Maria Black, only the seventh CEO in ADP’s 77-year history, often lingers before her predecessors’ portraits on the walk to her office. They’re a reminder of how ADP’s former leaders navigated previous moments of disruption, from financial crises to the dot-com bust. And when she’s facing a particularly tough conundrum, she says, “I go look at the oil paintings.”

Black, the self-styled historian of the $110 billion, 67,000-person HR software giant, says understanding a company’s past helps protect the best of its culture at times of heightened change. ADP’s hold on the payroll processing market, with more than 1 million clients, provides it with a distinct view into the current disruptive era and the heavily debated question of how much AI is reordering the workplace.

ADP has been tracking AI-exposed occupations, and has come to the conclusion that the technology isn’t wiping out entire careers but revaluing individual tasks within jobs instead.

But the shifts in employment wrought by AI are challenging enterprises to adapt to more dynamic teams, assessing people based on their skills rather than their place in a hierarchy, Black says. She’s optimistic about that prospect, but frustrated that fear-stoking predictions of an “AI job-pocalypse” are distracting from more urgent conversations about how to prepare the workforce for the changes that are now hitting it.

ADP’s payroll insights, meanwhile, have made it a valued source for investors seeking to understand the US labor market after the Trump administration’s cuts to the US Bureau of Labor Statistics raised questions about the reliability of official data. Black fields such questions diplomatically, saying their numbers are “both valid.”

Here’s how Black describes ADP’s insights into “a defining moment for human capital management.”

This interview has been edited for clarity and length.

Andrew Edgecliffe-Johnson: ADP and the Stanford Digital Economy Lab produce a “Canaries Dashboard,” showing where AI is and isn’t changing the world of work. What story is it telling you?

Maria Black: The early indicators are that jobs are not being reduced or going away; rather, it’s an unbundling of the tasks within jobs. If you look at IT jobs, what you see is that the low-value tasks are the ones that are ultimately being disintermediated by AI. Tasks that are list- and task-oriented are the ones that are getting disrupted. But the things that actually require judgment and complexity create more value. So ultimately employers are willing to pay more for jobs that include human complexity and judgment. So that, to me, is optimistic, but it’s also opportunistic, because it shows that there is room for both technology and humans.

Do you think we’re now at the point where we can say that the fear of an AI job apocalypse is overdone?

What I would suggest is that it’s never really been the right conversation. Every innovation cycle, every technology cycle has had an impact on jobs. Perhaps this one is the biggest of them all thus far. I don’t argue that. [But] the conversation quickly went from how naturally this continues to evolve the world of work and somehow took on a life of its own. I don’t want to sit here and judge how all those fears came about. But I wish we would have spent the last three years getting the world of work ready for upskilling, reskilling, converging, education, all of the things that I think will make this world, candidly, more human than it’s ever been.

Your private payroll data has always been looked at alongside federal data, the BLS payrolls. Are we seeing wider discrepancies between those two series at the moment than we used to?

Only time will tell. Partly because the BLS generally goes back and adjusts, and that’s not meant as a dig or a slight. In my mind, they are two completely different indexes. I think they’re both valid. BLS is survey-based. We are real-time-based. If you look at them in hindsight over time, they actually track pretty closely together.

Do you think that the official BLS data is now lower quality than it used to be?

I don’t know that I’m in a position to speak to that. I’ll leave that to the economists. I just think it’s a different measure, and I think it has its own value.

You call yourself ADP’s historian. What’s the value of studying history?

Every year, I teach our new executive management program the history of ADP through the lens of each CEO’s chapter. I’m fascinated by things that those before me did during heightened times of innovation, like [the advent of] the internet, or heightened times of economic disruption, such as the dot-com bust or the great financial crisis. [History provides] a bellwether on who you want to be and what it is that you need to protect. And I think the biggest thing that all of us as CEOs need to protect, assuming you have a good one, is the very culture of your company. Because in heightened times of innovation and transformation, you’re only as good as the culture. So [studying history] is informative for current decision-making. It gives you respect for where you’ve been. And it forces the accountability of protecting the things that are good.