06: AI Solutions
Welcome back to Supercharged: Insights from Keywords Studios, our quarterly newsletter.
This edition explores the ways game studios are uniquely positioned to lead in the AI era by leveraging the tools and personnel already at their disposal. By recognizing that the foundational skills used in game development are the same ones required for high-quality AI operations, studios can stop chasing "wow" moments and start engineering reliable, valuable workflows and better AI systems.
The Hidden AI Capabilities Every Studio Already Owns
By Quentin Staes-Polet, General Manager - Keywords Studios AI Solutions
Much of the games industry's AI conversation has focused on risk.
Will AI change how we develop games? Will it impact jobs? Will it alter production pipelines?
These are important questions, but I believe they risk obscuring a much bigger opportunity.
Over the past year, I've spent time working with some of the world's leading AI organisations. What surprised me most wasn't how different their challenges were from ours. It was how familiar they felt.
The capabilities required to build, test, evaluate and improve advanced AI systems increasingly resemble the capabilities the games industry has spent decades developing and applying to entertainment experiences, and are quickly becoming some of the most valuable inputs for AI.
What Matters
As AI models become more capable, the race is shifting.
The limiting factor is no longer simply access to compute or model architecture. Increasingly, it's access to high-quality human data, human judgement and robust evaluation processes.
AI companies need specialists who can identify errors, assess quality, evaluate outputs, test edge cases and improve performance over time.
These aren't abstract requirements.
They're operational capabilities.
And they're capabilities many games organisations already possess.
AI in Practice
There are two assets that I believe many studios continue to undervalue.
The first is their people - Artists, QA specialists, localisation experts, community teams and live-service operators all bring forms of judgement that AI systems increasingly depend upon. Many of the roles currently supporting game production have direct applications in AI evaluation, safety, red-teaming and model improvement.
The second is production discipline - Building and operating successful games requires rigorous processes, quality control, global coordination and the ability to deliver under pressure. These same disciplines are becoming essential for organisations looking to build reliable AI systems.
During my time leading AI-powered user-generated content initiatives, one lesson became clear very quickly: the challenge was rarely the model itself. The challenge was creating the human feedback loops needed to continuously improve outcomes.
That remains true today.
Inside Keywords Studios
We're already seeing this transformation take place.
Over the last year, Keywords Studios has supported one of the world's leading AI organisations through specialist teams focused on data creation, evaluation and quality assurance. What began as a relatively small engagement rapidly expanded into a large-scale operation involving hundreds of specialists supporting millions of interactions and helping improve advanced AI systems.
The project reinforced something we've known for years: exceptional outcomes depend on talented people, operational excellence and relentless attention to quality.
The same foundations that underpin successful game development are proving equally valuable in AI.
Why Localisation Matters More Than Ever
One of the most overlooked challenges in AI is multilingual quality.
Building models that work across languages isn't simply a translation exercise. It requires cultural context, linguistic nuance and human judgement.
The games industry has spent decades mastering those skills, so as AI becomes increasingly global, localisation expertise may become one of the most valuable capabilities in the entire AI ecosystem.
For years, studios have viewed many of these capabilities as essential production functions.
The organisations that thrive in the next wave of AI may be those that recognise these same capabilities are becoming strategic assets.
The question isn't whether AI will change our industry. It's whether we fully recognise the value of what we've already built.
Get The Latest From Our Experts
Click below to read more about Quentin's 25+ years of experience in the technology, media and AI space, or get in touch with him via LinkedIn
Quentin Staes-Polet
The Great Data Inversion
By JC Langley, Head of Operations and Strategic Partnerships: Keywords Studios AI Solutions
Why entertainment's most distressed services companies are its most valuable AI assets.
CuriosityStream was losing money. Streaming company, documentaries, not exactly Netflix. Then someone realized the library, 210,000 hours of it, was worth more as AI training data than as a streaming service. Net loss to $8.2M EBITDA. The content didn't change. The buyer for it did.
The AI training data market is $2.7 billion today and heading for $11 billion by 2030. And the companies sitting on the most valuable datasets in it have no idea they have them. They think they're in the translation business.
CuriosityStream was content as data. A library someone could point to and say “license this.” What comes next is stranger, because the most valuable datasets in entertainment aren't libraries at all. They're operational exhaust.
Reddit locked in $60 million a year from Google. Then renegotiated because the first deal was too cheap. Total AI licensing: roughly $130 million annually. Ten percent of the company's revenue. For text that users typed for free.
Shutterstock made $104 million licensing images to AI companies in 2023. By 2025 that was $203 million. Twenty-one percent of total revenue, growing faster than everything else in the business. They signed a six-year deal with OpenAI. The stock photo company is quietly becoming a data company.
Then it got strange. OpenAI offered $500 million for Medal.tv. A platform where gamers upload clips of themselves playing video games. Two billion clips a year. Ten million users. The primary asset is teenagers screen-recording their Fortnite kills.
