Monks Made a Puma Ad in 5 Weeks With AI. The Other 60% of Their Agency Is Still Catching Up.
By Notice Me Senpai Editorial
Agencies have spent the last two years telling clients they’re “investing in AI.” Mostly that meant buying a few ChatGPT Enterprise seats and putting together a deck about it. Monks, the agency formerly known as Media.Monks under S4 Capital, just made those decks look quaint. They produced a 60-second Puma commercial using AI agents, from concept through final edit, in five weeks. And 40% of their 7,000 employees are actively using the internal AI platform that made it possible.
Those are not aspirational numbers on a pitch slide. Those are operational numbers. And they’re creating a gap that’s going to be very hard for traditional agencies to close.
Five weeks for a national spot is not normal
A traditional 60-second commercial for a brand like Puma typically takes three to five months. Strategy, creative rounds, pre-production, shoot days, post-production, review cycles. Monks did it in five weeks using a combination of Runway’s video generation models, Google Veo, and NVIDIA’s Omniverse platform for 3D scene composition. The process handled “from conceptualizing all the way through to the final edit with minimal human interaction,” according to the official release.
I think “minimal human interaction” is doing a lot of heavy lifting there. Someone still had to brief the AI, evaluate the outputs, make subjective creative decisions. But even accounting for that, compressing a multi-month production into five weeks is a significant change in what’s economically possible. The cost structure shifts dramatically when you remove most of the shoot logistics, the talent scheduling, the physical post-production pipeline.
On paper, this sounds like a pure win for agencies. Faster delivery, lower costs, higher margins. And sometimes it is. But the uncomfortable follow-up question is what happens to the people who were doing the work that got compressed.
The 40% number is the one that actually matters
Monks built an internal platform called Monks.Flow that connects AI models across research, content generation, and media planning. Eight of their top ten clients regularly use AI-generated outputs from this system. Around 40% of employees across the agency actively use it in their daily work, and the agency reported 2.8x faster delivery of agentic workflows through the platform.
That 40% figure is more interesting than the Puma spot itself, honestly. It means adoption is real but not universal. Sixty percent of a 7,000-person agency is not yet using the primary AI platform daily. Some of that is expected (not every role needs it), but some of it probably reflects the same inertia you see at every large organization trying to get people to change how they work.
From what I’ve seen across agencies and in-house teams, the adoption curve tends to create two distinct groups pretty fast: people who build AI into their workflow and become measurably more productive, and people who keep doing things the old way and gradually become harder to justify at the same cost. I don’t think Monks is unique in this. They’re just the first agency putting real numbers to it publicly.
Nvidia is turning this into infrastructure, not just a demo
The hardware side of this story matters more than most coverage is giving it credit for. Nvidia VP Richard Kerris told Adweek that their Vera Rubin (next-generation GPU architecture) can achieve under 100 milliseconds time-to-first-frame for AI video generation. That’s not quite real-time, but it’s close enough that creative teams can iterate on video concepts the way they currently iterate on static images.
“That speed changes everything,” Kerris said. “Today, with agentic AI, you’re conversing with your computer.” Which sounds like a press quote (it is), but the underlying shift is real. When video generation takes seconds instead of hours, the entire creative review process changes. You can try forty variations of a scene in an afternoon instead of two.
Runway has been training a new real-time video model specifically on Vera Rubin hardware, according to PetaPixel. That puts pressure on OpenAI’s Sora and Google’s Veo, both of which still operate in slower, batch-style generation modes. And Runway’s own announcement frames the Nvidia partnership as a long-term infrastructure play, not just a one-off demo.
This is where the structural advantage starts to compound. Monks has the Nvidia partnership, the custom internal platform, the trained workforce (well, 40% of it). Agencies without those relationships are going to find it increasingly difficult to match those production timelines and costs, even if they adopt the same general-purpose AI tools.
The NAB demo hints at where this goes next
At NAB 2026, Monks also demonstrated AI-powered automation for live broadcasting, including 5G-enabled camera switching controlled by AI agents. That’s a different application entirely from commercial production, but it signals a broader ambition: Monks wants to be the agency that treats AI as core infrastructure across every type of content production, not just a tool for making social ads faster.
And to be fair, this isn’t entirely new territory. Agencies have been automating parts of production for years. Dynamic creative optimization, programmatic ad assembly, template-based versioning. What’s different now is the scope. Going from “we can auto-resize a banner ad” to “we can generate a national commercial in five weeks” is a pretty significant jump. The tools just weren’t capable of that twelve months ago.
The gap is structural, and it’s widening
Personally, I think the most important takeaway here isn’t about Monks specifically. It’s about the emerging divide between agencies that are building AI into their operational infrastructure (custom platforms, hardware partnerships, trained teams) and agencies that are still treating AI as a set of point solutions their employees can optionally use.
The first group gets compounding returns. Each project makes the platform smarter, the team more fluent, the client relationships stickier. The second group gets the same general-purpose tools as everyone else and competes on the same margins they always have.
If you’re at an agency or managing agency relationships in-house, the question worth asking this week is pretty specific: does your creative partner have an internal AI platform with measurable adoption numbers, or do they just have a slide about AI in their capabilities deck? The difference between those two things is about to start showing up in timelines and invoices.
YouTube recently handed creators AI video tools for branded effects, and that’s another data point in the same direction. The infrastructure for AI-generated video is becoming available at every level, from solo creators to 7,000-person agencies. The question isn’t really whether this technology works anymore. It seems to work fine. The question is who builds the operational layer around it first, and how wide the gap gets before everyone else catches up.
I don’t have a confident answer on that timeline. But five weeks for a Puma commercial is a data point that’s hard to argue with, and most agencies can’t do that yet. Probably not even close.
