josh Rosen, VP, production and content, North America, Gravity Media Group: The most consistent daily gains come from finding the right content faster — through searchable transcripts, speaker or scene detection, and structured metadata — and from reducing repetitive post-production tasks like masking, cleanup, and versioning. Savings are strongest when AI output feeds directly into existing production and post pipelines rather than living in standalone tools. In practice, the biggest impact is usually increased capacity and throughput, not automatic quality improvements.

Santiago Miralles, founder and CEO, Knox Media Hub: By implementing cloud-native MAM systems with embedded AI and end-to-end workflow automation, organizations are increasingly able to insource operations that were previously outsourced, while significantly reducing manual effort in areas such as metadata management and pre-delivery compliance. In some cases, the savings can reach an order of magnitude, dividing by 10 the previous cost. Beyond cost savings, these systems give teams the flexibility to launch new products, services, and digital business models without adding meaningful operational overhead.

Scott Goldman, general manager, U.S., Qibb: Automated tagging and metadata enrichment have significantly reduced manual logging and review. Teams spend less time searching for footage and more time reusing archived content for new programming and digital distribution. These gains show up daily in faster turnaround times and reduced production overhead.

Charlie Dunn, executive vice president, products, Telestream: It’s not an unusual or remarkable thing to convert an audio track to text, but there are special considerations when converting audio to captions or subtitles that take some unique knowledge. If you compare this to what you had to do 10 years ago, that meant hiring a caption/subtitle service to manually create captions and subtitles which was a significant cost for an organization not to mention the time it took to send content out and then match and time it with the original. Then if you send the content out for international distribution, you may have to create 10-20 different language translations and again, you had to send out for this work to be done and then match it with the content on the way back.

Derek Barrilleaux, CEO, Projective: Our approach is to pick concrete use cases for our customers and provide value with those. For example, improving search and driving more impact from their existing content and the framework that Strawberry provides.

Sam Peterson, COO, Bitcentral: The biggest gains come from automating high-volume tasks like transcription, metadata generation, and archive search. Tools like Bitcentral’s Fusion Insights can surface relevant clips across large content libraries in seconds, significantly reducing prep time for producers. Automated highlights and story segmentation have also shortened turnaround times during breaking news and live events, delivering measurable efficiency.

Advertisement

Danny Hollingsworth, director, post production product marketing, Avid: Built-in tools for transcription, dialog search, and automated assistance with repetitive editorial tasks reduce time spent on manual logging and review. Editors can locate specific moments or lines instantly and move more quickly from raw footage to usable sequences. These gains compound across large projects, especially in unscripted, news, and factual environments where volume is high.

Steph Lone, global leader, solutions architecture, media and entertainment, games, and sports, Amazon Web Services: Video understanding and the metadata it extracts are becoming increasingly prevalent. It’s an area where AI is being used to automate processes that traditionally required hours of manual review, like celebrity detection, quality control (QC), compliance verification and ad break detection. AI can process content in minutes, allowing organizations to refocus talent on more creative endeavors. Localization and dubbing are other areas where AI is accelerating workflows and saving teams money.

Phil Petitpont, co-founder and CEO, Moments Lab: AI-powered search and discovery is changing the way many companies approach their video workflows. For example, Global production company Banijay has seen an 80% reduction in the time it takes to locate the right clips. This efficiency has lowered production costs from about $800 per clip to just $10 per clip and has put them on a path to “100X”-ing their social media output, driving more revenue from platforms like YouTube.

Zoe Liu, CEO and founder, Visionular: The most measurable daily impact comes from bandwidth reduction without perceptual quality loss, reduced storage requirements, lower CDN delivery costs, and faster encoding throughput through AI-assisted optimization. In high-volume streaming environments, a 20-40% bitrate reduction translates into immediate, recurring cost savings. That is not theoretical ROI — it shows up directly in infrastructure budgets.

Miguel Coutinho, head, NDI: In controlled spaces like studios or conference rooms, repetitive camera adjustments can largely take care of themselves. Operators don’t need to constantly correct framing when a presenter shifts position or walks across the stage. Over a year, that adds up to a real reduction in hands-on work and mental load.

