From Google voice to AI-first communication: Dialpad's Brian Peterson on leading AI adoption

Dialpad CTO Brian Peterson built Google Voice for 10 million consumers. Then he left Google, co-founded Dialpad, acquired an AI company eight years ago, and built real-time AI into a unified communication platform from scratch. WorkOS CEO Michael Grinich sat down with Peterson at HumanX 2026 to talk about what it actually looks like when a company goes all-in on AI—not as a feature, but as a mandate.

The origin story

Dialpad's roots trace directly to the Google Voice team. Peterson and his co-founders took the same cloud-first approach that scaled to millions of consumers and pointed it at business communications. Eight years ago, they acquired an AI company and started building real-time AI capabilities on their own models—long before the current wave.

Today, Dialpad is a fully unified customer service and sales platform: omnichannel, with both automated AI agents and augmented human AI that provides live coaching and insights during calls. The AI isn't bolted on. It's the foundation.

No Code without AI

At Dialpad, AI adoption isn't optional. It's policy.

"Engineering is now required to not write a single line of code without AI," Peterson said. But he's clear-eyed about where things stand: "I'm not saying it's going to solve all your problems and make you 10x. But there's still a lot you can get productive with."

The company published an internal manifesto covering AI adoption across every function—marketing, customer service, sales, engineering—and hired dedicated AI leads for each department, each responsible for moving their function from manual processes to AI-assisted ones. That's operational infrastructure.

Lead by example

When asked what advice he'd give other CTOs and CIOs trying to drive AI adoption, Peterson didn't hesitate: lead by example.

Get your most respected senior people using AI visibly. Have them demo what they've built. "People look at that and think, 'The person I look up to is using it—now I need to use it,'" he said.

Then repeat the message constantly. Track adoption, but don't micromanage. The goal is cultural momentum, not compliance theater.

The jevons paradox is real

One theme echoed across multiple conversations at HumanX: productivity gains from AI don't lead to smaller teams. Peterson confirmed it.

"Studies show people are working more, not less," he said. "When you remove the simple things, you're free to do really creative things. Whenever there's an abundance of something, there's higher demand—not the opposite."

This is the Jevons Paradox applied to labor productivity. The original paradox, observed by William Stanley Jevons in 1865, described how improvements in coal efficiency led to greater total coal consumption, not less. Peterson is drawing an analogy: when AI makes certain tasks trivial, people do more of them—and tackle harder problems they wouldn't have attempted before. Whether this analogy holds precisely is debatable (Jevons described consumption of a single resource, not reallocation of human effort), but the directional observation tracks with early data on AI tool usage.

And competitively? "No one who's actually going to do something innovative is going to say, 'We can do the same amount with less people, so let's do that.' Leaders aren't thinking that way."

Personal AI: From soccer schedules to performance reviews

Peterson's personal AI usage goes deep. He uses Claude to automatically check his calendar for his kids' soccer games and text the schedule to his parents. At work, Dialpad automated performance reviews with AI—humans still verify the output, but the drafting and synthesis happen automatically.

"I'm trying to automate everything," he admitted. "I don't know if that's normal. I think that's a little crazy."

The people who push the tools hardest are the ones who find the failure modes fastest.

Grounded optimism

Despite his enthusiasm, Peterson is measured about the timeline. Customer service—Dialpad's core domain—will see dramatic improvement within a year. "No more press-one-press-two," he said. The technology is already there; it's a matter of deployment and trust.

But for rolling out AI broadly across an organization's internal functions? Peterson is more cautious.

"I don't think it's as magical as people think. We're still limited to the large language model. It's still non-deterministic."

His estimate: 80% of best-case within five years.

That's the kind of grounded take that's easy to overlook in a room full of AI optimists, but it matters. The companies that build lasting AI adoption aren't the ones chasing 10x overnight—they're the ones systematically pushing toward 80% of what's possible and compounding from there.

This interview was recorded at HumanX 2026 in San Francisco.