How Crunchbase’s People team became the company’s test case for AI

Company

Crunchbase, private company data and predictive intelligence (about 250 employees, remote-first, headquartered in San Francisco)

Engagement

Hands-on AI workshop for the People team

Scope

A 15-minute intake call with each person, three 2-hour sessions a week apart, a team portal, a shared Slack channel

Partner

OneTwo Growth Studio

Timeline

August 2026

Key outcome

The People team now builds its own automations and agents, and team members now assist in builds for other teams across the company

Time AI saves Kelly in a typical week

10+ hours

Kelly’s data synthesis, before and after

“Forever” → 10 minutes

Yearly cost of the performance management tool the team no longer needs

$30,000

Crunchbase is the leading solution for headless private company data, helping business decision-makers stay ahead of private market shifts, identify opportunities earlier, and act first wherever they work. Only Crunchbase combines nearly two decades of proprietary data, relationship context, and AI-powered intelligence with unique engagement signals from millions of users to provide market insights and forecasts into funding, growth, exits and more.

Kelly Scheib joined Crunchbase in 2022 as its first Chief People Officer. She chose HR in college, studying labor and industrial relations and then human resources at Penn State, and she’s worked in the field since 2006. Before Crunchbase she was VP of Human Resources at Proterra. She teaches a range of courses at Anderson University at the undergraduate and graduate levels, holds a PhD in Leadership, and is the former board chair of Fostering Great Ideas, a nonprofit that advocates for children in foster care.

Kelly’s team already knew AI mattered at Crunchbase. The company’s CEO is “so AI forward,” she told me, that her team had always felt “we must play, we must be in the space.”

But being in the space looked different depending on where you sat. Engineering and product were already, in Kelly’s words, “eons beyond most.” The operational teams, People included, were busy with the work in front of them, and the tools changed faster than anyone had time to learn them. “Without enablement, I just have people who are lost and confused,” Kelly told me when we first spoke in June.

Her team was already using AI and applying it to real work. “Everything was a nail and they were using the AI hammer for every nail,” she said. “What I wanted us to do was be significantly more thoughtful about where we applied it.”

When something broke, there was nobody to ask. A connection to an external tool wouldn’t work, or a permission got in the way, and people stopped there. Kelly hit the same wall in her own work. “I don’t have a technical resource behind me to say, oh, this is what you need to do, or maybe this is a permissions thing.”

When I first met Kelly, I said she was ahead of her team based on the topics she was bringing up. But Kelly saw it differently. “I wouldn’t say that I was the most AI-forward on my team. I was probably the biggest pusher.”

Her team had been through prompt workshops before, so the sessions needed to be at a deeper level. Kelly didn’t want a live version of an online course. In June she described what she wanted. “I want to be the team that writes about what we’re doing. I don’t want to be the team that reads about what other people are doing.”

Kelly wanted the People team to go first. She called it “test case zero” for the rest of Crunchbase’s operational teams. The idea was to prove it on one team that wasn’t technical before taking it anywhere else.

That meant asking a busy team to give up six hours across three weeks, plus whatever time they spent trying things in between. With such a commitment on a topic that can feel heavy for many, there was almost no hesitancy in the room, so I asked Kelly how she’d set it up.

Kelly said she was honest with the team that the world had changed and that these were skills each of them would need, for their own sake as much as the company’s. “I would rather you lead from the front wherever you are,” she said, “be it at Crunchbase or be it somewhere else.”

Before the first session, I had a short call with each person to learn what they worked on and where they got stuck. Several had already drawn clear lines around the parts of their jobs that should stay human.

Everything was tailored on the tools the team already had, ChatGPT Enterprise and Zapier, plus the likes of people-specific work such as Teamtailor, the system they use for recruiting. So people learned on the same setup they’d go back to the next morning.

Then we ran three two-hour sessions, a week apart, giving enough time for experimentation.

The first session built a working understanding of how these tools operate. For a team with no technical support to call on, that understanding mattered most, because people who know why a tool behaves the way it does can judge its output, work out why something broke and decide what is safe to hand over. The second moved to the team’s own work, covering where human judgment belongs and the difference between being AI-assisted, which means doing today’s work faster, and AI-forward, which means asking whether the work should be done differently at all. The third looked at what people had built in between and went deeper into agents and sharing work across the company.

Between sessions, everyone had work to take with them and a Slack channel to share what they tried.

I talked to Kelly again about six weeks after the last session. To prepare, she’d gone back and listened to the session recordings again.

One change she named was how carefully the team now chooses where to use AI. “Just because you can do things one way using AI doesn’t mean it’s the best way to use AI,” she said.

The AI-assisted and AI-forward idea had stuck. Kelly had asked a member of her team to turn it into a guide other teams at Crunchbase can use on their own workflows, so they know where AI should help with the work as it is and where the work itself should change. She remembered the example we’d used in the sessions. “Maybe you don’t need a dashboard. Maybe what you’re looking for is a completely different way of presenting what your ask is and utilizing AI to do that.”

