How to Know If AI Is Actually Right for Your Business

Every week there’s a new headline about AI transforming some industry, and every week a business owner asks me some version of the same thing: should we be doing something with this? Usually there’s a little worry underneath it, the sense that everyone else already figured this out and they’re falling behind.

Here’s the honest answer. Sometimes yes, sometimes no, and telling the difference isn’t complicated.

Think about a table saw. Nobody buys one because their neighbor got one and won’t stop talking about it. You buy one because you’ve got a lot of wood to cut, over and over, and doing it by hand is slowing you down. AI is the same kind of tool. The question was never “is AI powerful,” it’s “do I actually have the wood to cut.” Plenty of businesses don’t, and there’s no shame in that.

It’s also worth knowing that “everyone’s doing it” and “everyone’s succeeding at it” are very different claims. McKinsey found 88% of organizations now use AI somewhere, but only about a third have scaled it. And S&P Global Market Intelligence reported that the share of companies abandoning most of their AI projects climbed from 17% to 42% in 2025. The gap between those numbers is the whole game. Adopting AI is easy. Getting anything out of it is a discipline, and it starts with knowing whether you should be doing it at all.

The question to ask before anything else

Before you think about tools or vendors or models, ask this: do you have a task that happens over and over, follows predictable rules, and right now needs a person to do it?

If yes, that’s a candidate for automation. If no, if the work is genuinely judgment-heavy, built on relationships, or different every single time, AI probably isn’t your lever.

Sounds obvious. But most of the AI projects that fail skipped this exact question. Someone heard about a tool, got excited, and started building before they could name the problem they were solving. That’s not the technology letting them down. It’s a framing failure, and no fancier model fixes it.

Three signs AI is a good fit

1. You’ve got a bottleneck made of repetitive work. Processing documents, entering data, handling intake forms, routing customer requests, summarizing long threads, tagging and sorting content. If your team spends hours a week on something that follows a consistent pattern, that’s automatable. It’s also where the upside is biggest. McKinsey figures generative AI could automate the activities that take up 60 to 70 percent of employees’ time, and most of that is exactly this kind of pattern-following work.

2. A mistake isn’t the end of the world. AI isn’t perfect. It makes errors. If those errors are easy to catch and cheap to fix, that’s a fine risk profile for automation. If a mistake would be catastrophic, think medical decisions, legal filings, moving money without review, you need a person in the loop, full stop.

3. You’ve got enough volume to make it worth building. If a task happens twice a week, automating it probably isn’t worth the effort. If it happens dozens of times a day, the math flips in a hurry.

Three signs it’s probably not the fit

1. The work is mostly judgment and relationships. Sales, strategy, managing clients, creative direction. These lean on nuance, context, and trust in ways AI can help with but can’t replace. Don’t automate the thing that makes you worth hiring.

2. Your data’s a mess. AI is only as good as what you feed it. If your data lives in spreadsheets that don’t match, scattered across emails, or in one person’s head, you’ll spend all your time on cleanup instead of on the AI. Fix the data first. This is one of the most common reasons projects stall out. The technology’s ready long before the inputs are.

3. You’re doing it because everyone else is. This is the trap I see most. “We need an AI strategy” is not a strategy. “We need to cut the time our team spends on invoice processing” is a real problem, and AI might be the right answer to it. That 42% abandonment number is built mostly out of projects that started with the first sentence instead of the second.

Why so many of these get abandoned

It’s worth sitting with that abandonment number for a second, because it’s the best argument for being deliberate. When companies say why they scrapped an AI project, the reasons rhyme: unclear business value, cost, data and security headaches, a skills gap. Notice what’s not on the list. “The model wasn’t good enough.” The technology is rarely the bottleneck. The bottleneck is starting without a clear problem, a way to measure success, and the data to back it up.

The businesses that win treat AI like any other investment. They decide on the outcome first, run a small test, and measure it before scaling. The ones that lose treat it like a mandate to “do AI” and work backward from the tool.

What to do if you’re not sure

Start small. Pick one process, scope it tight, and build something real. A focused pilot with a number attached tells you more than any amount of planning.

Decide what success looks like before you build. A number you can point at, like “cut intake from six hours a week to one.” If you can’t name that number, you’re not ready to build. You’re ready to investigate.

And if you want a shortcut, have someone who understands both AI and your business walk through your operations and tell you where the real payoff is. Half the value is often hearing “not here, but definitely there.”

Frequently asked questions

Is my business too small for AI? Almost certainly not, but “too small” is the wrong test. The right test is whether you’ve got a repetitive, high-volume task with clear rules. A two-person shop with a daily document bottleneck is a better candidate than a 200-person company with no clear problem to solve.

Do I need to hire AI specialists to start? No, and for most small businesses a full-time hire is premature. A focused engagement to find and build the right first project is usually the smarter move. You get the results without the fixed cost, and you learn whether a bigger investment is justified.

How do I avoid becoming part of that 42%? Define the problem and the success metric before you pick a tool, start with one tightly scoped pilot, and make sure your data is clean enough to support it. The projects that get abandoned almost always skipped one of those three.

What if AI just isn’t right for us? Then the most useful thing an honest advisor can do is tell you that before you spend the money. Sometimes the answer is a simpler automation, a better-organized spreadsheet, or a change to how you work. “You don’t need AI for this” is a real and useful conclusion.

That’s the kind of work I do in my AI consulting engagements. Not pitching technology, but finding the problems worth solving first, and being straight with you when the answer is to wait.