AI won't fix a business that doesn't know its numbers
A lot of companies are using AI right now. That isn't the same thing as AI helping the business.
In McKinsey's 2025 survey, 88% of respondents said their organization regularly uses AI in at least one business function. Only 39% said it had any impact on enterprise earnings, and most of those said it was under 5% of EBIT. The survey ran in June and July 2025, with nearly 2,000 respondents. You can read it here.
I don't think that gap is mysterious. A lot of teams added tools. Fewer teams started with a number they were trying to move.
If you can't name the number, you can't tell if it worked.
"We're using AI" is not a result. Neither is a stack of logins, a pilot that never ended, or a slide that says the team is "exploring use cases."
The useful question is simpler. What in the business should get better, and how will you see it?
Time on a repeated task. How long a lead waits. How often a handoff gets dropped. Where deals stall. What it costs when a process fails. Those are numbers you can write down before anyone opens a new tool.
If you don't have that before, the after will always sound like a success. People remember the demo. They forget to measure the week.
AI makes a fuzzy process faster. That's not a compliment.
I've said this in a shorter way before: do the work manually, then automate it. And: start with the person, not the tool.
The same idea applies to the business as a whole. If five people run the same process five different ways, a model won't pick the right one for you. If nobody owns the handoff between sales and delivery, a chatbot won't fix the handoff. If you don't know which work actually makes money, better copy will just help you do more of the same.
This is why I start with the numbers. Not a 40-tab model. The basic ones. Where does revenue come from? Where does time go? Where does work wait? What would "better" look like in a figure you already trust?
The companies getting value tend to change the work, not just add a tool.
That same McKinsey research found that the companies seeing the most value from AI were much more likely to redesign the workflow, not drop a model into the old steps. That matches what I've seen. The tool is the easy part. Deciding what the person does, what the model drafts, and what should never go out unchecked is the actual work.
You can do a small version of this without a big program. Pick one process. Write how it works today. Choose one number. Change the steps. Measure. Keep it or drop it.
What you shouldn't do is buy a pile of software and hope the results show up somewhere in the P&L later.
This is also how I'd want a partner to show up.
I'm Co-founder of AI Answered. We work with established companies as a growth partner. The first job is to look at the numbers, agree on what we're trying to move, and only then build. We price on the value we can deliver, not a package that ignores the business in front of us.
If a partner can't talk about your numbers first, they're guessing. Guessing is expensive, even when the invoice looks tidy.
If this is the kind of help you want, I laid out how we work on the growth partners page. It starts with a review of your numbers. If we aren't a fit, that should be clear early.
For people who want the simple version of the tools themselves, Noah Kruthaupt and I wrote AI for Boomers. The idea in the book is the same idea here. Technology should simplify your life, not complicate it. A company is no different. If AI can't be tied to a number you care about, you're not late to the future. You're just not ready to spend on it yet. Get the numbers clear, then start.
The practical first step is where to start with AI in your business. If you're choosing who to hire, here's what to look for in an AI consultant.