Opinions expressed by Entrepreneur contributors are their own.
Every founder I talk to has an AI strategy now. Almost none of them can tell me what it’s actually saved them. That gap is the whole story.
I’ve spent years running operating businesses across advertising, healthcare and technology infrastructure, working with ad budgets in the tens of millions along the way. Different industries, same problem underneath all of them: most of the work inside a growing company is repeatable, and most companies still solve repeatable work by hiring a person to do it by hand.
That’s the actual cost center. Not slow software, not weak talent. Just headcount sitting atop a process nobody bothered to automate.
There’s a habit most growing companies fall into without noticing. Volume goes up, someone feels the strain, and the default response is to post a job. Nobody stops to ask whether the strain is actually a headcount problem or a workflow problem wearing a headcount costume. Those are very different problems with very different fixes, and only one of them shows up permanently on your payroll.
Before I approve a new hire now, I ask one question. Is this job mostly judgment, or mostly repetition? If it’s judgment, I hire. If it’s repetition, I automate it first and see what’s left over. Most of the time, what’s left doesn’t need a full-time person attached to it.
Three examples
In one advertising business I’ve worked with, campaign reporting used to require someone pulling data from several platforms into a weekly deck by hand. We built a pipeline that pulls the same data automatically and flags anomalies instead of just reporting numbers. The hire we were about to make never happened.
In a healthcare growth operation, client onboarding used to run through email threads and a shared spreadsheet, with someone manually checking every new account for missing paperwork. We automated the checklist and the follow-up emails. Onboarding got noticeably faster and nobody had to pick up the slack manually. Nobody lost a job. We just never added the extra coordinator we were about to bring on.
In another operating business, the monthly close used to take the better part of a week, chasing numbers across different systems by hand. Now it takes a fraction of that, because the systems talk to each other and exceptions get flagged automatically instead of getting buried in a spreadsheet. The person who used to spend that week chasing numbers now spends it on decisions that actually need a person.
A fourth pattern shows up almost everywhere once you start looking for it: customer support and account management. Most of the incoming volume in any support queue is the same handful of questions asked in slightly different words. Once those get routed and answered automatically, what’s left is the genuinely hard cases, the ones that actually need a person with judgment. That’s a better job for whoever is doing it, not just a cheaper one.
None of this took a fancy model. It took sitting down for every role in the company and writing out, honestly, what that person actually does all day. How much of it is a decision, and how much of it is just information moving from one place to another.
Why most AI spending never shows up on the bottom line
Most companies buy AI tools the way they buy software licenses. Add a subscription, hope productivity goes up, never check it against a specific role or cost. That’s a bet, not a plan.
The companies where AI actually moves the margin do it backwards. They start with the org chart, not the tool. Find the roles that are mostly repetition, get honest about what that repetition actually costs, then go find the tool or workflow that replaces it. The automation answers a cost question. It isn’t a general innovation initiative somebody greenlit at an offsite because a competitor mentioned it.
It also changes how you plan headcount. The question before a new hire isn’t “do we need more hands.” It’s “is this repetition or judgment, and if it’s repetition, why hasn’t someone fixed that yet.” Ask that honestly and a lot of hiring plans get a lot shorter, which frees up budget for the hires that actually need a human being making judgment calls.
What this means if you run a small company
You don’t need a data science team for this. You need a list of every recurring task in the company, an honest look at what each one really costs, and the discipline to automate the boring majority of it before you hire for it.
Start with the roles that feel the most routine, not the ones that feel the most urgent. Urgent gets attention on its own. Routine is where the waste quietly builds up month after month until it looks like a staffing problem instead of what it actually is, which is a process nobody ever got around to fixing.
Most rooms right now are debating what the model can do. Wrong question if you’re actually running something. The right one is simpler: which of my current roles is a workflow problem wearing a job title, and what is that costing me every month I haven’t fixed it.
Fix the workflow first. The headcount you don’t add is the return. Everything else is noise.
Every founder I talk to has an AI strategy now. Almost none of them can tell me what it’s actually saved them. That gap is the whole story.
I’ve spent years running operating businesses across advertising, healthcare and technology infrastructure, working with ad budgets in the tens of millions along the way. Different industries, same problem underneath all of them: most of the work inside a growing company is repeatable, and most companies still solve repeatable work by hiring a person to do it by hand.
That’s the actual cost center. Not slow software, not weak talent. Just headcount sitting atop a process nobody bothered to automate.