The 7 Things Every Streaming Exec Should Know About AI
My interview with Metadata Guru Rebecca Avery of Integration Therapy
Rebecca Avery lives where Hollywood storytelling meets streaming operations.
For 18 years, she’s helped platforms like Pluto TV, TelevisaUnivision, Allen Media Group, and NatGeo turn chaos into growth.
She thrives in that awkward middle stage — when streamers outgrow their early hacks but haven’t yet built real enterprise muscle.
And she’s a data nerd in the best way possible.
So when it came to asking how AI is changing streaming, I knew exactly who to call — the Owner and Principal of Integration Therapy and Chair of the Metadata Working Group. (Thx to the great Andy Beach for connecting us!).
10-Second Takeaways
I noticed these seven themes from my Q&A with Rebecca. She and I cover them roughly in this order:
Localization Was the First Big AI Win — Netflix uses AI to translate content with tone and compliance—fueling cheaper, faster global releases.
Metadata Cleanup Is Now Lightning Fast — At Pluto TV, AI replaced months of manual cleanup—speeding delivery across Roku, Samsung, and PlayStation.
Discovery Is Only as Good as Your Metadata — FilmRise stays competitive by using AI to sharpen recommendations and fill platform-specific data gaps.
AI Is Reshaping Rights Management — One client used anonymized viewership data to train AI—without violating any privacy or licensing terms.
Ad Pods Can Be Real-Time Revenue Machines — SVTA’s live project uses AI to match content and ads instantly—boosting CPMs and ad relevance.
Promos Are Already Auto-Cut by AI — One unnamed streamer uses AI to clip shows for TikTok—full audience targeting is next.
Licensing Content to AI Is a New Business Model — Curiosity Stream earns millions licensing content to train LLMs—while laws and risks are still evolving.
The Q&A below is edited for clarity based on three meetings we had recently.
My interview with Metadata Guru Rebecca Avery of Integration Therapy
Q: What was the first big AI win for streamers?
One of the first things that got clear for AI in media is localization.
Getting content into more languages, faster, and at a fraction of the cost.
That was the first major breakthrough.
And now, AI isn’t just translating word for word.
It is doing robust localization services, including contextual translation that makes sure tone, humor, and nuance land OK.
Because what’s OK in North America might not be OK in North Africa.
It can also flag and revise content for local compliance rules like ratings, decency standards, and language restrictions.
For streamers looking for growth, that changes everything.
More content, released more quickly, in more countries, at lower cost. That is unprecedented growth potential while still protecting margins.
Q: How does AI affect search and discovery in streaming?
Search and discovery takes tons of metadata labor to do well.
That kind of shift frees companies to focus on audiences and strategy instead of spreadsheets.
To add complexity, every platform—whether Roku, Samsung, or PlayStation—has different algorithms.
AI is critical for success because it fills in those data points quickly with far less manual labor.
Back at Pluto TV, we would spend months and months on metadata cleanup projects that AI can now do in hours.
That lets streamers double down on strategy instead of spending all their time patching data gaps.
It also gives flexibility. If you want to pivot or update your discovery strategy, you can do it faster and cheaper.
FilmRise cannot outspend Disney, but they can use AI to keep their metadata clean and their recommendations sharp.
And clean metadata isn’t just about speed—it’s also about consistency.
For example, Curiosity Stream might tag a title as both a documentary and a Western, while Cineverse might list it only as a Western (or vice versa).
That kind of variance can impact how content surfaces across platforms, and AI can help normalize those differences.
Looking ahead, generative AI will take even bigger strides.
In the next five years, companies like FilmRise or Cineverse could input their full catalogs, values, and brand identities.
And AI will be able to understand and process all of that to generate much more accurate, personalized metadata at scale.
Q: How does Rights Management play into all this?
Rights management sounds dry, but it is one of the most expensive pain points in streaming.
It is the rulebook for the content license that says when, where, on which business model, and how many times a piece of licensed content can be published on a channel or platform.
The open secret is that many streamers still struggle to track this accurately.
They burn huge amounts of time and money reconciling plays and paying out partners.
AI can automate that. It can read the contracts, map the rights, and reconcile usage with far less overhead.
For example, I’ve worked with clients where anonymized viewership data from one business unit was moved to another corporate entity to train AI tools—without violating data privacy contracts.
It’s the difference between saying “Jackass 2 was the number one title the month Netflix launched in France,” and knowing exactly who watched it.
The data stays anonymized, which makes it viable for training without breaching rights.
And once you can track rights more effectively, you unlock more than efficiency. You unlock creative freedom. You can design new kinds of deals, explore new windows, and think about how to build value with content rather than being limited by the system that chokes the entire finance and legal departments every quarter.
There’s even a working group—the Content Authenticity Initiative—developing technical standards for rights and AI-generated content, like embedding provenance into metadata.
That’s going to matter more as AI becomes a bigger part of production and distribution.
That is huge—both for today’s operating margins and for the new opportunities you can build in the future.
Q: What’s AI’s impact on Ad Inventory?
Ad inventory is still under-optimized, and this is another race—not just to beat competitors, but to clear the bar on profitability.
We’ve all seen those clunky ad pods: “Be Right Back” slates, or the same promo repeated three times.
That is wasted space.
With AI, you can combine content metadata, user data, and ad data in real time to serve the right ad to the right person at the right moment.
Ideally, it’s a fast, automated exchange: take the last 15 seconds of content that aired, pair it with the metadata about what that content was, and push that into the ad marketplace to pull in the most relevant ad for that moment, for that viewer, on that platform.