Geoff Stedman, chief marketing officer, SDVI: New AI-powered tools are analyzing content archives and extracting far more metadata than had been possible before. This metadata enrichment enables more efficient media workflows where content needs to be modified from its original form to meet requirements for new distribution platforms such as AVOD or geographic destinations with different content restrictions. AI tools can generate time-based metadata to guide final manual operations in use cases such as locating optimal ad insertion points, identifying content moderation items for regional requirements, and enhancing user viewing experiences with intro, recap, and credits detection.

Simon Wocka, regional sales manager, Pixitmedia: Automated cataloguing and search are making a massive difference along with automated data movement between storage tiers, sites, and cloud. When you can orchestrate that intelligently rather than having people manage it manually, you free up teams to focus on actual content creation. Productions are being tasked with quicker turnarounds and smaller teams, so using AI for the right purposes provides real benefit.

What measurable benchmarks are organizations using to evaluate AI’s impact on workflows?

Yang Cai, CEO and president, VisualOn: Organizations are measuring AI impact through concrete metrics such as processing time reduction, compute cost savings, and output consistency at scale. In video workflows, benchmarks often include bitrate efficiency versus perceived quality, error rates, and reductions in manual intervention. Long-term success is evaluated by workflow stability and the ability to scale AI-driven decisions without increasing operational complexity.

Aitor Falcó, sales manager, Knox Media Hub: Typical benchmarks include ingest-to-publish time, reduction in manual processing steps, faster content discovery, improved reuse rates, fewer compliance issues, fewer systems, more autonomy in building workflows, and higher throughput per operator. These metrics are tracked end-to-end across the media supply chain.

Scott Goldman, general manager, U.S., Qibb: Organizations commonly measure AI impact using KPIs such as time‑to‑content, search success rates, archive reuse volume, and reductions in manual logging. These benchmarks directly reflect how AI improves speed, efficiency, and content utilization across teams. Over time, they help quantify ROI and build the case for scaling AI deeper into production operations.

Rich Welsh, president, SMPTE: I don’t think business KPIs change when it comes to AI overall. Ultimately, AI will either improve efficiency or enable new revenue streams, or it won’t. The measure of its impact on a granular level is no different than a new hire or system investment decision.

Phil Petitpont, co-founder and CEO, Moments Lab: Organizations tend to measure AI’s impact by looking at how much easier and faster everyday work becomes. They track things like whether AI reduces time spent on repetitive tasks such as media logging and scheduling, how much better content discovery and search relevance are, and how intuitive AI search feels compared to traditional MAM systems. Another common benchmark is simple time savings — how quickly AI can generate video titles, descriptions, or hashtags versus doing the same work manually.

Zoe Liu, CEO and founder, Visionular: In encoding and delivery environments, benchmarks are concrete: Bitrate reduction percentages, full visual quality metrics such as VMAF, and proprietary no-reference perceptual quality metrics that we have developed in house and that have been extensively validated by customers operating large-scale video platforms. Additional indicators include CDN cost savings, encoding time improvements, compute resource reduction measurement, and storage footprint reduction. These are quantifiable and repeatable.

How are cost savings, efficiency gains or quality improvements being documented internally?

Russell Trafford-Jones, industry engagement manager, Techex: No broadcaster is the same as any other, but key metrics for live broadcasters are a mix of the traditional such as “seconds of black to air’”mixed in with more modern metrics like number of impacting routing errors. By monitoring accuracy, broadcasters are hoping to demonstrate increased efficiency as they bring AI into managing workflows.

How long does it realistically take for AI deployments to show operational value?

Michael Chan, VP, delivery operations in corporate, Accedo: If done right, organizations can expect to see operational value very quickly, say in less than six months. The initial operational benefit comes from efficiency gains because there’s a reduction in the roles that require manual effort with minimal variation in the way tasks are completed. Having said that, big technology companies are probably outliers, in that as technology enablers themselves, they’re able to extract value from AI much faster than most media companies are able to.

Phil Petitpont, co-founder and CEO, Moments Lab: The speed and success of an AI deployment is tied to the data a system has available to it and how organized their existing content is. For AI search and discovery, the process can be completed in a few short weeks — a bit longer if an organization has decades’ worth of footage to integrate. Once that process is completed, however, the value can be seen almost instantaneously.

Where has AI failed to deliver expected ROI despite technical success?

Aitor Falcó, sales manager, Knox Media Hub: ROI typically fails when AI is deployed as isolated point solutions without integration into the MAM and orchestration layer. Even accurate AI outputs deliver limited value if they cannot drive automated, governed actions across the media supply chain.