Okta’s president and COO has an agent on his team. He names them Leo, Sloane, Hank, and Walker, and they appear in Business Review with his human staff. He personally booked a flight to Bangalore and spent the entire trip launching an open source agent on a separate machine, which he then assigned to every member of his leadership team as a deliberate act of immersion. “This flight was transformative for me in terms of realizing what the capabilities of this technology are,” he said at the COO Summit this week.
Still, the hardest part isn’t the technology, he added. It’s the managers.
“We’ve trained all of our managers around the world to think about how many employees they have,” Kelleher said. “Our managers have spent decades learning how to think about headcount and payroll.” The changes he’s advocating for at Okta, forcing managers to explicitly budget for both human and digital workforces and consider rosters that include AI agents as true colleagues, are “a much harder problem than getting people to experiment with Claude code,” he said.
“One of the things I really advocate within Okta is getting managers to think about how they design work to include human and digital workers,” Kelleher told a room of executives. “Everyone has a mission. [to adopt AI]”The other thing that I’m most concerned about right now is that we’ve trained all managers around the world to think about one thing when they go into a budget plan or cycle: How many people do we have?” What does the organizational chart look like? Who reports to whom? How many layers are there? How is this scope controlled?” That idea doesn’t apply right now, he added.
This was suitable for sessions hosted by Cognizant. New jobs, new worlds: How AI is reshaping the organizational charthead of research Ollie O’Donoghue and chief business officer for AI Sushant Warikoo explore this topic.
Kelleher’s comments crystallize a growing frustration among executives, where companies largely understand how to experiment with AI, but remain in collective denial about how to actually redesign work with AI.
From number of employees to “work plan”‘
Kelleher’s proposed solution is deceptively simple. The idea is to stop thinking about work in purely human terms. His fix? Push the token budget down to the HR manager. The idea is to specifically calculate the workforce, including AI agents working alongside human employees, and visualize those trade-offs in the budget itself. “What we want to know is how rotas evolve when digital workers are working alongside human colleagues,” he said. The current debate is too focused on AI displacing jobs, he said, and “the nature of the jobs themselves hasn’t changed.”
Kelleher’s remarks came as Cognizant released new research showing that the AI revolution is moving much faster than anyone expected, yet its value is not being realized. In 2023, the company predicted that 90% of jobs would be destroyed by AI by 2032. Today, that number is already at 93%, six years ahead of schedule. However, the expected productivity gains that followed did not materialize.
O’Donoghue described this as the “activation gap,” or the gap between what AI can theoretically do and what companies are actually achieving. “There’s some disconnect between theory and reality,” O’Donoghue said, citing an analysis of 80,000 different tasks carried out in each of the past three years. “Ninety percent of the tasks we analyze still require some form of human involvement.”
The issue of organizational redesign is therefore becoming more urgent, not less. If humans are still involved, the question is not whether to replace them, but how to restructure their role around machines that can now perform the transactional part of the job.
more difficult management problems
Several executives in attendance explained that they are trying to solve this problem from various angles. John Blotner, president of Wayfair, said the company pivoted from top-down AI mandates and instead gave all employees access to Claude, Gemini and ChatGPT, and then watched teams begin to reinvent their roles. “We’re seeing people reinventing their jobs and saying, look, we’ve basically automated the job,” he said. “That person is very precious.”
Cognizant’s Warikoo agreed that was the unglamorous core of the problem. “Humans and agents have equal privileges,” he said. “But the entire enterprise architecture is built around the notion that humans approach business workflows using static application architectures.” AI agents require persistent context and operate continuously. This is a fundamentally different model than the temporary, batch-driven systems that companies have built.
“This is not an AI issue,” Warikoo said. “At the end of the day, it’s about people. It’s about expanding human potential and enabling humans to do more high-value work.”
In Kelleher’s assessment, most organizations are not there yet. Today, the instinct is still to think of digital workers in the same way that companies once thought of software: as tools for employees to use, rather than as a category of labor that should be managed, budgeted for, and integrated into organizational charts alongside people.
“I can see the future now,” Kelleher said. luck “And it’s clear to me, we can’t go back,” he said on the sidelines of a panel discussion.The turning point for him was a confrontation with a staff member when he asked them for the name of an OpenClaw agent. “In that exercise, AI became a colleague rather than a tool, and the impetus was valuable.” It’s similar to the introduction of electricity, he agreed, when entire factories realized late that they didn’t need old steam engines anymore. He said this is similar to how current AI deployments are “like asking people to add chatbots.”
Later that afternoon, Kelleher told other executives that his team was starting to realize that digital agents were colleagues of sorts. “It’s very uncomfortable, but it’s very transformative.”
“We are evolving from workforce planning to work planning,” Kelleher said at the venue. “What I’m realizing now is that this is a really big leap for people.”
