The McKinsey report One Year of Agentic AI: Six Lessons from the People Doing the Work distills insights from more than 50 deployments of agentic AI, highlighting both opportunities and challenges. Agentic AI refers to systems built on foundation models that can act in the real world and execute multistep processes, but success depends less on the agent itself and more on redesigning workflows. The six key lessons are:
- Focus on workflows, not agents – value comes from reimagining end-to-end processes, not just building impressive agents.
- Agents aren’t always the answer – simpler tools like rules-based automation or predictive analytics may often be better suited.
- Stop “AI slop” – invest in rigorous evaluations, continuous feedback, and trust-building with users.
- Track and verify every step – embed monitoring to catch errors early and refine agent performance.
- Prioritize reuse – develop reusable agent components to avoid redundancy and reduce wasted effort.
- Keep humans essential – people remain vital for oversight, compliance, judgment, and edge cases, with workflows redesigned for effective human–agent collaboration.
The report concludes that while early missteps are normal, companies that treat agent deployment like workforce development—combining training, evaluation, and workflow redesign—are more likely to realize productivity gains and sustained adoption
