#444: A Gym Owner's AI Playbook: How BFT Tysons Is Running on AI in 2026

Here's the plan the best gym owners are following right now: https://www.youtube.com/watch?v=uMPx7b3_LOA

The gyms winning with AI right now are not using it to write emails. They built a machine learning algorithm that predicts who's about to cancel. They fired their ad agency and their bookkeeper. They have an AI employee living in Slack that nobody on the team can tell isn't human.

Pav spent 20 years in Fortune 500 companies before opening BFT Tysons. His financial planning background and tech instincts put him in a different category than most gym owners touching AI right now. In this episode, Mike Arce sits down with one of the most advanced operators in the GSD community to break down exactly what they've built, how they built it, and what's coming next.

In this episode you'll learn: — How Pav built a machine learning algorithm that ranks the 25 members most likely to cancel before they decide — The five member behavioral avatars the model uses and which signals carry the most weight — How BFT Tysons uses Claude to run Meta ads, optimize campaigns from the grocery store line, and launch new ads in five minutes instead of 45 — The AI marketing analyst that pulls weekly ad data, surfaces red flags, and sends recommendations directly to Slack — How Victor works as an AI employee inside Slack and the story of it onboarding a new team member better than the humans did — Why the era of information is over and what the era of imagination means for gym owners right now — Studio OS: the custom internal dashboard Pav's team built to track memberships, financials, and cancellations in one place — How AI analyzed BFT's three membership tiers and identified which members were most likely to upgrade — The connection between selling your highest tier first and actually getting members the results they came for — What Pav is focused on in Q3 and how AI identified three specific reasons revenue per member was declining — The security risks gym owners should actually worry about when building with AI — Why the AI revolution mirrors the printing press and what that means for the jobs being created right now

If you think you're using AI because you asked it to rewrite an email, this episode will show you how far behind that really is.

Episode chapters: — 0:00 — Introduction: Pav Grewal, BFT Tysons, and two years in GSD — 1:12 — From Fortune 500 financial analyst to gym owner: Pav's background — 2:08 — How their AI journey evolved from basic copy to machine learning — 2:59 — The attrition prediction algorithm: how it works and what it watches — 4:57 — Attendance as the top signal and the early dropout problem — 6:03 — The first 90 to 100 days: why onboarding drives long-term retention — 7:17 — Victor: what an AI employee actually looks like inside Slack — 9:21 — Victor's second day: onboarding a new team member better than everyone else — 10:56 — How Victor handled a HYROX registration list and corrected a missing name — 12:24 — Mike's P1, P2, P3 priority system and how Victor scheduled a meeting end to end — 13:21 — Pav's wife wants Victor too — 13:55 — Firing the ad agency: how Claude connects to Meta and runs paid ads — 14:46 — Building avatar research, hooks, and copy with AI before the ad is even created — 19:30 — Generating five to ten ad variations simultaneously and testing at scale — 20:58 — Optimizing campaigns from the deli line at the grocery store — 22:58 — The AI marketing analyst: weekly performance reports delivered to Slack — 23:46 — Competitor research through the Meta Ads Library: offers, angles, patterns — 26:04 — The difference between using AI for flyers and actually building with it — 26:58 — Firing the bookkeeper: Scott's move and QuickBooks integration — 28:23 — Why custom-built tools will replace bloated