Medal said no. Spun out an AI lab, raised $134 million from Khosla and General Catalyst, and kept the data. Half a billion dollars on the table and they walked away because the data was worth more.
Text, images, video, audio. All getting licensed. And gaming services companies, the ones generating the most structured, most commercially specific operational data in entertainment, are pricing that asset at zero.
They're pricing the labor. They should be pricing the data.
This has happened before. In 2016, Quintiles merged with IMS Health. Quintiles ran clinical trials for pharma companies. IMS counted pills across pharmacies. Neither was exciting. Together, they became IQVIA. Market cap peaked north of $45 billion. The services produced proprietary data. The data made the services defensible. Neither half could have built the other.
That was healthcare. The same structure exists in entertainment right now. Nobody is building it.
The data nobody's pricing
So who actually owns this data?
Most people looking at gaming services stop at "translation." They see localization companies and think labor arbitrage. But a company that's localized 5,000 games across 60 languages isn't a translation vendor. It's sitting on a parallel text dataset. The same content, professionally translated into dozens of language pairs. That is exactly what multilingual AI models need for training. And the companies producing it every day have no idea what it's worth.
One company already proved it works. Flitto, a Korean localization platform, pivoted its multilingual parallel corpora into AI training data licensing. Revenue grew 77% year-over-year. Over 65% of revenue now comes from overseas data markets. They won a $7 million export award for it. Lionbridge and RWS both launched dedicated AI training data divisions. The localization-to-data pipeline isn't theoretical. It's operating.
Three of four are ready now. Localization data is already being traded. Trust and safety data is in massive demand: 80% of platform moderation budgets go to AI tools, and the market is growing at 26.6% CAGR toward $10 billion. Voice and dubbing connects to the $20 billion voice AI market. ElevenLabs just raised at $11 billion on $330 million in revenue. They need voice data. These companies have been recording it for years.
One caveat. This is the kill assumption, so I won't bury it: ownership. Standard localization contracts give the client ownership of the translation memory. The vendor is just holding it. Which means a localization company's text corpus might belong to its clients.
That's checkable in diligence. And it shifts the thesis toward the data types where ownership is less contested. Moderation decision logs. QA testing patterns. Voice recordings. These are operational byproducts, not deliverables. Nobody wrote a contract clause over a company's internal bug-tracking data. The localization data is still valuable if the vendor retained rights. But it's the upside case, not the foundation.
The Publicis precedent
Publicis acquired Epsilon's first-party data for $4.4 billion in 2019. Analysts were skeptical. CNBC ran the headline. By 2025, Publicis has dramatically outperformed WPP on every financial metric. Annualized ten-year return: Publicis +7.4%, WPP -9.2%. The re-rating wasn't immediate. It compounded over five years as data revenue grew. Wall Street doubted the deal, then spent half a decade watching Publicis pull away from every competitor that didn't make the same bet.
Where we are in the cycle
Every services-to-platform conversion follows the same four-phase pattern. I've tracked it across six sectors. The sequence is remarkably consistent.
Phase 1: First movers often prove the concept, but move over before taking meaningful territory in the space. PhyMatrix in healthcare. Rolled up physician practices in the '90s, collapsed. Hipgnosis in music catalogs. Proved the asset class, then the valuation collapsed. Both of them validated the thesis and cleared the field.
Phase 2: The vacuum. Smart operators build quietly while everyone else looks the other way. US Physical Therapy built during the post-PPM bust from 2002 to 2010. PE had fled healthcare services. No competing bidders. Distressed multiples. Eight years of compounding without a bidding war.
Gaming services is in Phase 2 right now, but Keywords has gone against the grain and proven to be the exception in an otherwise consistent trend. 60+ acquisitions, a proven consolidation model and a massive organisational restructure, committing to AI Solutions as a new service offering, and leveraging 25+ years of gaming experience to expand into data annotation, synthetic worlds and training for AI models..
Services-to-Platform Cycle · Six-sector pattern
- First Mover: Proves the concept, then exits or flames out. PhyMatrix in healthcare. Keywords Studios in gaming. Hipgnosis in music. Each validated the thesis.
- The Vacuum: Smart operators build quietly, gaining ground before the competition digs in. This phase makes the next decade, and we’re currently in it.
- Platform Expansion: Standalone units at 3–4× become platforms at 10–15×. Multiple expansion happens before market maturity produces the big outcomes.
- Data Differentiation: Platforms with proprietary data separate permanently from pure services players. The gap compounds. Publicis vs. WPP is the analog.
The question isn't whether this plays out. The analogs are too consistent. The question is who's building during the vacuum.
Contact Us Today
Contact us
Get in touch with our service teams to imagine more for your IP.
