When Brian Crofts joined BambooHR, the mandate was not simply to move faster.

BambooHR had already built one of the most recognizable products in the mid-market HR software category. The company was known for ease of use, a clean employee experience, and a product that made HR systems feel more approachable than many of the alternatives in the market.

But the next chapter required something broader.

“I think the clearest mandate to me was kind of going from maybe a single product to multi-product to platform,” Brian said.

That shift had already started before his arrival. BambooHR had acquired payroll capabilities years earlier, then spent time bringing HR, payroll, and benefits closer together into a more native experience. In Brian’s words, the work was less about “syncing of data” and more about building “integrative, native payroll HR, Ben admin experience.”

That platform work matters more now because the market has entered a very different phase. BambooHR is no longer just thinking about expanding modules or improving workflows. The next product era is being shaped by AI agents, autonomous systems, and a fundamentally different expectation of what HR software should be able to do.

“Now I think we’re in that platform AI era,” Brian said.

The roadmap becomes the AI strategy

One of the most interesting parts of the conversation was how Brian described the responsibility BambooHR feels from its customers.

Most HR teams are not trying to DIY their own AI transformation - building their own agents, stitching together foundation models, or deciding which technical architecture will still matter six months from now. They want their core systems to help them make sense of the shift.

Brian said BambooHR hears that directly from customers.

“Your roadmap is our AI strategy,” he said.

That line says a lot about where the market is right now. HR leaders are under pressure to explain how they are using AI, but many of them do not want to own the technical experimentation themselves. They want vendors they already trust to absorb the complexity and translate it into usable products.

Brian described BambooHR’s own AI transformation in three parts. First, how the company changes the way it works internally. Second, how it changes the way software gets built. Third, what BambooHR actually ships to customers. The product is the visible part, but the internal operating model has to change too.

“The way we approach the work has to be totally different,” he said. “The way we think about value exchange is different.”

Customers are asking BambooHR to help define what AI should mean for HR operations.

Building while the ground is moving

The hardest part of building in AI right now is that the ground does not stay still for very long.

A few years ago, RAG (retrieval augmented generation) felt like the center of the conversation. Then attention moved to agents. Then MCP. Then new model capabilities, new orchestration patterns, and new assumptions about what software should be able to do.

So I asked Brian how a product leader is supposed to build confidently in a world where the foundational technology keeps shifting.

His answer was one of the more thoughtful moments of the interview.

He described the need to operate with “loosely held” assertions. Not uncertainty for its own sake, and not overconfidence disguised as strategy. Something in between.

“[I need to say], ‘here’s my belief, here’s an assertion that I have that’s loosely held,” he said. “In this point in time, if this is how we see the future, this is how we think it’s gonna happen. And based on that, this is how we’re building.”

The important part is that those beliefs have to be tested constantly. Brian talked about the speed at which teams can now build prototypes, put them in front of customers, and learn whether their assumptions hold up.

“I think it’s about the velocity of learning versus maybe even the velocity of shipping new product,” he said.

That feels like the right distinction. For a long time, software companies were rewarded for shipping more. More features, more workflows, more modules. In this environment, experimenting at the frontiers of this new horizon may be the most important quality.

Brian summarized the mindset with an internal phrase: “learn at all, not know at all.”

It is a simple line, but it captures something real about this moment. The companies that pretend to know exactly where AI is going may look confident in the short term. The companies that stay curious and humble may end up being right more often.

When the interface matters less

BambooHR has historically had a real advantage in the market because of its user experience. For many buyers, especially in the mid-market, BambooHR has been the approachable option in a category that can often feel bloated or intimidating.

That raises an interesting question in the AI era.

If more work starts happening through chat interfaces, Slack, embedded agents, or other conversational experiences, does the value of a great UI compress?

Brian did not dismiss the concern. He said BambooHR talks about it often.

“I do think that one of our assertions is that the interface will continue to fragment,” he said.

That does not mean UI stops mattering. Power users will still need full systems. HR admins will still need places to configure, review, govern, and understand what is happening. But the everyday employee experience may look very different.

