Boards want an AI strategy by next quarter. Vendors promise one by next week. The headlines swing between Uber exhausting its entire 2026 AI budget in four months and research claiming 95% of organizations see no measurable return on generative AI. Somewhere in the middle sits the CEO of a $400 million manufacturer, a regional insurer, or a construction services firm. They know the next two years will look nothing like the last two, and they have no clear idea what to do on Monday.

Only 30% of CEOs are confident about revenue growth over the next 12 months, down from 56% in 2022, and most are struggling to convert AI spending into returns. Half of CEOs believe their job stability depends on getting AI right this year. In our conversations with leaders across fields like construction, manufacturing, healthcare, and software, CEOs arrive in one of three states. Each needs a different starting point, and treating them the same is one of the fastest ways to stall a transformation before it begins.

Type one: The CEO who knows what to do

This is by far the smallest group. A notable minority arrive ready to reimagine the entire company. More often, they have one clearly defined problem to solve. About one in five leaders who come to us can identify a specific bottleneck and put a cost to it.

A billion-dollar e-commerce company we work with is a good example. Its leadership team, many of them Amazon alumni, is technically sharp and knew exactly what it wanted to build. What it couldn't do was hire senior AI engineers fast enough without spending millions to compete for that talent. We embedded a dedicated pod of AI engineers alongside its own team, working through its roadmap on a steady cadence.

This type of CEO doesn't lack ideas, but capacity and speed. A partner brings both capacity and expertise. A team working at the forefront of AI every day knows which tools and practices are worth adopting, helping a leader with a clear vision move faster than hiring alone would allow.

What they need: The initiative with the highest return at the lowest cost and risk. Build it, prove it, and let it point to the next one.

Type two: The CEO who knows they need to do something

This is the largest group. They have already rolled out ChatGPT or Claude licenses, and employees can produce slide decks and contract summaries faster without changing existing workflows. What this CEO lacks is a map, not technology. Most companies have never documented how work actually moves through them, so we start with what we call "the art of the possible" conversations and business process mapping. Two questions do most of the work:

  • Where do multiple people do a similar job? Automating a task one person does rarely pays off. A department of ten doing the same work is a real opportunity.

  • Who is your best employee? Imagine another one who never sleeps, works every day of the year, and moves faster. What would that be worth?

From there, we plot candidate initiatives on a simple two-by-two matrix, weighing cost and risk against potential return. Most companies above $100 million in revenue have several initiatives worth $500,000 or more a year, and we target at least a 3x return in year one.

What they need: Map the work, find the repeated workflows, and pick the quadrant with low cost and high return.

Type three: The skeptic

The skeptic, to one degree or another, doesn't believe AI applies to their business. Sometimes the objection is about people ("our team has specialized knowledge no system could learn"). Sometimes it's about money, sharpened by every headline about runaway budgets. Both concerns are reasonable, but neither is a reason to sit still.

Skepticism is useful when it's aimed at the right question. AI works, but that doesn't mean it makes sense for every business or every process. The question worth asking is whether it can deliver results in your operation, at a cost you can see, without betting the company. A few tests can help answer it:

  • Pressure-test the specialized-knowledge objection. If your most experienced person quit tomorrow, you'd hire someone and teach them how the work is done. An AI system learns the same way. Whatever can be taught to a new hire can be built into a workflow, and whatever can't is exactly where that person's time should go.

  • Measure growth, not just cuts. "Doing the work with half the team" is the wrong frame for most mid-market companies. Ask what the business could do if it doubled revenue without adding headcount. The math is the same and it's a number a whole leadership team can get behind.

  • Run the experiment on the edge. Pick three or four people and a slice of customers, give them a new way of working, and measure the results against everyone else. The rest of the organization keeps running as it does today.

  • Judge it by your own numbers. Another company's case study proves little about yours. A small proof in your own operation settles the question either way.

What they need: One small, contained test that shows whether the numbers work in their operation.

What all three have in common

Whichever camp a CEO starts in, the work is not a technology problem, but a business process problem. Today's models can handle almost any well-defined workflow. Teams need to identify which workflows are worth reimagining and how. BCG's research finds the companies most likely to succeed treat AI transformation as a business-led strategy rather than an IT initiative.

The other constant is imagination. Most leaders are used to the technology they originally built their companies around, so it's hard to imagine a two-day bidding process taking just ten minutes. But it's the CEO's job to get the best ideas on the table. When reimagining workflows isn't a leader's expertise, the answer is to bring in people who have done it dozens of times. MIT's State of AI in Business 2025 research found that AI initiatives developed through external vendor partnerships reached deployment 67% of the time, compared with 33% for internally built tools.

We have three CEOs with three different starting points, but the same destination. Each wants a business that generates more revenue and margin per employee. If an initiative can't demonstrate a clear path to that outcome, don't pursue it. Keep looking until you find one that can.