Launching Before Consensus

Welcome to "Zapier's AI Transformation Journey." In this course, you'll unpack insights from Brandon Sammut's Transform 2025 session, where he traced Zapier's two-and-a-half-year climb from a CEO "code red" post — when the majority of the company wasn't on board — to 97% of employees using AI in their core work. Across three units, you'll explore how to launch an AI mandate before it feels safe, scale fluency through existing work rituals rather than competing programs, and set transformation goals that play for multiples rather than marginals.

Let's start with the moment that kicked everything off — and why waiting for consensus would have been the costliest mistake Zapier never made.

You'll recall that in March 2023, about six months after ChatGPT 3.5 exploded into mainstream awareness, Zapier's CEO Wade published what Brandon called a "code red" — a company-wide post calling 100% of employees to get hands-on with AI models. The critical detail: only about 40% of the company welcomed it. For the other 60%, as Brandon put it, "it felt too soon and pretty alarmist." Two and a half years later, his reflection was unequivocal — "I'm really glad we didn't wait" — because experimentation became the primary way people understood why AI mattered and worked through their fear, uncertainty, and doubt.

But a mandate alone isn't enough. As Brandon emphasized, you need to bundle the call to action with specific enablers within days, just as you would with any major organizational change. Zapier shipped three: AI use guidelines designed to maximize creativity by defining a sandbox rather than constraining thinking, enterprise LLM licenses for 100% of the company with "unconstrained budget," and the repurposing of their next quarterly hack week into an AI-focused event open to everyone — not just engineers. The principle was clear: don't let the call to action be "a bit of a just a platitude." Give people toeholds they can grab immediately.

Building on those enablers, the redesigned hack week became one of Zapier's most powerful activation tools — and the design choices are what made it work. Brandon described having participants submit in advance the thing they wanted to experiment with, often a frustration or opportunity from their daily work. Because themes naturally emerged, people were grouped into twos and threes around shared problems. Each group was paired with a power user — sometimes pulled from other departments — who could help them get started and get unblocked. They built for "scarcely 90 minutes," then presented demos.

The results were both emotional and practical. Brandon described this first hackathon with the People team as the only time he "almost cried" at Zapier — watching people who had been genuinely scared of AI present what they'd built with visible pride as their FUD dissolved. More than sentiment, though, two of the tools built were "production ready within about a week." The lesson for any people leader designing these events: intrinsic motivation around real work problems, not technical skill, is the engine. Power users provide the scaffolding, but voice and choice provide the fuel.