DETROIT — Eric Kelleher, President and COO of Okta, urged business leaders to treat AI agents as colleagues and incorporate them into workforce planning during his remarks at the Fortune 500 Innovation Forum in Detroit on November 16–17, 2026. He said the hardest part of the AI revolution is not the technology but managers’ inability to redesign work to include AI agents as colleagues.

"One of the things I’m really advocating for within Okta is to get our managers thinking about how to design work to include human workers and digital workers," Kelleher said. He noted that most organizations still view digital workers as tools rather than a category of labor to be managed, budgeted for, and integrated into organizational charts.

Kelleher proposed allocating token budgets directly to people managers to force a concrete reckoning with a workforce that includes AI agents alongside human employees. At Okta, he has named AI agents Leo, Sloan, Hank, and Walker and includes them in business reviews with human staff. He cited a turning point when he asked his team to name their OpenClaw agents, saying, "In that exercise, AI became a colleague as opposed to a tool and that catalyst is valuable."

During a flight to Bangalore, Kelleher set up an open-source AI agent on a separate machine and assigned the same task to his leadership team. "That flight to me was transformative in how I recognized what the capabilities of this technology are," he said.

Supporting context came from Cognizant, which reported that 93% of jobs are already disrupted by AI—six years ahead of its 2023 projection that 90% would be disrupted by 2032. Ollie O’Donoghue, Head of Research at Cognizant, said humans remain involved in 90% of tasks, citing analysis of 80,000 tasks conducted each of the last three years. "There’s a bit of a disconnect between theory and reality," O’Donoghue said.

Sushant Warikoo, Chief Business Officer for AI at Cognizant, said enterprise architecture was built for humans using static applications, not for continuously operating AI agents requiring persistent context. "Humans and agents have equal privilege," Warikoo said.