Blab AI
MIT Sloan CIO Symposium Highlights Challenges of Deploying AI Agents in Corporate Workflows
On May 19, the Royal Sonesta Hotel in Cambridge, Massachusetts, hosted the 2026 MIT Sloan CIO Symposium, a gathering that has drawn technology leaders and academics alike since 2004. The event’s theme—"Human‑AI Synergy: Redefining Leadership for a Transformative Future"—set the stage for a candid look at how AI agents are being woven into everyday business processes.
A short video produced by the MIT Sloan Management Review captures a panel of industry and academic heavyweights discussing the gulf between the promise of agentic AI and the realities of its deployment. The panel included Thomas H. Davenport, a professor at Babson College and a fellow at the MIT Initiative on the Digital Economy, and George Westerman, a principal research scientist and senior lecturer at the MIT Sloan School of Management.
Davenport warned that human‑in‑the‑loop (HITL) oversight is becoming performative. "People are being pestered to approve things rapidly, so they don’t really have a chance to engage," he said. He cautioned that employees are expected to sign off on AI‑generated outputs without enough time or context to assess quality or fairness. According to Davenport, policy alone will not solve the problem; organizational culture and workflow design must evolve to give auditors meaningful engagement.
Westerman offered a complementary view, arguing that "agents are not really ready for prime time in most organizations." He explained that the term agent is often applied to tools that are not yet sophisticated enough to act autonomously, inflating expectations without delivering real value. Westerman advised that companies should automate only where it makes sense, not where it is easy, and that processes should be rebuilt around desired outcomes rather than around the technology itself.
The panel also explored practical strategies for building trust in AI systems. One speaker described a "micro‑agents" approach, in which small, narrowly scoped agents are introduced gradually. The idea is to move from placing humans at every step to placing them at the right steps, allowing teams to build confidence in the system’s outputs over time.
Another key point was the distinction between "in the loop" and "on the loop." The former refers to agents that execute tasks directly, while the latter involves agents that clarify what the human user actually wants. Understanding this split helps organizations design interfaces that keep humans in the decision‑making chain without overburdening them.
The video concludes with a reminder that trust is built incrementally. Westerman likened the rollout of AI agents to a driver learning to navigate from local roads to a highway: small experiments should precede full deployment, and success metrics should be tracked closely.
Production credits for the video list Abbie Lundberg as editor‑in‑chief and M. Shawn Read as multimedia editor for MIT Sloan Management Review. The symposium itself has been an annual gathering of CIOs and technology leaders since 2004, and the 2026 edition continued the tradition of bringing together academic insight and industry practice.
In sum, the 2026 MIT Sloan CIO Symposium underscored that while AI agents are increasingly present in corporate workflows, both the technology and the people who manage it are still catching up. The event highlighted the need for thoughtful oversight, realistic expectations, and gradual scaling to achieve reliable, trustworthy AI integration.
A short video produced by the MIT Sloan Management Review captures a panel of industry and academic heavyweights discussing the gulf between the promise of agentic AI and the realities of its deployment. The panel included Thomas H. Davenport, a professor at Babson College and a fellow at the MIT Initiative on the Digital Economy, and George Westerman, a principal research scientist and senior lecturer at the MIT Sloan School of Management.
Davenport warned that human‑in‑the‑loop (HITL) oversight is becoming performative. "People are being pestered to approve things rapidly, so they don’t really have a chance to engage," he said. He cautioned that employees are expected to sign off on AI‑generated outputs without enough time or context to assess quality or fairness. According to Davenport, policy alone will not solve the problem; organizational culture and workflow design must evolve to give auditors meaningful engagement.
Westerman offered a complementary view, arguing that "agents are not really ready for prime time in most organizations." He explained that the term agent is often applied to tools that are not yet sophisticated enough to act autonomously, inflating expectations without delivering real value. Westerman advised that companies should automate only where it makes sense, not where it is easy, and that processes should be rebuilt around desired outcomes rather than around the technology itself.
The panel also explored practical strategies for building trust in AI systems. One speaker described a "micro‑agents" approach, in which small, narrowly scoped agents are introduced gradually. The idea is to move from placing humans at every step to placing them at the right steps, allowing teams to build confidence in the system’s outputs over time.
Another key point was the distinction between "in the loop" and "on the loop." The former refers to agents that execute tasks directly, while the latter involves agents that clarify what the human user actually wants. Understanding this split helps organizations design interfaces that keep humans in the decision‑making chain without overburdening them.
The video concludes with a reminder that trust is built incrementally. Westerman likened the rollout of AI agents to a driver learning to navigate from local roads to a highway: small experiments should precede full deployment, and success metrics should be tracked closely.
Production credits for the video list Abbie Lundberg as editor‑in‑chief and M. Shawn Read as multimedia editor for MIT Sloan Management Review. The symposium itself has been an annual gathering of CIOs and technology leaders since 2004, and the 2026 edition continued the tradition of bringing together academic insight and industry practice.
In sum, the 2026 MIT Sloan CIO Symposium underscored that while AI agents are increasingly present in corporate workflows, both the technology and the people who manage it are still catching up. The event highlighted the need for thoughtful oversight, realistic expectations, and gradual scaling to achieve reliable, trustworthy AI integration.
