Case Study — Google Engage ’26
Ready, Set, Go! — How We Built an AI Game Show for 10,000 of Google’s Closest Friends
How an AI game show for Google’s sales organization went from a sandbox brief to three arenas in three weeks — and the game underneath it that almost didn’t survive Christmas.
David Ebert · Creative Director | Director · For Google GBO, with SpecialGuestX
Ready, Set, Go! live at the Michelob Ultra Arena, Las Vegas — March 4, 2026.
3 cities — Dublin · Las Vegas · Macau · 10 minutes on the main stage · 90 seconds on the prompt clock · 5 departments, one team · 1 dress rehearsal
Nobody set out to make a game show. The brief, when it reached me in the fall of 2025, was not a brief. It was a sandbox: Google’s GBO — the Global Business Organization, the sales teams — wanted to play with the AI tools the company was shipping.
Here’s why that mattered, and it took me a while to actually hear it. GBO makes the money that enables Google to build these tools. Ads, sales, partnerships — not the glamorous end of the business, but the engine. And the people in that engine felt removed from the tools their work paid for. They sold a future they never got to touch.
So the real assignment wasn’t “demo the AI.” It was: let these people feel the impact of what they fund. Let them see themselves in it. That sounds simple. It was not simple.
Phase One — Developing the game
We tried everything first. That’s not a figure of speech, that’s the process. We tried vibe-coding video games live. We tried building animals out of blocks — the anonymous animals, the elephant and the capybara that pop up in a Google Doc when someone won’t sign in. We tried updating the Chrome dinosaur, the little T-Rex you race when the wifi dies. The technology wasn’t reliable enough, and honestly, that wasn’t even the real problem.
The real problem was that none of it had anything to do with what GBO does.
I want to be fair to those ideas. They showed off the technology beautifully. Image generation. Veo. Gemini. A tour of everything Google was launching, with confetti. But a demo of AI is not a job, and these people’s job is incredibly specific: they take a product and they bring it to market. They sell it. That is the whole craft.
I lost my mind about this for a while. It was around Christmas. I was not fun to be around. Every version I built answered the question “what can the AI do?” and the question I needed to answer was “what do these people do — and can the AI do it with them?”
Two things saved me. The first was scale. My early thinking was broadcast thinking — Google I/O, a few hundred people in a room, thousands watching a stream, informationally dense, a little dry. Then I understood where this was actually going to live: arenas. Dublin. The Michelob Ultra Arena in Las Vegas. Macau. The pitch deck described it as playing to “10,000 of your closest friends and co-workers,” which is pitch copy, but the rooms were real. You don’t think I/O for that. You think the Eras Tour. Where does the crowd look? When does the energy dip? How simple does an idea have to be before ten thousand people get it at the same time?
The second thing that saved me was my kitchen table. I stopped designing a demo and built the game on paper first. The assignment was a product that doesn’t exist for a customer who shouldn’t want it. A surprising thing meets a surprising audience — an oven for goldfish. Backpacks for cats. A video game console for vampires. I played it with my family. Toasters for ghosts. We named them, we pitched them, we argued about whether a ghost can push a lever (it can’t — so the toaster has to be voice-activated, obviously, ghosts can make spooky noises). We voted, ranked-choice, and somebody won.
It worked immediately. Because it wasn’t a tech demo anymore. It was their job, made stupid and fast and fun: take an impossible product to market.
Then I found the tools to build it for real. The team at SpecialGuestX had been building apps in Google AI Studio, in AI Labs. I started building the game there myself. Not a deck about a game. The game. It generated the brief — spin, and a product meets an audience. Then it ran the launch, step by step, and here’s the part I’m still proudest of: each step belonged to a different department.
One player sets the vision — the name, the problem this ridiculous product solves, the pitch. The next does the brand — logo, design, the merch shot. The next designs the product itself. The next writes the slogan and the print ad. The last one makes the commercial. Five departments — LCS, GCS, GP, gTech, GTM — five handoffs, and at the end there’s a complete product launch for something that was invented ninety seconds earlier. Their actual org chart, turned into gameplay.
Play with it — The original AI Studio build: Ready, Set, Go! https://ai.studio/apps/ceb5111d-e937-4617-9b39-a5a3b91f7650
I had rules for this, and I wrote them down on December 23rd because I needed them where I could see them:
The players are the Iron Chefs. AI is the oven, the knives, the ice-cream maker. A chef wouldn’t ask their oven to come up with the menu — so our players don’t ask AI to have the ideas. And AI cannot be a gimmick.
Then we fought about the clock and about choice, and both fights were really the same fight: how real do we let this be? Teams got 90 seconds to write their prompt and accept what came back. (The app started at 60. Sixty was too short — not tense, just broken — so we expanded it.) And I wanted players to choose: give them four outputs, let them pick. SGX wanted to take choice away. Their argument was volume, and time, and the one rule nobody could break — nothing inappropriate could ever, ever end up on a forty-foot screen in front of Google’s entire sales organization. I lost that one, mostly. The app that shipped literally describes itself as “one prompt, zero choices, pure execution.”
I still think I was right about the game. They were right about the show. That distinction was about to become my whole life.
Phase Two — Developing the game show
A game on a laptop is not a show. A show has to survive a room. So we tested it the way you test a show: on performers, in New York, at the Upright Citizens Brigade. I performed at UCB for years, and I wanted exactly those people — fast thinkers, collaborative, and crucially, not marketing people. If the game only worked for people who launch products for a living, it didn’t work. I barely briefed them. We ran it, and I watched what flowed and what stalled, and I went home and rebuilt the parts that stalled.
