Will AI kill my small business?

AI is most likely to hurt the small businesses that wait. Nearly nine in ten companies now use AI, yet only about 6% of the companies McKinsey surveyed get real profit from it. Using AI means typing a question into ChatGPT and saving a few minutes. Adopting it means rebuilding a real process in your business around AI so it saves hours every week. US Census data shows AI use climbing in firms with 20 or more people and standing still in firms under 20. Only 6% getting real results means it’s still early, and the best time to start is now.

Have you actually adopted AI? Or are you just using it?

Using AI and adopting AI are completely different.

The majority of us have logged into our computer, opened up a web browser, gone to ChatGPT, and entered something into it. Simply by doing all that, you can say that you have used AI. For your business, simply using something like ChatGPT or Claude is not actually producing you with anything that is making you money or making you more productive. At most, it might be saving you a few minutes of time for things like responding to an email by writing it for you. Nearly 9 in 10 organizations, or around 88%, use AI in at least one function. And 77% of US small businesses use AI regularly, which is up from 48% in mid-2024. But simply opening up the chat window, typing in your question and copy-pasting the answer doesn’t mean you’ve actually adopted AI into your business. This is the exact reason that you might have dropped AI. You used these tools, but it never really felt like they got anywhere, because they weren’t actually saving you much time.

Adopting it goes much further. Adopting AI means integrating it into an actual process within the business, and that integration actually saves you time and gets real work done. If using AI means copying an email, pasting it into ChatGPT and asking for a response, then adopting it means creating an agent that does email triage over the entire inbox. This agent would act in the capacity an executive assistant would in terms of organizing the inbox. This means logging on in the morning and reading over all of your unread emails, checking which ones require you to reply to them and drafting a response for review, as well as marking spam and marketing emails as read.

The main difference here is that when you adopt AI and it stops working, it feels as if an employee isn’t working on it anymore.

If we look at the numbers of who’s actually benefiting from AI, we see that only 37% have seen any profit impact at all, and only about 6% of the companies McKinsey surveyed get 5% or more of their profit from AI. Most companies are still in the first group, only using AI.

The businesses that get the most stuck in that first group are the smallest ones.

Sources: The state of AI in 2026 (McKinsey) · Stanford AI Index 2026 and Intuit 2026, as cited in Boris Blum, The AI Innovation Wave (MY CEO Advisory Research, Sept 2026)

Why are businesses under 20 people falling behind?

The US Census found that from December 14th, 2025 to May 3rd, 2026, AI use rose in firms with 20+ employees, but didn’t move much at all for firms under 20. McKinsey found that large companies scaling AI agents went from 27% to 40% in about a year, while smaller organizations stayed flat at around 22%.

So in the beginning, large firms used AI at around 1.8 times the small business rate in early 2024, then down to 1.2 times by August 2025, because the tools got cheap. But now that gap is reopening on the doing rather than the access.

So if 77% of small businesses say they’re using AI, why is the gap opening? Once again, I’m going to drive home the point that using is not adopting. Let’s take a look at the actual way that these large businesses are going from that 27% to the 40%.

A competitor with 30-plus people can buy the same tools that a smaller business can. However, this competitor has actually adopted AI. That means they’ve rebuilt full end-to-end processes so that AI is fully incorporated into them. This means that things like automated lead follow-up, proposal drafting, or onboarding flows all happen as a result of something that AI did. They are answering, quoting, and onboarding their clients faster with fewer people. All while the small business is still pasting into the chat window.

This has nothing to do with what you can afford or how technically inclined you are, either. AI is getting cheaper by the day, and proportionally both the small business and the larger competitor can easily afford to bring AI into their business. And the competitor doesn’t need some IT or software development team to implement any of this either. Most likely, since the tools are readily available, they hired somebody who knew what they were doing to build it and to train their team to use the tools and keep them running.

So, if the tools are cheap and everyone has them, why is most of the money going nowhere?

Sources: Large Firms With at Least 20 Employees Biggest AI Users (US Census Bureau) · The state of AI in 2026 (McKinsey) · Business.com, as cited in Boris Blum, The AI Innovation Wave

Measure yours in 3 minutes: the free Task Audit scores how much of your recurring work could run without you.

Why does most AI spending go nowhere?

Sorry to throw more numbers at you, but if we look at them again, it’s easy to see why the record spending is getting flat results. By some estimates, more than 80% of AI projects fail. 42% of companies scrapped most of their AI projects, which was up from 17% the year before, and 46% of AI pilots were killed before launch. This isn’t to say that the problem is with AI itself, rather, it’s a people problem. RAND found that the top causes were either focusing on the tool instead of the business problem, or just focusing on the wrong problem entirely. The problem was never model choice or budget.

Firstly, we tend to put our money in the wrong place. One MIT study found that about 50% of AI budgets go into sales and marketing. However, the best returns and the biggest savings are found in back-office automation. Again, that’s the invoicing, client updates, proposals, and onboarding, the admin that eats away at your week.

