MIT looked at the flood of enterprise AI spending and landed on a brutal headline: about 95% of companies are getting zero return. That’s not because the models are weak. It’s because most organizations are deploying generative AI like a software add-on instead of rewiring how they operate. We get specific about what the 5% do differently, and why “speed of execution” now matters more than having the fanciest tools.

We introduce a concept we’re calling AI ultraculture: a workplace where fear gets replaced by ownership, and where people understand why AI is important to the mission. From there, we unpack three practical moves: urgency that feels uncomfortable (think 90 days from pilot to production), a bottom-up AI roadmap that ends shadow AI and reduces confidentiality risk, and treating people as the moat because models and compute are quickly becoming commodities. We also talk about the investor lens, P&L pressure, and why cutting headcount can backfire by crushing morale, tribal knowledge, and real output.

Then we go hands-on with tactics like agentic loops that continuously assess what your tools can do, what your team can do, and what the business needs next. We also challenge “token maxing” and argue for token mining: same output, 10–20x lower token cost by using premium models for strategy and lighter models for production. To make it concrete, we walk through Beast Industries as a case study and how AI culture, workflow redesign, data, and distribution could compound into outsized growth.

Subscribe for more, share this with a leader who’s still stuck in AI pilots, and leave a review if it helped. What’s the single biggest bottleneck stopping your company from getting real AI ROI?

Support the show