Work On The Business. Not In It.
Every business owner has heard some version of the advice: get out of the day-to-day and work on your business, not in it. The idea is not new. Michael Gerber built an entire philosophy of small business management around it in The E-Myth Revisited, published in 1995, and it’s been repeated by nearly every business coach since. Almost nobody disputes it. Almost nobody has been able to act on it either.
That’s not a discipline problem. The advice always assumed a step most owner-led companies couldn’t afford: someone else to do the “in it” work so the owner could step back and do the “on it” work. For a 40-person company, that someone was a general manager, an operations lead, or a second layer of management the P&L couldn’t support yet. Working on the business instead of in it was correct advice sitting on top of an expensive prerequisite. So owners nodded along, kept doing both jobs, and the advice became something you heard at a conference and forgot about by Monday’s schedule crisis.
What AI actually changes about the old advice
AI doesn’t replace a general manager, and it isn’t trying to. But it does something Gerber’s original framework never had an answer for: it absorbs a real slice of the “in it” work directly, without adding a salary line. SBE Council’s Small Business Technology Use Survey, published in March 2026, found small business owners using AI save a median of about five hours a week personally, with their businesses recovering roughly eleven and a half employee hours a week on top of that. Separately, a national survey of 1,000 SMB marketers, fielded by Talker Research for ActiveCampaign between May and June 2025, found marketers getting back around 13 hours a week, and the heaviest users (people using AI at least once a day) saving 57% more time than occasional users.
The exact number shifts depending on who’s measuring and which part of the business they’re looking at. What doesn’t shift is the shape of it. The hours coming back are disproportionately the “in it” hours: the status report, the first draft nobody wanted to write, the research that used to eat an afternoon before a client call. None of that required the owner’s judgment specifically. It required somebody’s time, and for years the owner was the only somebody on the org chart with any to spare.
The hours coming back aren’t automatically “on it” hours
Here’s the part that gets skipped when this turns into a simple win. Freed-up time doesn’t relocate itself. In that same ActiveCampaign-commissioned research, only 41% of respondents said the time AI freed up actually went toward strategic growth work. Thirty-five percent used it for other higher-value tasks, and the rest went toward things like backfilling other tools or just absorbing more of the same operational load. Time saved and time redirected are two different outcomes, and the gap between them is a decision nobody’s software makes on their behalf.
This is the piece of the “work on it, not in it” advice that Gerber’s book couldn’t solve for either, because it was never really about hours. It was about which hours get an owner’s attention. An owner who gets five hours back and spends them clearing more of the same inbox hasn’t changed anything about how the business runs. An owner who takes one of those five hours and actually looks at pricing, or the hiring plan, or which client relationships are worth protecting, has made the exact shift the advice was describing thirty years ago. AI can hand the hours back. It cannot decide what they’re for. That part is still the owner’s job, maybe the only part of the job that always was.
Where this actually shows up
We see this distinction play out directly in the companies we train. The tasks that get automated first in a cohort (research, first drafts, status reporting, the copy-paste between two systems that don’t talk to each other) are almost never anyone’s real judgment calls. They’re the tax paid to get to the judgment calls. Once that tax gets smaller, the honest question isn’t “what did we save.” It’s “what is the owner doing with the five hours that used to disappear into scheduling and email.” Owners who answer that on purpose, naming in writing which one or two owner-level decisions get first claim on the freed time, are the ones where the shift actually sticks. Owners who don’t just get a quieter version of the same treadmill, with better tools running underneath it.
None of this requires a bigger AI rollout or a more sophisticated model. It requires treating the freed hours as a decision to make, not a side effect to enjoy. That’s the actual difference between an owner who’s technically using AI and an owner who’s finally doing what the advice always asked for: working on the business, because for the first time, something else is covering the in it.