In almost every leadership meeting I sit in now, there's an "AI moment." Someone has seen a demo, a competitor has announced something, or a board member has asked, "What's our AI strategy?" The pressure to do something is real. But doing something isn't the same as getting results.

In PwC's January 2026 global CEO survey, 56% of chief executives said AI had delivered no meaningful cost or revenue improvement over the previous year. The technology works. The problem is usually where it's being applied, and how.

After 25 years leading technology organizations, I've found that the most useful AI question isn't "Should we use AI?" It's "For each opportunity in front of us, should we build, buy, or wait?" Answer that honestly and you'll avoid most of the expensive mistakes.

Three Options, Clearly Defined

Buy means using AI that someone else has already built: the copilots and assistants inside Microsoft 365, Google Workspace, your CRM, your ERP, or a specialized SaaS product. You're paying for speed and someone else's R&D budget.

Build means connecting AI models to your own data and workflows to do something specific to your business. It doesn't have to mean a team of data scientists. Often it's a focused integration. But it does mean you own it, maintain it, and are accountable for it.

Wait means deliberately choosing not to act yet, because the data isn't ready, the process isn't stable, or the risk isn't understood. Waiting is a legitimate strategy. Drifting is not.

A Simple Scoring Model

For each AI idea on your list, score it from 1 (low) to 5 (high) on five factors:

Business value. Can you tie it to measurable dollars, hours, or customer outcomes? "It would be cool" scores a 1.

Data readiness. Is the data it needs accurate, accessible, and reasonably clean today — not after a future cleanup project?

Process stability. Is the underlying process well defined and consistently followed? AI amplifies whatever process it sits on, good or bad.

Differentiation. Is this part of how you win against competitors, or is it the same work every company does?

Risk tolerance. If the AI gets it wrong, is the consequence a minor annoyance (high tolerance) or a regulatory, financial, or reputational problem (low tolerance)?

Then apply three decision rules:

• Buy when differentiation is low and business value is real. Drafting emails, summarizing meetings, and searching documents are the same everywhere — don't build what you can license, and check whether you're already paying for it.
• Build when differentiation, value, and data readiness all score high. This is where AI can become a competitive advantage, because it's trained on knowledge and data only you have.
• Wait when data readiness or process stability scores low, or when risk is high and you don't yet have governance in place. Fix the foundation first.

What This Looks Like in Practice

A Composite Example

Consider a 140-person industrial distributor I'll describe as a composite of situations I've seen. The leadership team came in with a list of 22 AI ideas, collected enthusiastically from every department. Nobody knew where to start, so nothing was starting.

We scored every idea in a single 90-minute working session. Four landed squarely in Buy — and it turned out they were already paying for AI features in their productivity suite that nobody had turned on. One idea, a quoting assistant built on a decade of pricing and margin history, was a clear Build. The remaining 17 went to Wait, most because the underlying data lived in disconnected spreadsheets.

Within a quarter, the Buy items were in daily use at essentially no additional cost, and the quoting assistant had cut typical quote turnaround from two days to the same day. Just as important, the Wait list became a data-cleanup roadmap instead of a source of guilt.

"Wait" Is a Strategy, Not a Failure

Many leaders treat waiting as falling behind. Done well, it's the opposite: it's preparing to move faster than competitors who rushed in.

Waiting well means:

• Naming the specific blocker (data, process, or risk) and assigning an owner to fix it.
• Putting a lightweight AI governance policy in place so the next decision is faster.
• Running small, time-boxed experiments to learn without committing budget.
• Revisiting the list quarterly, because the vendor landscape changes fast and today's Build may be next year's Buy.

Three Traps to Avoid

Paying for AI you don't use. AI add-ons are being bundled into nearly every software contract. Audit what you already have before buying more.

Building what a vendor will ship in six months. If the capability isn't differentiating, a major platform will likely offer it soon. Your build becomes technical debt.

Waiting as avoidance. If "wait" has no owner, no blocker, and no review date, it's just a decision you haven't made.

Where to Start This Week

Gather your leadership team and list your top ten AI ideas. Score each one on the five factors. Aim to leave the room with one Buy you can activate now, one Build worth scoping, and a clear list of what you're waiting on and why.

That alone puts you ahead of most organizations, which are still debating whether to start.

If you'd like an experienced, independent perspective on where AI fits in your business, I help leadership teams run exactly this exercise and turn it into a practical roadmap. Schedule a complimentary Discovery Call to get started.