Her team had already done this. A teammate had built a site to show managers their team’s PTO balances. Emily Martin, an HR business partner, replaced it with a monthly Slack message sent through Zapier. “She did it and it was five minutes and it’s always correct,” Kelly said. “I do think there has been significantly more creativity with how things get done.”

Bigger pieces of work changed too, like the engagement survey. “It used to take them over a week of a team working on it to understand,” Kelly said. “And now it’s all done, customized through AI, extraordinarily quickly.”

Kelly’s team also built its own performance management process, and she no longer pays for a tool that costs $30,000 a year. Her reporting used to be a PowerPoint deck that someone on the team updated by hand. Now it runs through AI and sites built with AI. “That’s not happening anymore,” she said of the manual version. The team has also built “Zoe”, a policy agent that answers employees’ questions about company policy, to replace the third-party knowledge base tool they used before.

Her own numbers moved too. Kelly wrote that AI saves her 10 or more hours in a typical week. Synthesizing data, she wrote, used to take “forever.” Now it takes 10 minutes.

Kelly’s own daily briefing project has grown into a set of agents she directs. She calls it “way better than it was,” with “so much more to go.” On the change as a whole, she was direct.

The fascinating half shows up in what the team now does on its own. Before the sessions, people had ideas but often didn’t know how to act on them. Emily Martin, an HR business partner, described where she started. “I feel like I had a lot of really good ideas, but I had no idea how to start.”

The first change was knowing what was possible. Emily credits the sessions for that PTO automation. “I wouldn’t have known about it if I hadn’t taken the workshops and really understood even what an MCP could do, but also how to utilize ChatGPT.”

Then the work began to cross team lines. Other teams started coming to the People team for help with their own automation needs. When Crunchbase’s VP of Customer Success wanted to automate a part of the renewal process, Emily jumped in to partner with the CS team to bring it to life.

In the same meeting, the VP mentioned the quarterly voice of the customer presentation and the customer-call data it needed. Emily had already connected ChatGPT to Gong, so she told the VP she could pull what customers were asking for on renewal calls in about five minutes, and then she did. For someone who supports the engineering side and doesn’t usually work with the customer side of the business, it opened up a new view of the company. “It honestly has helped me learn a lot more about Crunchbase,” she said.

The craft improved too. People learned that the quality of the result depends on how clearly they direct the tool. “I think I am defining what good looks like a lot more. I’m also defining what bad looks like,” Emily said. When she and a coworker tried the same request side by side, Emily’s version found the answer in 30 seconds and her coworker’s never did, which Emily puts down to how each of them gave the instructions.

The knowledge also spread, inside the team and beyond it. Emily now asks colleagues the same question again and again. “Have you tried to use chat for that?” she asks, meaning ChatGPT. Often, she said, the answer is no because they didn’t know it could, and she shows them how. “It really has just been an amazing learning opportunity for everybody involved,” she said.

Kelly is just as open about the harder side of the change. When she asked her team to map out core processes like onboarding and offboarding to see what could be automated, she knew it was “scary to people.”

Her answer to that fear is the People team’s responsible AI use framework. It sets out how AI can support and improve people processes, and it’s explicit that AI can’t replace human judgment or employment decisions, something the workshops focused on in detail. Kelly has, in her words, “an entire people team framework” for what “must stay human.” She wants every use of AI tested against one question. “Is it worth it for us to let this skill atrophy?”

Understanding how the tools work changed how they felt about them. Kelly wrote after the last session that the workshop “made AI feel more practical, more accessible, and more relevant to real work.”

For Emily, the tools are now part of every working day. “I don’t know how I would get through my day without ChatGPT at this point.” “It has been a very, very eye-opening experience just kind of across the board,” she said.

I asked Kelly what she’d tell another CPO or COO deciding whether to bring someone in for three weeks or let their team figure it out alone.

“Hands down, bring someone in,” she said. “It’s going to augment you so much faster than I think fighting with things. Not only that, there are tips and tricks that we learned that we would have never learned.”

Her Instagram feed is full of influencers selling three-day and four-day courses, and she sees how much they look like what we did. She takes online courses herself. What made the difference for her was that the sessions were, in her words, “extraordinarily customized to us.” They started from the team’s own problems and were built in its own workspace.

She set that against “a generic class where you can build a skill.” The work went as deep as whiteboarding her own system and “how Teamtailor’s MCP connects and what that means.” “That was really, really valuable to us,” she said.

She also knows the learning doesn’t stop when the sessions end. Unlike Six Sigma, which you can learn once and know, AI keeps changing, she said, and “it’s just so hard to keep up.” Keeping up is one of her biggest worries. “I don’t want to be left behind in this revolution.”

When I asked after the last session what she’d tell other leaders, she wrote back. “I would tell them that the value is not just learning a set of tools. It is learning a better way to work with AI while keeping human judgment, context, and accountability at the center.”