Industry

Debate at COO Summit: Are AI Agents Colleagues or Tools — and What Leaders Should Do

At Fortune’s COO Summit, executives disagreed over whether AI agents should be treated as colleagues or merely workflow tools, with Okta’s Eric Kelleher calling agents colleagues and Cisco’s Francine Katsoudas rejecting that framing.

Debate at COO Summit: Are AI Agents Colleagues or Tools — and What Leaders Should Do

At Fortune’s COO Summit, executives publicly disagreed about whether AI agents should be regarded as colleagues or treated as tools. Okta President and COO Eric Kelleher named several AI agents — Leo, Sloan, Hank, and Walker — and said these agents have become "a colleague" rather than a tool, noting they sit alongside human staff in business reviews.

A few hours later, Francine Katsoudas, Chief People Officer at Cisco, pushed back, arguing that agents are "part of the workflow, but not a colleague."

Why the label debate matters less than organizational change

Observers say the argument over terminology distracts from a deeper problem. Research from Cognizant found that 93% of jobs have already been disrupted, a shift that occurred six years earlier than the company had forecast. Despite widespread disruption, promised productivity gains have not materialized — a shortfall Cognizant describes as an "activation gap." The gap, analysts argue, is not caused by the technology itself but by organizational inertia: org charts, budget cycles, and performance review systems were designed to ration human headcount and have not been reengineered for a workforce that increasingly includes AI agents.

Practical takeaway for leaders

Executives and analysts at the summit emphasized that whether an agent is labeled a colleague or a tool is secondary to the need for structural change. The urgent task for managers is to redesign organizational charts, budgeting processes, and performance-review cycles so the presence of AI agents can translate into actual productivity and strategic value, while managing the transition risks that the current frameworks expose.