Zapier's Chief People Officer on What Separates AI Adoption From Real AI Transformation

Zapier's customer support team is the company's clearest proof that redesigning work around AI, not just speeding it up, is what produces transformative results.

The team's mission was simple and measurable: "fast, accurate support," tracked through average ticket handle time and customer satisfaction score.

About two years ago, support leader Lauren stood in front of her team after prototyping new AI-driven workflows and made a public commitment. Cut average handle time in half over 12 months while holding or improving CSAT.

That's a 2x productivity gain stated out loud, in front of the people who'd have to deliver it. They hit both targets.

Then something Sammut didn't expect happened: "While they were doing all this good, hard re-engineering of their core ways of working, their employee engagement scores went up 20 or 30 points on a scale of 100."

What made it work, according to Sammut, wasn't the technology. It was that Lauren "distilled the why and the why now all the way down to the individual level."

She held open "ask me anything" sessions with her managers with no questions off the table, which sharpened her own team's thinking as much as it clarified things for the group.

And critically, the redesign went past the surface: "the way that they work has been redesigned altogether, including the staffing model and the roll cards, aka the job descriptions."

That DNA-level rebuild, not a faster version of the old process, is what separates genuine transformation from a productivity bump. It's the same logic behind what's now being called AI native organizations: companies built around how AI actually changes the work, rather than AI bolted onto how work already happened.

Clarity of purpose is the foundation, but it only works alongside a culture where experimentation and failure are genuinely safe. Sammut named four ingredients Zapier leans on: clarity of purpose, a culture of experimentation, psychological safety, and trust in management.

"AI itself is a technology. It's like a medium... it doesn't tell us what to use it for. It's kind of a canvas," he said.

Organizations with clear goals already have a foundation for using that canvas well. Organizations without it just add uncertainty on top of uncertainty.

Psychological safety matters because nobody, including leadership, has this figured out yet. "None of us are ever feeling like caught up or totally knowledgeable," Sammut said.

"Being able to say, I don't know how to do this, I need help, I tried something, it didn't work," is what keeps people experimenting instead of hiding failures.

He was also candid about how fragile that culture can be. After Zapier's 2023 restructuring, engagement metrics around feeling safe to experiment dropped and took nine months to recover.

"You have to be really thoughtful about the knock-on effect of big changes or big events in the company that can impair some of the cultural ingredients for AI transformation."

How Do You Get Middle Managers to Actually Buy In?

The single most effective tactic for building a culture of experimentation is dedicated, structured time, specifically hackathons where senior leaders build in public.

Asked how to get mid-level managers past "just tell me what to change and I'll implement it," Sammut pointed to Zapier's builder sessions.

What makes them work is that leaders go first: "The whole senior leadership of whatever org is doing it is going to be there and not just be there, they're going to be experimenting too. They're bringing an idea to the table."

Leaders demo their unfinished, imperfect work alongside everyone else, which removes the stigma of showing something that doesn't fully work.

He described running one with his own people team: "I was almost in tears because you could see, this is a matter of like three hours, right? It's not a week long or even a day long thing. Just the boost in folks' confidence."

The format also solves a practical problem, since most people won't experiment without dedicated time, a clear use case, and someone to ask when they're stuck.

For managers specifically, Lauren's team ran small, unfiltered "ask me anything" forums.

Zapier also folded AI-related expectations directly into its company-wide performance criteria, which it calls impact behaviors. "It's just a core part of how we evaluate our performance," Sammut said.

Zapier hires against a four-tier AI fluency rubric, from unacceptable to transformative, and openly shares it so candidates know exactly what's expected.

Sammut walked through the framework live on the call. "Unacceptable" is a floor the company won't hire below.

"Capable" describes people dabbling with AI tools with a strong growth mindset but no proven results yet. "Adoptive" means they've translated that experimentation into measurable outcomes.

"Transformative" means they've rebuilt how the work gets done at a structural level and scaled that new way of working to others.

Notably, mindset counts as much as skill: "If you look across all four tiers, there's as much in here about mindset as there is about skill. And that's intentional."

Zapier doesn't rate existing employees against the framework in performance reviews. Instead, it asks people to self-assess and align with their manager on where they need to grow.

"We're more interested in the pace of progress for our existing team members than we are exactly where they are on the framework," Sammut said.

And crucially, not everyone needs to reach the top tier. On Lauren's 80-person support team, only two to five pioneers needed to operate at "transformative" to engineer the new way of working. The rest needed to be "adoptive" to execute it.

How Do You Address Job Security Fears Honestly?

The most credible answer to "will AI take my job" is the honest one: nobody knows, and pretending otherwise erodes trust.

When asked how Zapier keeps people experimenting despite visible anxiety about job security, Sammut didn't offer reassurance. He offered honesty.

"Are we going to still have recruiters in two years? I don't know. I don't know," he said, describing his actual answer to his own team.

"I deeply believe that humans have a role to play in recruiting, even with all the goodness that AI is bringing to the table. I do not know how many of which types of roles we will need two or three years from now."

He pairs that admission with a second, more hopeful point. Zapier can't promise headcount, but it can promise to be "the place where you learn how to be an elite AI supported recruiter."

Part of how the company invites hard questions in the first place is a practice called the "Rude FAQ." The team writing a major communication deliberately drafts the toughest, least polished questions employees might have and answers them head-on rather than in corporate language.

• Separate adoption from transformation before setting goals. Adoption produces 20 to 40 percent individual productivity gains; transformation targets 3x to 10x by redesigning the work itself, staffing model included.
• Choose your transformation lead for judgment and credibility, not job title. Tenure, cultural fluency, and relationships matter more than which department they sit in.
• Set a measurable team-level mission before introducing AI. Zapier's support team rallied around "fast, accurate support" with two clear KPIs, which gave the AI work a concrete target.
• Run recurring hackathons where leaders build and demo in public first. Dedicated time, visible leadership vulnerability, and access to skilled builders removes the biggest blockers to grassroots experimentation.
• Answer job security questions honestly instead of reassuringly. Admitting "I don't know" about future headcount, paired with a real commitment to skill-building, builds more trust than false certainty.

Brandon Sammut is Chief People Officer at Zapier, where he has led the company's AI transformation since its 2023 "Code Red." This article is based on his fireside chat hosted by Lead with AI PRO.