Hello HR Tech Enthusiasts,

Two quick housekeeping notes before we jump in.

This month’s article is a bit of a hybrid.

Expect some of the usual “what’s catching our eye,” weaved together with some insights from a few recent interviews with Matt MacInnis, Rippling’s COO, and Ian White, ChartHops’s founder and CEO/CTO.

The headline this month is pretty simple: AI agents have arrived in People Ops.

For the last couple of years, AI in HR has mostly meant some version of summarization or generation.

You can ask a question about someone’s benefit eligibility and get an answer. You can generate a job description. Draft a performance review. Get recommendations for a learning pathway.

All of that is useful and has materially saved time for HR practitioners and employees alike.

But, it has all been step-wise improvements on what software could previously do.

What we’re starting to see now is something different.

In my interview with Rippling’s COO Matt MacInnis, he showed how an AI agent can take a list of employees and automatically apply spot bonuses across multiple countries, handling payroll timing, currencies, and compliance without manual input.

ChartHop’s announcement included an AI agent building and running multiple headcount scenarios, then pushing those plans into action.

This is an eye-opening shift from AI surfacing information to AI actually taking action inside the system.

It’s important to stress that it’s we’re still early in this transformation. Most client experiences have taken place in beta tests and controlled scenarios. There’s a lot that we will still learn about the strengths and shortcomings as these tools get more exposed in the market.

But I feel comfortable saying that these announcements are a real inflection point and a new chapter of the AI in HR story.

What stood out to me in the interviews with both White & MacInnis was how quickly the conversation shifted to product architecture.

I asked White an elementary question: “What’s the difference between using ChartHop AI and just dropping data into ChatGPT?”

He immediately went to permissions. Robust permissioning is something ChartHop spent years prioritizing and making a calling card.

The result? Granular controls over who can see what, at what level, across what context.

MacInnis framed it slightly differently, but landed in the same place. According to him, the things allow Rippling AI to take agentic actions are:

Different words, same idea.

Every HR Tech vendor has access to the same frontier models (Claude, ChatGPT). But, the reason some systems are harnessing those models to deliver more impressive outcomes is because of how they were built beneath the surface.

For years, I’ve been blue in the face trying to educate buyers on the importance of product architecture. Why unified systems matter. Why stitching tools together creates downstream problems.

But it’s always been a bit abstract.

One of the biggest implications of the AI agent moment is that appreciation of product architectures will now be very concrete.

If an agent is going to move across your system, understand context, and take action safely, you can’t fake the foundation.

There’s another implication here that I don’t think gets talked about enough.

I believe that front-ends are about to matter a lot less.

For a long time, buying decisions were heavily influenced by UX. How many clicks something takes. Whether the menus make sense. Whether it “feels” clean.

Truthfully, those considerations mattered more than they probably should have in buying decisions.

But if everything is moving toward a chat-based interface, the question stops being, “How easily can I navigate this tool?” and starts being, “How reliably can agents navigate this tool?”

And the benchmark that buyers will expect isn’t their last HRIS.
It’s Claude. It’s ChatGPT. It’s that feeling of “Wow, this just works.”

There used to be a real tradeoff between power and usability. Buyers could credibly consider a more capable system that was harder to use, or a simpler system that was easier to navigate.

If the interface starts to look the same everywhere (i.e., chat windows), that tradeoff starts to disappear.

At that point, the question becomes pretty obvious, “Why wouldn’t you choose the system with the best underlying architecture and capabilities?”

The flip side of all of this is what it means for HR practitioners.

If AI agents can actually move through an HRIS and do real work - including things as complex as global payroll - that’s a pretty big moment.

And it’s reasonable for that to feel a bit uncomfortable.

There was a comment from MacInnis that stuck with me. “Your life is going to get easier, and then you’re going to start asking what your job actually is.”

My answer to that question is strikingly simple: My job is to find better work.

This is something I’m challenging myself to do in my business every day.

There are things - like building RFPs from requirements sessions - that used to take me days that AI can now do in a fraction of the time.

If my role is just to generate those documents, then yes, this is a scary moment.

But if I instead say, “That was low value work. What better work can I do now that it’s off my plate?” the feeling changes.

With the RFP example, we’re building a custom app that helps rethink the entire experience into a more delightful, dynamic experience. That’s genuinely fun!

If AI does something that you used to do, simply say, “Good, I'm going to find better work to do now.”

The constant process of finding better work to do is how you can help your organization, insulate your career, and move toward work that is more creative and less drudgery.

There are no shortage of reasons to be worried today, but I strongly believe in choosing to be an optimist.

And, I thought it was notable that both White and MacInnis, separately, in different interviews, gave me the exact same quote:

“What a time to be alive.”