Thinking · Newsletter
Booksy's creative playbook, read against the five stages of creative operations
Gus Murillo described how a marketplace app runs creative: read first, brief from the read, edit for volume, keep the human. Where it matches our method and the two places we would push further.
TL;DR. At Gamesforum in October 2026, Booksy's global director of creative and community, Gus Murillo, described how a marketplace app with millions of beauty-professional users runs creative: intelligence first (an ad-analysis tool reads what hooks, formats and personalities are working before any brief), briefs and scripts generated from that read, talent selection informed by it, and an edit step that turns one piece of creator content into "30, 40 different versions... actually hundreds." He also said pure AI ads, generated visuals and voice, "significantly underperform AI-edited influencer ads with a real beauty provider." That is a creative operation built the way we think one should be built, from a company that is not an agency. Here is the playbook read against the five stages of creative operations, and the two places we would push further.
Who said it, and where
Gus Murillo, Global Director of Creative and Community at Booksy, interviewed on Gamesforum's channel in October 2026 ahead of a San Francisco session on AI in creative production. Booksy is a marketplace for appointment-based beauty providers, the largest in its category by Murillo's account. Quotes are from the recording.
Stage 1, read before you brief
"If you think of pre-brief: how do we even monitor what is working and what isn't working on organic or paid? ... Motion app analyzes all of the ads that we have in Meta and TikTok and then delivers back to us what are the messages, the hooks, the formats, the styles, and the personalities that are working effectively for us. So AI is that intelligence layer that is currently telling us what's working and what's not working."
This is the stage most teams skip and the one we build everything on. Booksy reads its own account. We read the account and the category: Flixr Studio tracks every ad running in a client's category on AppLovin, Meta and TikTok and calls winners by what the networks keep funding, so the brief starts from what the market rewards, not only from what the brand has already tried. Same principle, wider lens.
Stage 2, the brief comes from the read
"Then we've been using AI to generate the briefs itself... tools that will analyze all of what's working and then help us generate the briefs of what we need to build next... instead of AI processes just helping us speed up the time it takes to make a script, we've been using it to inform what scripts we even deliver based off of what is working in platform."
The distinction he draws is the right one: speed is the small win, direction is the large one. A model that writes a script faster has saved an hour. A read that tells you which script to write has saved a shoot.
Stage 3, the edit is where the volume comes from
"We've been testing a lot of the tools, Opus and some of these editing tools that will allow for us to take large swaths of content from people and creators and then edit and re-edit and develop 30, 40 different versions of them. I'd say 30, it's actually hundreds... out of any individual piece of content, we can extract significant edits, do captions, titles, and pacing, and flow, and music."
Hundreds of variants from one piece of creator content is the concept-system method by another name, and it is the only way any team reaches the volume the platforms now reward. Our note on it is the edit is where creative volume is won or lost.
Stage 4, the human in the frame is not optional
"We've done pure AI ads, meaning AI developed visuals, AI developed voice-over, and I've seen those significantly underperform AI edited influencer ads with a real beauty provider, a real nail tech, and a real barber."
And the reason, which is about the audience rather than the technology: "it is a trend in the beauty industry to literally say, I made this with my hands... there's an aversion that we're staying very close to... we've monitored sentiment when we've done fully AI content."
We land in the same place from the other side of the table: generative tools for environments, variants and edits; a cast human for the face that has to carry trust. The reasoning is in AI UGC: where it wins, where actors still do and we are AI-native and we still pay actors.
Stage 5, where we would push further
Two places.
Judge the read on purchases, not on engagement. Booksy's intelligence layer reads "messages, hooks, formats, styles, personalities." Those are diagnostics. They explain which part of a working ad did the work. The decision about what to make next should rest on the outcome, bookings or purchases per thousand impressions, with the diagnostics explaining it afterwards; otherwise three months of great hooks can sit on a flat account. See measure creative on purchases, not on the ad.
Read the category, not only the account. An account read tells you what your own ads did. A category read tells you what the networks are funding for everyone in your space, including the motivations you have not tried. For a marketplace fighting on both sides (providers and consumers) that second lens is where the new concepts come from.
Why this matters beyond beauty
Booksy is not an agency and has no reason to flatter the method. It arrived at read, brief from the read, edit for volume, keep the human, because that is what its own numbers rewarded. The same five stages are what we run for games and commercial brands, and the same two gaps (outcome over diagnostic, category over account) are the ones we see most often when we open a new client's account. The full definition is in creative operations, defined.
Questions people ask
What is Booksy's creative process? Per Murillo: an AI intelligence layer reads what is working in paid and organic, briefs and scripts are generated from that read, talent is selected against it, and AI editing tools produce dozens to hundreds of variants from each piece of creator content.
Do AI-generated ads work for Booksy? He said fully AI ads (generated visuals and voice) significantly underperformed AI-edited ads featuring real providers, and that the audience has a measurable aversion to AI-made content.
What tools did he mention? Motion for ad analysis, Claude and Motion's "Run It" for briefs, Opus and similar tools for editing variants.
How does this compare with how Flixr works? Same five stages. Flixr adds a category read through Flixr Studio and judges the loop on purchases or installs in the client's own accounts rather than on hook and format metrics.
Source: Gamesforum, "Inside Booksy's Creative Playbook with Gus Murillo", YouTube, October 2026. Quotes transcribed from the recording; interpretation is Flixr's.