What I didn’t know at the time: that test was also an audition. Done+Dusted, who were producing Engage overall, were watching. Google was watching. Google, I would learn much later, was nervous — and had every reason to walk away if this didn’t work.
The LA test, Feb 10, 2026 — sizzle highlights (2:43). The format, nearly finished.
The second test, in Los Angeles in February, is where it became a show. Higher production value, a real stage, and the decisions that would survive all the way to the arenas.
The products were locked in advance — SUVs for space aliens, sneakers for chickens — settled with Juan at SGX and approved by Google. The audience experienced them as random. That’s not deception, that’s show business: you don’t improvise the parts that can break. The alien SUV became “Phone Home” — no doors, bio-integration dashboard, full-spectrum headlights. The chicken sneakers became “Roosts,” with an egg with legs for a logo.
The clock became theater. Ninety seconds is tension, but there’s no real punishment for losing — the consequence is that the other team was funnier. And the winner was decided the oldest way there is: the audience made noise, and we measured it.
One thing from LA I’m unreasonably proud of: the play-off music. When a player finished, we played them off, like the Oscars, like a hook at the Apollo. It told the audience “that bit is over, laugh now, here comes the next one” without a word of explanation. It worked in LA, so it went in the show, and it stayed in the show on three continents.
Behind the curtain (7:53) — direction off-camera, and an audience rehearsing how to clap. The least glamorous, most honest footage we have.
Then we had to cast it, and casting a game show for a company is a strange and wonderful problem. The contestants had to be real GBO employees — that was the whole point, the audience had to see themselves up there. So we ran it like a real game show. I defined the submission — what to send, how to send it — and around 300 videos came in. We sifted them by department, built teams out of strangers from around the world, and then I ran dry runs with the players in the app, teaching them the game while trying very hard to teach them nothing else.
A game show dies the second the players look rehearsed. I wanted them briefed just enough to not be lost, and in the dark about everything that would make show day feel like the first time. Because it was the first time. The outputs they saw on stage, they were seeing for the first time too.
Phase Three — Bringing it to the big stage
Here’s a number that concentrates the mind: three. There were three performances of this show, total, on Earth. Dublin, February 25–26. Las Vegas, March 4, Michelob Ultra Arena. Macau, the following week. No tour. No second season. You get one dress rehearsal in Dublin and then you are correct, in public, in an arena, or you are not.
And we had been squeezed. The segment had been built at twenty minutes, slotted at thirty in the early decks. On the day, we had ten. Ten minutes to explain a game nobody had ever seen, play it with people who had never played it, and launch a product that didn’t exist that morning. Everything I learned at UCB is in that sentence. Blocking is explanation. Pacing is explanation. A music cue is explanation. I directed the segment in all three cities — blocking, pacing, when the music comes in, when it gets out of the way — and because I came in with a theatre brain and a run of show I could defend, I got final say on all of it.
The Big Finish, Las Vegas — the winning team hits the button and the product launches.
My other job in those rooms was Brandon Baum, our host. A host can carry a format or he can stand next to it, and the difference is ownership. Brandon was exceptional at the job, and my work was to make sure he felt like it was his — that he was the one taking care of the players on that stage, because an audience can tell in about four seconds whether the host is protecting the contestants or just reading near them. He protected them. In every city, he protected them, and the players played bigger because of it.
The format, in the end, was four demos and a finish: build the pitch, build the product, work the crowd, find the market — then the winners hit a giant button and their product launched. Confetti. Eco-jets in Macau. A Veo montage of world domination, sales charts, fake news coverage, delighted customers for a sneaker a chicken will never wear. The audience, playing the client, green-lit it all by applause.
“One of the three best moments of Engage.”
It worked. I’m allowed to say that because Google said it first. EMEA “went GREAT,” per the email I got after Dublin, in all caps, which is the correct amount of caps. There were shout-outs to the game in Google’s internal comms. Strangers from the sales org were proud of a game show about their jobs.
And then, in Macau — the last city, after the last show — somebody finally told me. Google had tried to do game shows before. They had never worked. And the team had been willing, at basically any point in this process, to pull the plug on ours if it started to wobble. Nobody told me. Not during the Christmas meltdown, not during the UCB test that was secretly an audition, not during the one dress rehearsal in Dublin.
I wrote to Brandon afterward that I love learning we did the impossible only because no one had told us it was. I meant it then and I mean it now. If someone had sat me down in December and said “by the way, this format has a perfect record of failure and everyone is ready to cancel it,” I would have made a safer show. A safer show would have been a worse show. It might even have been the fourth failure.
Instead we made a game where the oven doesn’t write the menu, where ten minutes is enough if every second explains itself, where a sneaker for a chicken gets a world-domination montage — and where the people who pay for the future finally got to stand inside it, on stage, and play.
That’s the whole trick, as far as I can tell. Nobody told us it was impossible. So we just did the show.
Client: Google — Global Business Organization (GBO), Engage ’26 · Role: Creative Director | Director — game design, format development, segment direction (all three cities) · With: SpecialGuestX (SGX), SpecialGuest, 1stAveMachine · Production: Done+Dusted (Engage ’26) · Host: Brandon Baum · Shows: Dublin, Feb 25–26, 2026 · Las Vegas, Michelob Ultra Arena, Mar 4, 2026 · Macau, Mar 2026 · Built with: Google AI Studio, Gemini, Veo