What Leaders Wish They Knew Sooner

As AI agents go beyond the hypothetical and enter actual workflows, many leaders see a gap between the promise and the reality. Are the agents ready? Moreover, are the humans? At the 2026 MIT Sloan CIO Symposium, we sought expert perspective and advice. We asked technology and business leaders, “What have you learned this year about humans and agentic AI working together?”

What came back wasn’t a technology story but a management story.

Thomas H. Davenport, a professor at Babson College and a fellow at the MIT Initiative on the Digital Economy, struck a cautionary note. He warned that human-in-the-loop oversight of AI tools is becoming performative: “People are being pestered to approve things rapidly, so they don’t really have a chance to engage.” He worries that most humans simply won’t want to serve as auditors of what AI is doing and asserts that no amount of policy will easily fix that.

George Westerman, a principal research scientist and senior lecturer at the MIT Sloan School of Management, said that agents “are not really ready for prime time in most organizations.” He noted that the word agent is being slapped on things that aren’t that sophisticated yet — inflating expectations without delivering value. His advice: Automate where it makes sense, not where it’s easy, and rebuild processes around the desired outcomes.

Watch the video for more lessons from the leaders in the room, including:

• Micro-agents and the trust fabric. Find out how one organization evolved from placing humans at every step to placing them at the right steps.
• In the loop versus on the loop. Acknowledge the fundamental split between agents that execute tasks and agents that clarify what you actually want.
• Building trust gradually. Move from small experiments to full deployment, just as a new driver moves from local roads to the highway.

Video Credits

Abbie Lundberg is the editor in chief at MIT Sloan Management Review.

M. Shawn Read is the multimedia editor at MIT Sloan Management Review.