Sidekick Orchestration

Does AI actually pay off? What the data says, and how much to invest

What the research and real client numbers say about AI's return for small businesses, why $500,000 of freed-up capacity often turns into $100,000, and how much to invest.

// In this article

People ask me how much they should invest in AI. Usually as a percent of revenue, sometimes per employee.

I don't think a percent of revenue is the right place to start. It's not spending, it's investment. So start with what it pays back, and be honest about what decides whether it pays back at all.

A quick disclosure first. I sell Fractional AI leadership. Getting these results for clients is my job, so I'm biased. That's why the sources are up front. You can disagree with me, but here's the data.

What the data says

  • The leaders pull away. The 7.5% of companies furthest along with AI deliver 2.4 times the revenue growth and 2.8 times the profit growth (EBITDA, roughly operating profit) of the companies furthest behind (BCG, Sept 2026). It's a survey of 1,330 senior executives, so it reflects bigger companies more than small ones.
  • Frequent users grow more. 68% of small businesses that use AI frequently grew revenue in 2025, compared with 37% of those that rarely or never use it (PYMNTS, Oct 2026, 533 US small businesses).
  • Few say it hurt. 43% of US small and midsize businesses using AI say it raised their revenue and 2% say it lowered it. 29% say it cut costs and 17% say it raised them (QuickBooks 2026 AI Impact Report, survey of 34,000+ owners in four countries).
  • Profit too. Small businesses using AI are at least 7 points more likely to report sales, profit and workforce growth (US Chamber of Commerce, 2026).

A fair caveat: these are surveys. They show a link. They don't prove AI caused the growth, and growing businesses may simply be the ones with room to invest.

Why most businesses don't see it

In a 2026 survey of 1,256 owners in Goldman Sachs' 10,000 Small Businesses program, 76% use AI. Only 14% say it is fully built into how their core work runs (Goldman Sachs 10,000 Small Businesses and Babson, 2026).

Almost everyone has tried it. Very few have changed how they work around it. I've written more about that gap in Where AI rollouts go to die and How to scale with AI.

Real numbers from clients

These are before-and-afters clients have shared with me, for specific tasks:

  • HR consulting firm: a social post went from about an hour to about 5 minutes. Standard proposals take 75 to 80% less time. The first draft of a large RFP went from days to hours.
  • Local home builder: proposals went from about 40 hours to about 4.
  • Advisor: preparing for a client call went from about 45 minutes to about 5.

Here's what those numbers don't mean. Cutting one task by 90% doesn't make someone ten times more productive. One task got much faster, and the time went somewhere else. The HR firm described it well:

“The greatest benefit has not been replacing work, but redirecting effort from administrative tasks such as searching, formatting, graphic creation, and document drafting toward higher-value activities including business development, client relationships, proposal strategy, and revenue-generating work.”

An HR consulting firm we work with, from their quarterly AI update

The real question: what do you do with the time?

If all you do is bank the efficiency, you end up asking what to do with your staff. If you put the time into growth, the bottleneck moves to the work that needs people.

That's what happened with the builder. Proposals that took a week now take half a day. In his words, he's the busiest he's ever been, and he has the highest win rate he's ever had. Revenue is growing. But he can't get all the value he otherwise would, because he doesn't have enough skilled tradespeople to do the work. Right now he's trying to hire a master carpenter.

That's the real world. On paper, AI might free up $500,000 of value. In practice, it moves the bottleneck, and the new constraints in your business decide how much of that you actually get.

For a lot of businesses, using AI to grow revenue and headcount is worth far more than using it to cut hours. One of the biggest places to look is whether buyers can find you at all. Our free AI Search Check shows whether AI tools recommend your business when buyers ask.

Napkin math for a 50-person company

How much time does AI actually save? The research lands in a fairly consistent range for people who use it regularly:

  • More than 8 in 10 workers who use AI say it saves them at least an hour a day (Indeed Hiring Lab, a survey of 80,000+ workers in eight countries). A separate study by the London School of Economics and Protiviti found workers with AI training can save upward of 11 hours a week.
  • The median AI user in Europe saves about 3 hours a week (European Central Bank, 2026).
  • Workers say AI saves them about 11 hours a week, but they spend about 6.4 hours checking and fixing its output (Glean Work AI Index 2026).

After the checking and fixing, 3 to 5 hours a week is a fair range for someone who uses AI every day. Using the top of that range, and assuming all 50 people become daily users:

$517k

50 people x 5 hours a week x $45 an hour x 46 weeks = about $517,000 a year in freed-up capacity.

Now the honest part. That's capacity. It isn't money in the bank.

Not everyone will use it every day. Not every saved hour goes somewhere useful. Changing habits, attitudes and how a business runs takes time, and you need data to decide where to move people and effort. In my experience, the businesses using AI well have a clear answer for where the time goes. I think that's a big part of why the leaders in the surveys above report stronger revenue and profit growth. A lot of businesses don't have anyone inside who can make those calls.

So the same capacity can turn into very different results:

How well the time is usedRough economic value a year
Mostly absorbed, little changesabout $100,000
Some of it moved into better workabout $500,000
Moved into growth: more clients, more work won$1 million or more

These ranges are my own estimates from client work. No survey measured them. In my experience, the top row is where many businesses land without someone owning it. The bottom row is possible because time put into winning more work can be worth more than the hours themselves.

If you're making a case to your board or partners, plan around the top row. If the decision still makes sense there, it's an easy one.

How much should you invest?

  • The tools are the cheap part. Most of what you're paying for is the time, judgment and leadership to change how work gets done.
  • A reference point, not a target: large companies have gone from spending about 1% of revenue on AI in early 2025 to about 3.3% today (BCG, 2026).
  • My rule of thumb: use the conservative number. Spend up to what that pays back in about six months. Pilot first, measure, then expand.

What a rollout looks like

One person owns it and is accountable for the result. Leaders get good first, then the rest of the team. In my experience, most teams can get pretty proficient in about 90 days. It won't all be perfect by then, and behaviour keeps shifting after that, but people know how to use it and where the time should go.

That's the work I do with Fractional clients, and it's why I keep coming back to change management. The tools are ready. The harder part is the people and the decisions. More on that in How to adopt AI at your company.

Run your own numbers

If you want to see where AI could save time in your own business, start with the free AI Opportunity Map. It takes about five minutes.

Next step

Run the numbers for your own business.

Book a free 30-minute call. Bring your headcount and the jobs that take the most time, and we'll work out what AI could free up and where that time should go.

Have a question first? Ask it here, text Chase at 778-999-0985, email, or connect on LinkedIn.