That does two things: it increases CPMs, and it builds advertiser trust.
If you can prove placement is precise and effective, advertisers will pay more.
That’s exactly what we’re working on at SVTA right now. One of our live projects is focused on leveraging content metadata for both brand safety and increased ad value—and I believe that’s now public.
That said, there is still a lot of dust to settle around ad inventory—moving from broadcast to digital, and between social and streaming.
There are a lot of levers that streaming needs to pull to bring in more advertisers, but AI and metadata will help settle the dust in the right direction.
Q: What about the impact of AI on creating content?
There are two buckets: promos and actual content.
On promos, it is already happening.
Some streamers are scanning long-form content, cutting clips automatically, and pushing them straight to TikTok.
One streamer is already doing this today, though when asked if they could segment audiences—for example, a version for women 30–45 and another for men 18–24—they said not yet, but we’ll be there in a couple of years or less.
On content creation, the frontier is still taking shape. Studios are using AI for animation and VFX.
Disney, for example, has research teams speeding up rotoscoping and background duplication.
And at the same time, we have all seen the rise of AI-generated personalities.
Tilly Norwood, for instance, is everywhere right now. And whether you love her or hate her, she is a preview of what is coming.
Between legal developments and tech developments, nobody really knows what content will look like in just a couple of years.
What we do know is that technology is moving so fast that the law will always be scrambling to catch up. That is the new normal.
Q: What do you hear about streamers licensing the content they own out to AI? Curiosity Stream generates millions of dollars doing this.
This is the new secondary rights market.
Many major streamers already have deals with Google, AWS, and others to let their content train AI models.
Personally, I’m excited that big tech has so much content to work with, because it’s vital for AI/content innovation.
I know a lot of people are feeling cautious in this area.
The laws are still evolving. California has passed new regulations around data use, and Europe is drafting others.
We are in the early stages of figuring out what is protected, what is not, and how compensation should work.
But the upside is real. Licensing training rights creates a new revenue stream.
The risk is also real.
If contracts are not explicit, you may end up training a competitor’s AI without realizing it.
Q: You’ve told me that the #1 blog article you’ve written is “You’re Not Netflix, and That’s a Good Thing.” Why do you think that landed?
I’ve worked in streaming for 15 years, and I’ve heard dozens of problem-solving sessions devolve into, “What does Netflix do?”
And my answer is: it does not matter. You are not Netflix.
Netflix had a very specific path.
They went from DVDs by mail to becoming one of the first major adopters of AWS, and their early demands helped shape what cloud infrastructure became.
They embraced an identity as a technology and data company before most media companies even realized they needed to.
They leaned hard into product innovation and had the right timing to scale during the streaming boom.
That’s kind of the point: you don’t have the same history Netflix had. You’re different.
So think about your own values. Go to your own mission.
That is not a playbook anyone else can just pick up and copy. What worked for Netflix worked because of who they were and when they did it.
Personally, I have deep respect for Apple TV as well.
I’d still give Netflix the crown because they’re such a big aggregator, but the point remains: other players bring different strengths.
That message resonated because it gave executives permission to stop chasing an impossible benchmark and lean into their own uniqueness.
Don’t be a worse version of someone else—be the best you that you can be.
Q: Why are Netflix and Apple TV number one and two at data among the streamers?
Netflix understood early that data was gold.
They organized it, tested against it, and turned it into predictions.
They have been doing that for 20 years. Every thumbnail, every category, every A/B test compounds into insight.
Apple, on the other hand, comes from an experimental culture.
Steve Jobs did not just ask what customers wanted.
He shipped what he believed they would use.
That mindset carried through. Apple has the cleanest device-level identity graph in the business.
They know you seamlessly across iPhone, iPad, and Apple TV. That clarity makes their data razor-sharp. Traditional broadcasters and studios started from a very different place.
Hollywood is the master of storytelling. It has shaped culture for a century.
But legacy is a double-edged sword.
Having such a broad reach, such a deep history, and such a huge cultural impact, and then being asked to reinvent themselves as data-driven companies in just a handful of years, has frankly been an unfair disadvantage.
That is something I empathize with deeply.
As much as I work in technology now, I am a product of Hollywood jobs.
My heart is in Hollywood. I know how much effort goes into creating culture through content, and how difficult it is to shift that weight of history overnight.
That said, the industry is learning.
You can see it in how much more traditional Hollywood content shows up on streaming platforms now, and in the flexibility of the business models being tested.
Netflix has the benefit of time and scale. Apple has precision and culture.
And Hollywood still has the crown in storytelling, while working through the challenge of merging that legacy with a data-driven future.
If you’d like to contact Rebecca, you can reach out to her at Integration Therapy.
Final Takeaways
For Media Executives:
• Metadata is now a growth lever—clean it fast, scale discovery, and boost platform visibility.
• Rights and ad ops can be automated—AI cuts costs and creates new revenue paths.
• AI training rights = real money—Curiosity Stream proves it, but contracts must protect your IP.
For AI Executives:
• Streamers need metadata help—tools that clean, tag, and personalize will win fast.
• Training deals are heating up—offer transparency and control if you want premium content.
Thanks for reading!
Rob Kelly, Creator & Host of Media & the Machine
p.s. How to reach me?
If you want to reach me, the best way is to subscribe below and reply to any of my emails. There’s a free option — and your reply comes directly to my inbox.