NAB 2026: AMC Global Media's AWS-Driven AI Journey
My visit to NAB 2026 skewed heavily toward talking about workflow changes with AI. I tried to keep my travels on the convention floor to very specific examples of AI’s impact on streaming workflows, and this article will explore the most interesting ones I found. First up is how AMC Global Media (newly rebranded from AMC Networks) is working with AWS to leverage AI to create better content access.
AI Content Management
On the second day of the show, AMC VP of Architecture, Strategy and Emerging Technologies Sam Phillips spoke on a panel in the AWS Theater titled PRISM—AMC Networks Success Story: Adopting GenAI to Streamline Workflow. Phillips heads up the AI and emerging tech program at AMC Global Media, and he and his team worked with AWS to design PRISM as essentially a MAM overlay for testing the waters on how generative AI could help them manage content assets and also provide access to a wider base within their company who need access to these assets.
Prior to developing PRISM, Phillips explained, AMC Global Media faced two major challenges. “Our inventory is spread out all around the world. It's in all different storage locations. It's in inconsistent formats. When you look at the sheer scope of that and you think, ‘OK, if I want to deploy a solution that leverages AI to discover across that entire library, suddenly that becomes very overwhelming,’” Phillips said.
Thus arose the million-dollar question: "Do I need to consolidate into a single MAM? Do I need to consolidate to a single storage environment? That's often a non-starter for most companies," Phillips said. The goal was to deploy an architecture that could be sufficiently abstracted, without waiting for other parts of the supply chain to get organized.
A typical build vs. buy dilemma followed. The idea of spending three to six months evaluating vendors and technologies, choosing one, and then being faced with a good probability of those models and methods becoming out of date quickly emerged as a major concern. "That's really not a place we can afford to be," said Phillips. "The other aspect of that is we're also downstream of their product decisions, which means that if they go in a direction that is different to where we want to go within our business, then we're going to be stuck."
The MAM Overlay
AMC worked with AWS to develop what they characterized in the session description as “AI functionality to HD/4K video content with multi-track audio support, automated metadata generation, intelligent search, secure content management, and performance metrics.”
PRISM incorporates the following AWS components:
- AWS Media Lake
- TwelveLabs Marengo Video Analysis model
- Amazon Bedrock
"We got inspired by this media lake concept, the idea of moving up in the architecture," said Phillips. "I also think it’s why we were able to deploy into production so fast, because we were able to move without needing to worry about what is my current state."
Panelist Paul Dulac, Senior Media Application Architect, Media and Entertainment at AWS explained that having Amazon S3 services as the single “source of truth for those assets across systems” meant that AMC was “able to immediately provide references or pointers from the AMC PRISM system to their downstream systems and the editors can immediately be working with the new assets. It's really about being able to quickly get a catalog on board and maybe interact with an existing MAM system, but leverage models that are not necessarily tied to your current provider of choice, while also being able to integrate in ways with your downstream systems."
The aspect of PRISM that makes it an effective tool for AMC team members who need to access media assets across those systems is agentic AI search.
“When you think about the scale of a media inventory for a global media company,” Phillips said, “you’re going to get a lot of noise in those results too, even if they're brilliant results. I think if you’re looking at this from maybe a contextual advertising perspective, product placement is a huge piece that we're very, very excited about.”
The workflow begins with source assets like video, audio, or images being dropped into the S3 pipeline. "We put a video asset into a [Twelve Labs] Marengo 3 [embed] model and we get back visual embeddings which encompass everything that you're going to see on screen, who's doing it, and anything in the image. The audio tracks are split into two different modalities: everything non-dialogue and dialogue,” said Dulac. ‘I think there's support for 36 languages, and you can search across all of them simultaneously.”
They can combine modalities in search and weight them, depending on the semantics of what's in that search. They can choose to weight visuals or audio more heavily weight. "The other thing that the model gives us is the ability to take that semantic search and pair it with an image,” Dulac added. “This is incredibly powerful for us because you can put anything in that: logos, products, specific vehicles, celebrity faces, or the faces of people who might not be celebrities yet and aren't in your dataset for a general recognition model.”
"From a metadata perspective, we hooked up MediaLake by API to interface with our metadata management system,” Phillips said. “Every time media hits MediaLake, it hits a rest API call and pulls a very shallow payload back from the metadata system and that serves as our title metadata for within the PRISM platform.”
They didn't want complex hierarchies because it's not a MAM; it's designed for a new search experience. They also put business rules in place on where content was or what type of content needs to be processed where to minimize egress costs.
The first place they decided to use this was social. "Our social media team produces tens of thousands of assets a year,” Phillips said. “Each of those assets takes a long time to find what you're looking for. You need to know the specific ID, the storage, potentially the time within the asset where you're going to find that moment that you want to put on TikTok.”
