AI can handle payroll, product development, in-boarding, talent recruitment, and safety seminars – to an extent. But most of us would contend that humanity still plays a role in this type of enterprise administration. After all, HR pros navigate systems that were made by people, and the human touch, even in a massively automated age, is still important.

I heard a lot about this in a panel discussion at our Imagination in Action event this past April, at MIT. (Disclaimer: I help run the IIA events annually.) The panel, led by Michael Hayes of Practical.ai, explored some of the ways that HR and office work will go in light of the emergence of agentic swarms.

Along with Hayes, the panel consisted of Lisa Simon of Revelio Labs, Dirk Jonker of Crunchr, Michael Krebs, co-founder Goldstar AI, and AI consultant Amit Mohindra.

Toward the beginning, Simon talked about surveys of workers in this new economy.

“When we really zoom in on what people are saying, positively and negatively, skill development is actually positive,” she said. “So they feel better about skill development, but workload and culture, mostly workload, is the driver that pulls the sentiment down.”

“Everyone talks about AI, but the internal compliance and guidelines are still needed to catch up,” Jonker added. “The real traction is what we're going to see when we see multiple agents working together in networks. I think HR is starting to get there, but we're basically halfway, or maybe just getting halfway.”

Jonker referred to “three foundations” and challenges that AI has to address: scattered implementation, underinvestment in tech, and undeveloped capabilities.

“If you look at the people that lead HR organizations today, at headquarters, but also in the field,” he said, “if you look at their education, it never really included strategy, finance, analytics. So we have a bit of catch up to do on data technology and capabilities.”

“Very few people, at least in the past, joined HR to run spreadsheets or build models, let alone build agents,” he said. “So the journey to make HR more data-driven, more analytical, even more digital has been long.”

Mohindra suggested there’s an opportunity now for HR, if leaders can address it as an adaptive problem, poised to handle the ambiguity and uncertainty that exist.

“HR is a really rich environment for AI,” he said, “all kinds of AI, whether it's analytical, generative or agentic.”

Later in the presentation, Krebs expounded on how swarms can build capabilities.

“When you pair agents and swarms of agents with humans, it can be really powerful and really increase the output of a single employee, without increasing the quantity of hours that they're working,” he said.

Swarms of agents, he asserted, can handle deterministic tasks and atomic units of work, leaving humans free to work with the big picture.

“What it really does is it distills the most important things to the human beings at the center of these swarms of agents,” he explained of this new way of using LLMs.

“It's very carefully programmed to match exactly what that person's output would be, right?” he said. “To match their preferences and everything. So it's not just typing things into chatGPT and getting 10 different responses 10 different times and being frustrated with the response.”

“Mentoring and the experience of having mentored people is really having a big moment,” Simon said, addressing the role of leadership in the AI age.

“We’re debating: will AI replace humans? But for sure, AI has exposed leaders who never actually emphasized or tried to maximize human potential in the first place. So, you know, the traditional form of leadership, this transformative leadership where the leader has a vision, a compelling vision, they get people to follow them … the leader really needs to become an adaptive leader. And adaptive leaders help their organizations learn together through uncertainty. They mobilize them in a particular way. The only way that organizations are going to thrive with AI is if the leaders help the organizations learn faster than the technology develops.”

That’s a mouthful. But I think it’s appropriate to describe how this form of leadership works, right now, as humanity is at a sort of crossroads with AI.

Jonker added that vision, and the consolidation of vision, is centrally important.

“If you think about big companies, then there is an island of HR, an island of finance, an island of business operations, and the real power of AI is to connect these different islands,” he said.

Back to the role of swarms, Krebs came back to the idea that humans will use LLMs to “not sweat the small stuff” in tomorrow’s world.

“It's distilling the most important strategic high-cognitive tasks to the human, because it's funneling all of those down, while the agents are handling the more repetitive tasks, “ he said.

But then, as Hayes and the rest talked through this, the awareness emerged that this top-level leadership, all the time, might make the humans exhausted.

“It's distilling the most important strategic high-cognitive tasks to the human, because it's funneling all of those down, while the agents are handling the more repetitive tasks.

“Who would have guessed that it turns out, as we spend our days over the last decades at work and we're doing sort of hard thinking things … that we actually needed the more mindless work to give our brains sort of a break during the day. We never thought of that as a break, but it sounds like we're sort of learning a little bit.”

Indeed, all of us are learning, in an era that lays bare the real values and objectives of human societies contending with what some now call “God-like AI.”

Here’s a bit more at the end of the talk, where each panelist came up with his or her own productivity metric.

“I think this focus on productivity, this obsession on productivity is going to come back and bite us at some point,” Mohindra said. Bringing in his own prized metric, he mentioned something new called “zeitgeist,” not spelled traditionally like the German “zeitgeist,” which is, in his view, a measure of human comfort with AI.

Jonker pivoted to the idea of money, suggesting that parties invest in AI to generate more free cash flow, and suggesting that we measure AI that way.

“I have personally never seen a productivity metric that I think fully captures what we're really trying to measure,” Simon said at her turn, finally going with “revenue per person” as a good yardstick.

Krebs suggested token usage per employee, as a metric, to see how people are taking advantage of the technology.

This was an array of useful angles, and the talk, in general, struck me as an essential debate around workplace implementation of AI. This is certainly the time to have that talk. Stay tuned.