We asked an unaided AI to build a B2B strategy—and discovered that its biggest weakness looked surprisingly human.

Recently, we conducted an experiment while studying the emerging surgical robotics market.  J&J’s new Ottava surgical robot caught our attention, so purely as a strategic exercise, we decided to develop a strategy for it. J&J isn’t a current client, and no one from the company participated. We simply thought it presented a fascinating real-world strategic challenge.

We first ran the situation through Scout™, an AI system we’ve been developing that incorporates the strategic processes we’ve built and refined through more than 25 years of consulting & teaching. Then we wondered: what would happen if we skipped that structure, opened a general-purpose AI and simply asked it to develop a strategy?  What we learned wasn’t what we expected.

Surprise #1: AI Is Pretty Good at Strategy

We expected the unaided AI to produce something superficial. Instead, it generated a wide-ranging, thoughtful strategic plan. It identified meaningful market issues, surfaced competitive considerations and proposed plenty of intelligent actions. For those of us who have spent our careers doing strategy, that was humbling.

AI can absorb enormous amounts of information, connect ideas quickly and generate plausible strategic alternatives in seconds. Give it access to market research, customer information and competitive intelligence, and its capabilities become even more impressive. But as we examined its recommendations more closely, something interesting happened.

Surprise #2: AI Has Learned One of Our Worst Strategic Habits

The plan contained roughly 20 different strategic initiatives. Most were reasonable. Some were very good. The problem was that almost everything made the list.

We’ve seen the same behavior from human teams for decades. Ask a group to conduct a SWOT analysis and the walls quickly fill with strengths, weaknesses, opportunities and threats. Ask which opportunities matter most, however, and things get considerably harder. One of Tom’s old bosses had a colorful term for these sprawling collections of ideas: a “dog’s breakfast.” AI appears perfectly capable of producing one too.

Strategy isn’t about identifying everything you could do. It requires deciding what matters most, where scarce resources should go—and what you are deliberately not going to do.

Surprise #3: AI Can Prioritize. But By What Criteria?

There’s an obvious response: couldn’t you simply tell AI to prioritize its list? Absolutely. And it will. But what criteria should it use? Decision-maker influence? Customer needs? Segment attractiveness? Ability to win? Economics? Organizational capabilities? Those aren’t simply prompting questions. They’re strategic questions.

More importantly, strategy isn’t a linear exercise in which you answer all those questions independently and ask AI to assemble the answers at the end. Each decision changes what comes next—and sometimes causes you to reconsider what came before.

You might develop an initial hypothesis about the stakeholders who matter most. Understanding their needs may change that hypothesis. Those needs may reveal segments you hadn’t considered. Evaluating your ability to win may cause you to reconsider which segments deserve priority.

Good strategy develops through a sequence of hypotheses, decisions and revisions.

A Good AI Strategist Should Know When to Stop

This realization has significantly influenced how we’ve built Scout. Rather than asking Scout to ingest everything and spit out “the strategy,” we’ve embedded the questions, frameworks and linkages we’ve learned through more than 25 years of real-world strategic consulting. Scout develops hypotheses at key stages, brings relevant internal and external information to bear, and then stops. That’s important because the humans need to decide.

And those stopping points aren’t only about human judgment. They’re also about organizational alignment. In a B2B company, marketing may have one view of the most important stakeholders while sales has another. Product may see one unmet need while R&D or operations sees something different. If those disagreements remain hidden until everyone is presented with a finished strategy, alignment becomes extraordinarily difficult.

A better process creates alignment along the way: Are these the stakeholders that matter? Do we agree on what they need? Are these the segments we should prioritize? Do we really have an ability to win here?  Once people align at one strategic juncture, AI can incorporate those decisions and move to the next. That’s how we’ve designed Scout: hypothesize, stop, decide, align, refine—and continue.

The Real Opportunity Is Human + AI

AI still contributes something extraordinary to that process. It can synthesize customer studies, competitive reports, sales information and internal documents while simultaneously accessing relevant external information. That insight buried on page 137 of the fifth research report doesn’t have to remain buried.

But the human contribution is equally important: choosing the criteria, challenging hypotheses, exercising judgment and building organizational commitment as the strategy develops.

That’s a very different model from asking AI, “What’s my strategy?” The better question may be: How can AI help us make each strategic decision better—while knowing when to stop and let humans make it? AI is becoming remarkably good at generating answers.

The most powerful strategic AI may be the one that also knows when it’s time to ask a question.