He went on to explain that "the key thing is being able to then export that collection out to that downstream system. We already have integrations with one system currently for AMC. We're looking at adding more S3 services as a source of truth for those assets across systems, [which means we can] immediately provide references or pointers from the AMC PRISM system to their downstream systems and the editors can immediately be working with the new assets."
The result? "It reduces our time to find those moments and time to assemble those moments into a deliverable that makes a good viewing on a social platform," Phillips said.
AMC expects to be able to support different models as needed. The name PRISM comes from the solution’s ability to look at media from different perspectives. For more details search AWS Labs GitHub page for a white label version of the UI.
Data Intelligence
"I think the way that AMC operates is somewhat different than some of the other media companies. We try to get our content to everywhere there are viewers are," said Robyn Goldman, VP of Business Solutions, AMC Global Media, on the panel Making Data Pay - How Data is Becoming the Operating System for Media. "I have a team of business analysts. We're responsible for relationships with the vendors that our business stakeholders utilize. We service our business stakeholders to provide solutions for their problems across the media supply and ad tech chain."
Goldman explained that in earlier times, the most effective way to provide these types of data intelligence solutions and services was to build a dashboard or generate a report. "I would say about 10% of that report would contain the data that you actually need. You’d have to run this large report just to get to five rows of data that you might need.”
Today, she said she’s working with an internal team at AMC led by Sam Phillips to pursue a new and more streamlined approach “that's focused on data, AI, and emerging technology."
They are looking at understanding data, putting it into a place where it can be understood, normalized, and gathered across silos and then extracted. The goal, she said, is to be able to “address questions as opposed to just running large sets of data when you don't really know what the actual answer to the question might be.”
With this legacy report-running approach, she explained, “If another question is posed to you, now you have to either run another report or you have to go to your technology team and have them build you another report, which could take weeks. It could take months.”
Leveraging AI in their new strategy, by contrast, means “having immediate access to the data. Whether it's a C-level person that's asking you what content do we have available to air in a particular territory for a particular amount of time, or it's our content licensing team that's looking at how they can leverage and monetize inventory that's been sitting around dead for months, for years [having access to this data] bring that to life, now you're monetizing your inventory in ways that you haven't even thought of."
But she cautioned that "you still always have to have that single source of truth. It's always accessing that trusted source of data that has its source of truth."
The Distribution Path
By placing less emphasis on owned-and-operated environments, Goldman explained that AMC instead works with multiple distribution partners, which brings the challenges with sending content out into a multi-platform world of varying specs and requirements, "We are very partner-focused,” she said. “Our channels are on any platform that we can get them on. The management of those schedules is extremely time consuming."
Turning her attention to spinning up FAST linear channels, she said, "FAST channels are procedural [to build]. So you've got all of your episodes of The Walking Dead, you've got all of your episodes of Portlandia, and it's rinse and repeat. Or you have your category specific-type FAST channels. It would make things so much easier if we can create content collections and put the content in a pool, understanding what content fits best," she said.
AI deployments in AMC’s workflows have moved well beyond isolated applications or experimentation, she said. "We're heavily investing in AI across the supply chain. "One of the really cool things that we're doing right now is partnering with AWS to search our content so our social media team could create social posts off of anything without having to spend hours and hours and hours pouring through the long form source material to make these assets."
Looking ahead to future applications, she said, “it could be used for contextual advertising, it could be used to make clips for digital out-of-home partners. You could use it for a lot of different things and also just to search our content for anything that we may want to leverage our assets for."
Of course, first they have to find whatever it is they are looking for. Hopefully AMC Global Media’s AI journey scales well.
Related Articles
As we head into another NAB, I'm less interested in what looks impressive in a booth and more interested in what is actually ready to survive in production.
22 Apr 2026
Perhaps the booths that cropped up on my NAB 2026 itinerary and the sessions I squeezed in between simply skewed this way, but it seems like everyone's favorite sport in streaming this year is, well, sports. Certainly, when it comes to live, the pinnacle of achievement appears to be seamless delivery of premium sports events at global scale, with an eye to "meeting fans exactly where they are"—crossing national, lingual, and cultural borders via localization, and transcending generations through autogenerated highlights and verticalized mobile delivery. And when it comes to traditional landscape sports video streamed to conventional CTV glass, multiview more than ever seems to be the name of the game in 2026.
21 Apr 2026
In this Q&A, Samira Panah Bakhtiar, Amazon Web Services (AWS) general manager of media, entertainment, games, and sports, talks about how AWS and agentic AI are changing the ways streaming workflows are executed. In her interview, she spoke both to the business side of the house and the underlying tech.
14 Oct 2025