Everyone’s Using AI. Almost Nobody’s Using It Well.

In 1954, a milkshake machine salesman named Ray Kroc noticed something odd in his order book. A single hamburger stand in San Bernardino, California, was running eight of his Multimixer machines at the same time. Most restaurants needed one, maybe two. Curious, Kroc drove out to see what kind of restaurant could possibly sell that many milkshakes.

What he found was the result of a decision two brothers, Richard and Maurice McDonald, had made six years earlier. In 1948 their drive-in restaurant was already successful, with carhops and a long menu. They closed it anyway, for about three months, and rebuilt how it worked from the ground up. They cut the menu down to a handful of items, replaced the carhops with a walk-up window, and redesigned the kitchen around the flow of each order. To get the layout right, they drew the outline of every piece of equipment in chalk on a tennis court behind the family home, and then, as the story is usually told, brought their staff out to walk through making burgers, fries, and shakes on the chalk lines until the movements flowed. They called it the Speedee Service System.

What strikes me about that story is that the McDonald brothers didn't have any equipment their competitors couldn't buy. Every restaurant in America had access to the same grills, the same fryers, and the same Multimixers Kroc was selling door to door. Kroc showed up because of the machines, and he stayed because of the system. The advantage was the way the brothers had rebuilt the work, and the training that let their team run it the same way every time.

I think about that story a lot right now, because AI tools have become the Multimixers of 2026. Almost every business has them. Very few have rebuilt the work around them, and that's where the payoff is.

Nearly everyone has the machine

The numbers on AI adoption among smaller businesses have moved remarkably fast. Intuit QuickBooks recently published its 2026 AI Impact Report, built from survey responses from more than 34,000 small and midsize business owners along with anonymized data from more than 5.3 million QuickBooks businesses, developed with economists at the University of Chicago. Its headline finding is that more than three in four small and midsize businesses in the US now use AI regularly.

The report also shows where that use is happening. Adoption is highest in marketing, admin, and customer service, which makes sense, because those are the areas where an individual can pick up an AI tool on their own and get value right away. Write the social post, summarize the long email, draft the reply to the customer.

The most revealing finding, to me, is what happens when you look at spending instead of usage. Using anonymized payment records, the report tracked how many businesses are putting real money behind standalone AI tools, and found it's a small share, though most of those that do are staying invested. My read is that a great many businesses are using whatever AI shows up free or comes bundled with software they already own, and far fewer have made a deliberate decision to build it into how the business runs. To borrow from the McDonald brothers, most companies now own the Multimixer. Very few have redrawn the kitchen.

Why so many businesses stay at the surface

I don't think this happens because owners lack interest or ambition. The Federal Reserve Bank of San Francisco recently published research drawing on what small business owners said in the Fed's Small Business Credit Survey, and the barriers they described are revealing: the cost of the right tools, time and capacity constraints related to staff training and system upgrades, and knowledge gaps about how to implement the technology.

Two of those three barriers are about time, training, and know-how. A small business owner can sign up for an AI tool in five minutes. What's hard to find is the stretch of focused time it takes to figure out which process to change, rebuild it, and teach the team to work the new way, all while still running the business every day. The McDonald brothers solved that problem by closing their restaurant for three months, which most businesses can't do, and which is exactly why so many stop at the easy uses.

The same Fed research found owners who had deliberately chosen not to adopt AI, citing concerns about accuracy, the central role of human relationships in their business, or not seeing how it applied to their work. Those owners deserve real respect, and some of them are making the right call. In my experience, though, "it doesn't apply to us" very often means "nobody has shown us where it applies to us," and that's a gap that can be closed.

Four depths of using AI

It helps to have a way to describe how deeply a business has actually put AI to work. This is a framework we use with clients, and owners tell us it helps them place themselves quickly.

At the first depth, people use AI on their own to ask questions, brainstorm, or look things up. It's genuinely helpful and entirely individual, and the business itself is unaffected. If that person left tomorrow, the company would lose nothing.

At the second depth, people use AI to produce first drafts of real work, such as emails, proposals, job descriptions, and marketing copy. Output gets faster, but it's still one person and one tool at a time, and the quality depends entirely on who happens to be using it. This is where most of the businesses in those adoption surveys are today.