The Move
Entry multiples on gaming services companies are 5-8x right now. The cycle analogs say 3-5 years before those rise to 8-13x. Fifteen to twenty year runway to full maturity. This is the point where CrowdStrike entered cybersecurity. Nine years after SOX created the compliance demand. Well before market maturity produced $80 billion outcomes.
The play is straightforward. Acquire fragmented gaming services companies at services multiples. Architect for data capture from Day 1. Not as a bolt-on. Not as Deal 5 optimization. As infrastructure.
40% of PE return variance explained by entry price alone — across 50,000+ deals
The Publicis lesson is that data ownership is not something you figure out later. It's something you build for from the beginning.
Then the cross-sector positioning that no gaming services incumbent has pulled off. The same talent running QA for a major battle royale can run QA for BMW's virtual showroom. The same artists building game environments can build architectural visualizations. Qualitest proved it. Cross-sector QA positioning got them acquired by Bridgepoint at $200 million in revenue. BISim bridged gaming to defense simulation and exited to BAE Systems at $200 million. Same talent. Different label. 2x the multiple.
Right now, IT services firms like Globant, Accenture, and Capgemini are hiring gaming talent at a premium, marking it up 3-4x, and selling it to enterprise clients. That is a talent arbitrage being captured by the wrong intermediary. Build the $30-50 million version at gaming services entry multiples and you get enterprise exit multiples without needing sixty acquisitions to get there.
The obvious objection is synthetic data. If AI can generate its own training data, the scarcity premium on human-generated datasets collapses. Meta, Google, and Nvidia are all investing in exactly that. For commodity data, they're probably right. But operational byproduct data is definitionally resistant to synthesis. You can't synthesize authentic moderation failures. You can't generate the pattern of where real code actually breaks across 5,000 QA cycles. You can't fabricate how a native Korean speaker actually pronounces English game dialogue. Synthetic data works for the generic middle. It doesn't work for the specific edges where AI models actually fail.
Regulatory Timeline
EU AI Act enforcement begins August 2026. Article 10 requires AI training data to have documented provenance and licensing. Web-scraped datasets without clear rights chains become non-compliant. Companies with licensed, human-generated training data - localization firms, dubbing studios, QA providers - hold the only compliant supply.
Put simply: Licensed, provenance-tracked datasets become the only legal option for AI companies operating in Europe.
The window where these companies are priced as distressed services plays while sitting on appreciating data assets has a clock on it. Same deals. Same talent. Same EBITDA. Fifty to a hundred and ten million more in exit value from positioning alone. That gap is available right now because nobody in gaming services is thinking about it yet.
Keen for more insights?
Connect with JC on LinkedIn for more info on all things tech, gaming and AI, or get in touch to hear more about his work at Keywords Studios AI Solutions
JC Langley
The Assistant I Never Had
By Rich Tanksley, AI Senior Client Success Manager
My father had an assistant. Sometimes two. It used to be that your success as a leader came down, in part, to your assistants ability to solve problems and get things done. I have never had an assistant. But in the last six months something fundamental has changed in how I work, and I am now solving problems and getting things done exponentially faster than I was before.
For thirty years, my work looked roughly the same. I did things, talked to people, typed into a machine, produced things, and pushed them toward completion. My value to the company was largely my own input: the problems I personally solved and the output I personally produced. That is the part that has changed.
Today, much of my work is asking my AI assistants to produce things for me. It feels like having a really smart secretary. Or two. Or ten. It started with the boring stuff: find everyone attending this conference who fits our ICP and drop them into a spreadsheet. Then it got more complex: now add a column with each decision-maker's email, pulled from Lusha using our API key. Then more complex still: now add a draft email to each of them, referencing something in our product offering that speaks to a pain point they or their company has mentioned recently. Do that for all two hundred contacts.
I now have a team of very smart assistants doing the grunt work I used to do myself. They make mistakes. Sometimes they flat out lie. But they let me produce at two to five times the speed I used to. That is not a small thing. Almost every productivity gain in my career has been incremental. Even the computer only made you incrementally faster than the typewriter. This is different.
Here is what I did not expect. Handing off the grunt work has not made my judgment less important. It has made it the entire job. The assistant can build the list, find the emails, and draft the outreach, but it cannot decide which conference actually matters, which prospect is worth a real relationship, or which message will land because I met the person on the other end. The bottleneck used to be how fast I could produce. Now the bottleneck is how well I can think.
My father's assistants freed him to lead. These tools are doing the same for the rest of us, without the headcount and without the limit of ten. The grunt work is leaving. What remains is the part that was always the most valuable and the most human: knowing what is worth doing, and having the taste and judgement to recognize when it is done right. I have spent thirty years building that judgment. For the first time, I have the leverage to use all of it.
Want To Hear More?
Click below to learn more about Rich's background in the AI and Technology space, or connect with him on LinkedIn for more industry leading insights and opinions.
Rich Tanksley
In case you missed it - check out our previous editions:
Create
Globalize
Engage
