← The notebookField notes · 4 MIN READ

How to Tell If AI Will Actually Pay Off for Your Business

Every owner I talk to asks some version of the same question: "Will this actually make us money, or is it another expensive experiment?"

It's the right question. Most companies just ask it too late, after they've already bought the tools.

McKinsey's 2025 State of AI survey found that only 39 percent of respondents say AI has had any impact on their company's EBIT (basically, operating profit), and most of them say it's less than 5 percent. Nearly two-thirds say their companies haven't started scaling AI across the business. So lots of companies are using AI. Far fewer are getting paid for it.

The good news is you can tell ahead of time whether AI is likely to pay off. Here's how I look at it. If you'd rather walk through your own numbers with us, that's exactly what we do at AI Answered.

Start with the number you want to move

AI doesn't pay off in general. It pays off on a specific number.

Before anything else, I want to know which number we're trying to change. Hours spent on a process. Cost to serve a customer. How fast quotes go out. How many leads get followed up. Revenue per salesperson.

If nobody can name the number, the project isn't ready yet. I wrote more about this in AI won't fix a business that doesn't know its numbers.

Write down where that number is today

You can't prove something improved if you never measured where it started. That sounds obvious, but it's where a lot of AI projects fall apart.

Pick the process. Measure it for a few weeks the boring way: how long it takes, how often it goes wrong, what it costs. That's your baseline. Everything later gets compared to it.

Count the full cost, not just the software

The software is often the smallest part. The real costs usually include:

If the expected gain doesn't beat all of that, it won't pay off, no matter how impressive the demo looked.

Check that people will actually use it

This is the part spreadsheets miss. A tool nobody uses returns nothing.

McKinsey's AI measurement framework tracks adoption right alongside financial impact for this reason: who is using the tool, how often, and whether they accept its answers or keep overriding them. If your team sees AI as a threat instead of help, usage stays low and so does the return. That's why I always start with people before tools.

Look for the redesign, not the shortcut

The companies getting real value aren't just bolting AI onto the old way of working. In the same State of AI survey, redesigning workflows shows up as a key factor for the companies seeing the most value.

So the question isn't "where can we add AI?" It's "if we built this process from scratch today, what would it look like?" That's usually where the money is.

Test small, then decide

Don't roll it out everywhere at once. Pick one process, run it for a set period, and compare it to your baseline. If you can, run the old way and the new way side by side so you know the change came from AI and not something else.

Then make a real decision. Scale it, fix it, or kill it. Endless pilots are how companies end up "using AI" without ever seeing a return.

Signs it probably won't pay off

If two or more of those sound familiar, slow down and fix those first. How to tell if your company is ready for AI goes deeper, and if you're weighing who should do the work, read build AI in-house or hire a partner.

Why we look at the numbers first

This is how we run every engagement at AI Answered. Before we recommend anything, we look at the company's numbers and find where AI can actually move them. Then we price based on the value we can deliver, so you're never paying for a vague promise.

If you lead an established company and want a straight answer on whether AI will pay off for you, let's talk.

And if someone on your team still feels lost on the basics, our book AI for Boomers explains it in plain English.

More from the notebook ↗