Okta's COO says the hardest part of AI adoption isn't technology—it's redesigning how managers think about work to include AI agents as colleagues, not tools

Okta's President and COO Kelleher has named AI agents on his team (Leo, Sloan, Hank, Walker) and integrated them into business reviews; he assigned his leadership team to spend time building and experimenting with open-source agents as he did on a flight to Bangalore.

Kelleher argues managers must shift from budgeting purely for headcount to budgeting for both human and digital labor, treating AI agents as genuine colleagues on org charts rather than software tools—a change he frames as harder than getting people to experiment with Claude Code.

Cognizant research shows 93% of jobs are now projected to be disrupted by AI (up from a 2023 projection of 90% by 2032), yet productivity gains have not materialized; analysis of 80,000 tasks over three years found 90% still require human involvement in some way.

Wayfair reversed a top-down AI mandate and instead gave all employees access to Claude, Gemini, and ChatGPT, then observed teams redesigning their own roles—with employees discovering they had automated transactional parts of their work and remained valuable for higher-value tasks.