What I Tell Every CEO Who Asks Me About AI

TL;DR: When a CEO asks me about AI, I give them five things: start with one workflow, demand results in 90 days, insist on industry depth, budget for ongoing intelligence (not a one-time project), and ignore anyone who says AI will solve everything.

I get the same call about twice a week. A CEO — usually mid-market, usually running a $20M to $200M company — reaches out because they know they need to "do something with AI" but they don't know what.

The call usually starts the same way: "We're behind on AI and I don't know where to start."

Here's what I tell them. Every time. Regardless of industry.

1. Start With One Workflow, Not a Strategy

Every consulting firm in America will sell you an "AI strategy." They'll spend three months interviewing your team, produce a 60-page deck with a roadmap, and charge you $150,000. Then the deck sits on a shelf because nobody knows how to implement it.

What I tell CEOs: Pick one workflow. The one that's most painful — the one your team complains about, the one that causes errors, the one that takes too long. That's your first AI project.

Examples I've seen work:

• A legal firm started with document review (their paralegals were drowning)
• A construction company started with cost estimation (their bids were taking too long)
• A healthcare practice started with appointment scheduling (their no-show rate was 28%)
• A logistics company started with route planning (their fuel costs were climbing)

One workflow. Get AI working there. Measure the result. Then expand.

Why this matters: The companies that succeed with AI start narrow and go deep. The companies that fail start broad and go nowhere.

2. Demand Results in 90 Days

If someone tells you it'll take 12-18 months to see results from an AI implementation, they're either building something too complex or padding the timeline.

What I tell CEOs: Any AI project in the mid-market should show measurable results within 90 days. Not "the platform is built" — measurable business results. Fewer errors. Faster processing. Better conversion. Reduced costs.

If the vendor can't commit to a 90-day outcome, they're selling you R&D, not a solution.

The exceptions: If you're in a heavily regulated industry (healthcare, finance) where compliance review adds weeks to every deployment, 120 days is reasonable. But 18 months never is.

3. Insist on Industry Depth

This is where most AI projects fail. A vendor sells you a "general purpose AI platform" and then discovers — after you've signed the contract — that they don't understand your industry.

The AI that works for a law firm doesn't work for a hospital. The AI that works for a retailer doesn't work for a manufacturer. Not because the technology is different — because the context is different. The terminology is different. The regulations are different. The workflows are different. The buyer psychology is different.

What I tell CEOs: Ask your AI vendor three questions:

1. "How many companies in my specific industry have you deployed for?"
2. "Can you name the top 3 compliance frameworks that affect my industry?"
3. "What's the biggest AI implementation failure you've seen in my sector, and why did it fail?"

If they can't answer all three without hesitating, they don't have industry depth. They have a horizontal product they're going to try to wedge into your vertical.

4. Budget for Ongoing Intelligence, Not a One-Time Project

This is the one CEOs get wrong most often. They think of AI as a project with a start date and an end date. Build it, deploy it, done.

AI is not a project. AI is a capability.

A well-built AI system gets smarter over time. Every invoice it processes teaches it something about your billing patterns. Every customer interaction teaches it something about your service flow. Every scheduling decision teaches it something about your capacity.

But that only works if the system is maintained, monitored, and evolved. If you treat AI as a one-time project and walk away, you'll have a system that works great in month 1 and degrades by month 6.

What I tell CEOs: Budget for AI the way you budget for marketing — as an ongoing capability with a monthly cost and a monthly return. Not as a capital project with an end date.

The right question isn't "how much does AI cost?" The right question is "what's the monthly cost and what's the monthly return?"

5. Ignore Anyone Who Says AI Will Solve Everything

If a vendor tells you AI will "transform your entire business," run. If a consultant tells you that AI will "fundamentally reshape your industry," ask them to name one specific thing it'll do for you by next quarter.

AI is extremely good at specific, well-defined tasks. It's terrible at vague, open-ended mandates. "Use AI to improve our business" is a recipe for a failed project. "Use AI to reduce our document review time by 50%" is a recipe for a successful one.

What I tell CEOs: AI is a tool. A powerful tool. But a tool. It doesn't have a strategy. It doesn't have judgment. It doesn't understand your company's culture or your customer relationships. What it does — very well — is handle repetitive, data-heavy tasks faster and more accurately than humans.

That's not "everything." That's enough.

The Conversation I Wish More CEOs Would Have

The best conversations I have with CEOs aren't about AI at all. They're about operations. "Where are we slow? Where do we make errors? Where are our people spending time on work that doesn't require their judgment?"

AI is the answer. But the question has to come first. And the question is always about the business — not the technology.

If you're a CEO reading this and you've been putting off "the AI conversation" — start there. Not with vendors. Not with strategy decks. With your own operations team. Ask them what's painful. Write it down.

Then call someone who builds AI systems for your specific industry, and ask them: "Can you fix this in 90 days?"

If the answer is yes, you've found your starting point.