If an employee can ask for a pay stub in Slack, they may not care where the answer technically comes from. If a manager can start an action from a chat interface, they may never think about the underlying system at all.

But that creates a new problem.

The software still has to know what should be allowed.

Brian gave a simple example. An employee might be able to ask for their own pay stub in another interface. But can they approve payroll? Can they create a payroll exception? Can they ask for their manager’s salary? Of course not.

That is where BambooHR’s trust layer becomes more important.

“Do I trust that this system is going to behave accordingly in an autonomous way where it understands and respects execution authority?” Brian said.

That phrase, execution authority, is a useful way to think about where HR software is heading. In a traditional system, the UI helped constrain what users could do. In an agentic system, the controls have to sit deeper. Permissions, roles, compliance, governance, and data structure become more central because the user may not always be operating inside the native interface.

Brian put it plainly: “That’s what I think customers will be paying for is not just a nice UI.”

What frontier labs may miss about HR

The conversation eventually turned to frontier AI companies and whether they might start building into categories like CRM, HRIS, or ERP.

Brian’s view was nuanced. He did not dismiss the possibility. In fact, he said it would be naive to assume foundation model companies will not continue verticalizing into valuable software categories.

“They’ll look at CRMs, they’ll look at HRS, they’ll look at ERP systems, they’ll look at all of it,” he said.

But he also pointed to the things that are easy to underestimate from the outside.

Payroll is a good example. From a distance, it can look like a calculation engine with a user interface on top. In practice, it is deeply tied to money movement, tax compliance, payroll compliance, employee records, government filings, permissions, and error handling.

“They quickly learn that it’s not just about a simple UI that does a complex, or fairly complex calculation from gross to net on the back end,” Brian said. “It’s a very robust compliance driven system.”

That does not mean HR software companies are immune from disruption. But competing against HRIS companies is not just shipping a software platform. It will involve maintaining the system, hardening it over time, handling edge cases, respecting governance, and earning trust in processes where mistakes have real consequences.

That is why the conversation around AI replacing software often feels too simplistic. Intelligence is only one part of the system. Execution, compliance, maintenance, and accountability matter just as much.

Paying for outcomes instead of software

The most forward-looking part of the conversation came near the end, when we talked about where SaaS business models may be heading.

For years, software companies tried to avoid services. Services were lower margin, harder to scale, and often handed off to consulting partners. The classic SaaS model was to sell the product and let someone else handle the implementation, administration, and ongoing operational work.

Brian sees that line starting to blur.

He described a world where BambooHR builds first-party agents inside the product, while also using agents to power services that help customers get work done directly. The two sides can teach each other. What BambooHR learns from delivering services can inform the product, and what it builds in the product can make services more efficient.

The end state is not just better software. It is work getting completed.

“Just take it, just do all the work for us and then we’ll pay you for payroll being done,” Brian said, describing what customers are increasingly asking for. He mentioned examples like payroll, onboarding, rolling out a new policy, or changing a compensation planning cycle.

That is a different value exchange than traditional SaaS.

Instead of buying access to tools, customers may increasingly pay for outcomes. Payroll completed. Employees onboarded. Policies updated. Workflows handled. The software becomes less of a place where work happens and more of the mechanism that ensures work gets done.

Brian summarized the direction well: “[As a practitioner], I don’t do the work, I just make sure I know the policy and I can orchestrate and I can kind of oversee everything, but the work is just happening.”

That may be one of the bigger shifts AI brings to HR software. The value of the system moves from helping humans click through the work to helping humans govern the work as agents execute more of it.

Closing

The impression I left with from the conversation was not that BambooHR has every answer already mapped out.

It was almost the opposite.

Brian seemed comfortable acknowledging that the market is changing too quickly for rigid certainty. What mattered more was the posture: build around trust, learn quickly, stay close to customers, and keep returning to the parts of HR software that will remain important no matter how the interface changes.

Good employee records will still matter. Compliance will still matter. Permissions will still matter. Payroll accuracy will still matter. Governance will still matter.

The interface may fragment. The workflows may become more autonomous. The pricing model may shift closer to outcomes. But the need for a trusted system underneath all of that does not go away.

In some ways, AI may make that trusted foundation more important than ever