Most professionals treat AI like a search box. You type a question, read the answer, ask something sharper, and repeat until you get what you need. A recent piece from MIT Sloan Management Review argues that this habit, useful as it is, caps how much value you can get out of AI, and that the real prize sits one level up: designing a small team of AI systems that investigate a problem on their own, each from a different angle, then comparing what they find.
The article calls this "directing intelligence." Instead of one conversation, you set up several AI agents against the same company data, such as financial reports, sales records, or customer interviews, and give each one a different job, like applying a specific business strategy or comparing what shows up in interviews against what shows up in official strategy documents. A separate "orchestrator" AI then reviews all the findings and points out where the answers contradict each other.
That contradiction is the whole point. A single chat conversation tends to confirm whatever assumption you walked in with. A system of agents pointed at the same data from different directions is far more likely to turn up something nobody in the room had considered, like a company's biggest growth risk being that its entire competitive edge lives inside one senior expert who never once appears in the strategy plan.
This lines up with what the wider numbers are showing. A joint survey from MIT Sloan and Boston Consulting Group found that 35 percent of companies already use agentic AI in some form, and another 44 percent plan to adopt it soon. Yet many of those same companies admit they have no clear strategy for how to use it, which is exactly the gap this approach is trying to close.
It also explains a harder number. A widely cited MIT Media Lab study found that 95 percent of corporate AI pilots produce no measurable financial return. Gartner separately predicts that more than 40 percent of agentic AI projects will be cancelled by 2027, due to rising costs and unclear value.
Consulting firms are already moving this way. McKinsey has built internal AI platforms used by thousands of its own staff, and rival firms are doing the same, which suggests the largest players in the advice business see this as a real shift in how they work, not a side experiment.
For a manager or business owner, the lesson is not to buy more AI tools, it is to rethink how you use the ones you already have. Before handing a messy, recurring problem to AI, decide what different angle each "investigator" should take, and treat any surprising or uncomfortable finding as worth digging into rather than dismissing. The businesses that pull ahead will not be the ones chatting with AI the most, they will be the ones who learn to set it up properly and trust their own judgment on what it reveals.