At the third depth, AI is built into a specific workflow that the whole team follows. Every new service request gets summarized and routed the same way, or every completed job automatically produces a draft invoice for review. The steps are written down, people are trained on them, and the results are measured. This is the redrawn kitchen, and it's the point where the business itself starts to change.

At the fourth depth, AI agents carry out multi-step work across several systems while people supervise, such as following up on unpaid invoices, keeping the schedule current, or preparing the weekly operations report. People spend their time on judgment, relationships, and the exceptions that need a human.

Most of the real return, the kind that shows up in your margins and in your ability to grow without adding headcount, begins at the third depth. The jump from the second to the third is where the McDonald brothers made their fortune, and it's where most businesses are stuck.

"Customizing AI sounds expensive and technical"

When I describe the third depth, owners often assume it requires custom software, a developer, or a large budget. Sometimes it involves those things, but usually much less than people expect.

Most of the work of moving a process to the third depth is writing it down and giving the AI what it needs to do the job your way. That means your pricing, your templates, the questions you always ask a new customer, the way your best estimator structures a quote, and the rules for what needs a person's sign-off. Today's AI tools are quite good at following a clear process once someone has described it. What they can't do is invent your process for you, and that knowledge lives in your team's heads.

So the real cost of going deeper is time and attention, especially from the people who know the work best. That's a genuine cost for a small business, and it's the same one the McDonald brothers paid on that tennis court. It's also the kind of investment that keeps paying off, because once a process is written down and the team is trained on it, it works the same way every time, for everyone, including the next person you hire.

How to go one level deeper

The most valuable move for most businesses is taking a single workflow from the second depth to the third. You don't need to reinvent the whole company at once.

It starts with choosing one painful, repetitive process. Look for something that happens many times a week, follows a roughly similar pattern each time, and frustrates the people who do it. Quoting, scheduling, intake, and invoicing are common candidates in the businesses we work with.

It continues with learning from whoever already uses AI on that work. Someone on your team is almost certainly experimenting already. Their habits are the first draft of the new process, and asking them about it tells your team that their experiments are valued.

Then write the new process down, step by step, including what the AI does, what a person checks, and what happens when something looks wrong. Give each person who touches it a short, hands-on session using real examples from your business, and check in a week later, because that's when the real questions surface. Pick a single number to watch, such as hours per week, turnaround time, or error rate, record it before you start, and look again after thirty and ninety days.

After two or three workflows, something useful happens. Your team starts spotting the next opportunity on its own, because they've seen what the redesign looks like and what it does for their day.

Where Avolis stands

Nearly every business has the tools now, so we think the next few years in small and mid-sized business will be decided by who redesigns the work around them. That redesign is mostly people work: choosing the right process, writing down what your best people know, training the team, and staying with it until the new way holds.

That's the work we do. We start by finding where the work in your business actually gets stuck and which process would pay off most if it went one level deeper. We choose tools that fit and explain why, then build the workflows and agents around how your team already operates. We train each role on its own part of the new process, capture what your best people know so the whole company can use it, and set clear rules for what AI handles on its own and what needs a person. We think of it as drawing the new kitchen with your team on the tennis court, and we stay until they can run it without the chalk lines.

If you'd like an honest read on how deeply your business is using AI today, and which workflow would pay off most if you went one level deeper, a good place to start is our AI Readiness Evaluation.

Sources

• Wikipedia, "History of McDonald's": https://en.wikipedia.org/wiki/History_of_McDonald%27s

• Gemba Academy, "McProcess" (February 17, 2017): https://blog.gembaacademy.com/2017/02/17/mcprocess/

• Intuit, "2026 AI Impact Report: Mapping Adoption, Use, and Impact Across Small to Midsize Businesses" (2026): https://www.intuit.com/blog/global-stories/ai-impact-report/

• Federal Reserve Bank of San Francisco, "AI Adoption Among Small Businesses: Qualitative Insights from the Small Business Credit Survey," Community Development Research Brief 2026-01 (July 15, 2026): https://www.frbsf.org/research-and-insights/publications/community-development-research-briefs/2026/07/ai-adoption-in-small-businesses-2024-sbcs/