Moving from AI anxiety to real productivity gains in HR

It’s always energizing to walk into a room and realize that many of us are wrestling with the same challenges.

My team is responsible for HR systems, payroll, and people operations, and more recently, I’ve taken on AI enablement.

What I want to share isn’t a polished success story. It’s an honest look at a journey that started only a few months ago and is still unfolding.

What surprised me most wasn’t the technology. It was how human this transformation turned out to be.

Our biggest lessons had less to do with tools and more to do with psychological safety, fear of obsolescence, and how people experience change.

Who LegalZoom is and why that context matters

Most people know LegalZoom from forming a business, filing a trademark, or creating a will or estate plan. Since inception, we’ve helped form about five million businesses and are the number one filer of trademarks in the United States.

While LegalZoom can look like a B2B company, it’s really much closer to B2C. Most of our customers are solopreneurs, people early in their business journeys who are navigating complexity for the first time.

That context shapes how we think about outcomes, empathy, and scale, not only for customers but internally as well.

Our work is guided by operating values, and two of them became especially important as we started our AI journey: focusing on results and seeing the whole picture.

When we look at our internal customers, the parallels with our external ones are striking. People want clarity, confidence, and tools that actually help them do their jobs better.

What pushed us into AI

Our AI journey has been short but intense. We’ve really only been doing this seriously since June. In AI terms, that’s long enough to see dramatic change.

Two trends pushed us forward. The first is the exponential acceleration of AI itself. AI is now training AI, no longer constrained by human-generated data or physical limits. That’s why improvement feels so fast.

We’re a Gemini house at LegalZoom, and the difference between using Gemini earlier in the year and using it just a few months later is dramatic.

The pace forced us to rethink traditional change management, because technology simply doesn’t usually evolve this quickly.

The second trend surprised me even more. Employees and customers are beginning to expect AI and, in some cases, to trust it.

Data started circulating showing that more than half of employees believe AI is less biased than humans in compensation decisions.

The signal was clear: we had to engage with this, thoughtfully and intentionally.

Choosing tools with governance in mind

We’re fortunate to have a CIO who is deeply invested in AI and helped us make smart, governed decisions about tools. Our core tools are Gemini, NotebookLM, and Glean.

NotebookLM has been especially interesting. It’s our most popular tool in terms of satisfaction, with around a 90% OSAT score, but only about 39% of the organization knows how to use it well.

That gap alone showed us how critical enablement is.

Glean started as our intranet and internal search tool, but its AI capabilities transformed how people access information. Because it fits within our security model, it’s particularly valuable for teams working with sensitive people data.

The lesson wasn’t about having the most tools. It was about having the right ones, clearly sanctioned, and paired with education and support.

Structuring the journey for our people and places team

We decided early that we needed a journey, not just training sessions. We started with the People and Places team in September, and that work has since expanded across the company.

At the beginning, many people were new to AI. We focused on fundamentals like prompt engineering, what makes prompts effective, and how to move beyond generic outputs.

One of our most effective approaches was a series of sessions we call “Let’s see what AI can do.”

These were practical, role-based sessions where people demonstrated how AI could support real work, like managing inboxes or analyzing data. Seeing peers apply AI in concrete ways made the value tangible.

We also launched an internal hackathon called HackPnP. Instead of a two-day sprint, it ran for a month. Teams had time to engage deeply, collaborate with data partners, and let the technology mature alongside their ideas.

That time made a meaningful difference.

What we learned about engagement and safety

Two lessons stood out quickly. The first was that content matters. Engaging, well-designed presentations kept people attentive and reduced the tendency to tune out, especially in virtual settings.

We measured engagement through participation and feedback, and it made a real difference.

The second lesson was even more important: the power of real talk. The fear of being obsolete is real. Avoiding the question of whether AI might take someone’s job erodes trust.

Being direct, honest, and compassionate created psychological safety. When people felt heard and respected, they were far more willing to experiment and learn.

Using storytelling to humanize AI

To frame the journey, we leaned into storytelling. Initially, I thought about a Superman metaphor, but a colleague encouraged a shift to The Incredibles. That change mattered.

In that story, heroes are stuck in mundane roles, weighed down by bureaucracy. AI, in our framing, removes drudgery and allows people to operate at their best.

We introduced the idea of becoming a T-shaped resource. AI allows people to broaden their understanding across areas while deepening their expertise.

I experienced this personally as I learned more about benefits and leave processes while building AI workflows, despite my background being primarily in analytics.

This shift changes productivity. A single analyst, supported by governed AI tools, can become a “super analyst,” answering questions across domains without constant handoffs.

Naming what holds us back

We also found value in naming the barriers to adoption. Change fatigue is real, especially after years of disruption. According to Gartner, the biggest barrier to AI adoption is the belief that learning it takes too much effort.

We called this the “until spiral,” where people tell themselves they’ll try AI after the next deadline, and that moment never comes.

Naming it helped us confront it collectively, especially through HackPnP and hands-on sessions.

Listening to fear through anonymous feedback

We created space for honest, anonymous feedback through monthly surveys. Fear didn’t disappear overnight, even as adoption increased.

We consistently saw three groups: early adopters who were all in, a large group that was curious but needed skills, and another large group concerned about AI’s long-term impact on their careers.

We addressed these concerns directly by discussing what work lends itself to automation and what doesn’t. Critical thinking, empathy, and judgment remain deeply human. AI can support them, but not replace them.

We also used real-time feedback during sessions to catch confusion as it happened, helping people move from zero to one more effectively.

Solving the last mile problem

The hardest challenge was the last mile. People could experiment with AI, but struggled to turn experiments into real workflows.

That’s where “Let’s see what AI can do” sessions proved invaluable. An AI superuser would work directly with someone on a real problem, build a prompt together, then hand ownership to the user.

By the end of the session, the solution belonged to them.

One example involved a benefits analyst who answered the same questions repeatedly. Together, we built a prompt, aligned on tone and legal considerations, and gave her ownership.

She became one of our strongest early adopters.

This approach is high-touch and resource-intensive, but it works.

The results so far

The results have been encouraging. When we started, People and Places used AI at roughly the same rate as the rest of LegalZoom. Today, our usage is about double the organizational average.

We estimate four to 5% productivity gains in HR from these initiatives alone. Our long-term goal is a 10% productivity lift, which would correspond to roughly a 50% adoption rate.

Measurement is challenging as different teams use different tools, but the goal feels attainable. More importantly, the conversation has shifted. AI is no longer just hype or fear. It’s becoming a practical capability tied to real outcomes.

Closing thoughts

AI transformation isn’t just technical. It’s human. Tools matter, but trust, safety, and ownership matter more.

Adoption happens when people feel safe enough to experiment, honest enough to express fear, and supported enough to turn curiosity into impact.

This article is based on David's brilliant talk at our People Operations Summit. Check out our upcoming events.

Join our (free) Insider membership to access exclusive content, including ondemand footage from our events and ready-to-use templates you can apply in your role.

How we moved from AI anxiety to real productivity gains