Secondly, another mistake is trying to build it yourself. The same study found that bought or vendor-made tools worked around 67% of the time, while in-house builds worked only about a third as often. The specialist has done it many times and knows what they’re doing, while a team learning on the job while still running the business is less likely to succeed.

What the 6% have in common is a focus on the work itself. Whether or not you decide to hire somebody, your focus should be on redesigning your processes into AI workflows from end to end. McKinsey found that around three out of four of the high performers fundamentally redesigned their workflows, versus about one in four of everyone else. This doesn’t mean adding AI to the old steps. It means looking at the process and redesigning the steps to fit AI.

Sources: The Root Causes of Failure for Artificial Intelligence Projects (RAND) · AI project failure rates are on the rise (CIO Dive, on S&P Global) · 95% of companies get zero ROI from GenAI initiatives (MediaPost, on MIT NANDA) · Boris Blum, The AI Innovation Wave (MY CEO Advisory Research, Sept 2026)

Is it too late to start?

With every new technology, there is a general adoption curve. It starts with a few enthusiasts and innovators. Then some early adopters. Then the practical early and late majority. And finally, the stragglers and laggards.

We are currently sitting somewhere between the innovators and the early adopters. Some say we’re still very early and some say we’re a little further into it, but the real thing to look at is whether or not we’ve actually crossed what people call “the chasm.”

The chasm is simply where people stop asking “is this interesting?” and start asking “is it actually paying?” It’s basically realizing that you can make money from it.

If we take a look at where business is by usage, it looks like we’re in the late majority. If we measure by adoption, only about 6% are getting real profit, so we’re still at the very start.

The timing varies here and there, but it’s estimated that we’ll cross the chasm in around 12 to 24 months. So what does that mean for the small business?

You’ve probably heard the cliché that the best time to plant a tree was any time in the past 20 years, and the second best time is right now. This applies directly to AI. So, to answer the question, no, it is not too late to start yet.

But you may ask whether or not you should wait for the tools to develop and get better. My answer there is sure, you can wait for the tools to get better, but you can’t wait on the groundwork. We’re already seeing AI built into your CRM and your scheduling links, and it’s already in Gmail. What we’re not seeing is businesses building the groundwork to really benefit from those tools being implemented into their systems.

The groundwork simply means writing down your processes, how your teams are using them, what data you collect and from where, and basically anything AI could be using to benefit your business and save you time or money. The groundwork will make anything pay off no matter what happens with the AI tools in the apps. AI, no matter where, no matter what tool, only works as well as the data, the strategy, and the people under it. So if you wait on a tool to get implemented and you have no groundwork, you’re going to be switching on a feature and finding out that it has nothing useful to work with.

Sources: Boris Blum, The AI Innovation Wave (MY CEO Advisory Research, Sept 2026)

What is McKinsey telling big companies to build?

One of the biggest consulting firms in the world is telling large companies to build one thing: an “AI operating system.”

I’ve explained it in a previous newsletter, but McKinsey’s AIOS is simply an AI structure that knows how the whole company works, can answer anyone’s question about the company, recommends decisions, carries out approved ones, and gets better with every result. They call it a “second brain” for the business.

McKinsey walks through a made-up example where the CEO of a big insurance company hears about regulators in several countries planning to cap how much profit some products can make. The CEO asks their AIOS how exposed the company is, and within minutes it says that about 140 products across seven countries could be affected. The next morning, a manager for their German department looks at the numbers, pushes back on one of the system’s assumptions, and reruns it. From there it’s found that only nine products are actually in trouble, and by the next day the system has laid out the options for each one, such as changing the price or pulling the product entirely. The leadership team then picks a plan, the AI drafts the new prices and the letters to partners, and people approve every one. McKinsey says that’s normally weeks’ worth of spreadsheets going back and forth between teams, where their AIOS took three days.

McKinsey lays out their layers as follows. Their AIOS has a written-down map of how the business works, calculators that AI can use to get accurate numbers, the actual AI workforce in the form of AI agents, and a front door inside the tools that the people in the company already use. This is all built on top of what they already own, and nothing is ripped out.

They say that in their early results, it can get planning done up to 10 times faster, each new area implemented pays back 50% faster, and they can find their first value in about 12 weeks. They also have two firm build rules: own what you know and rent the tech, and redesign the work end to end.

But what does this mean for a small business? Every piece of the enterprise AI operating system has a direct translation to a small business version. And in my opinion, small businesses can benefit from an AIOS much, much more than these large corporations.

Sources: The AI operating system: A new way for CEOs to run the enterprise (McKinsey)

What does that look like in a 10-person business?

It’s validating to see a large consulting firm recommending something that people in my field and I have already been building for small businesses. To me, it seems like McKinsey’s version of an AI operating system is the same thing that I’ve been building, just scaled up. And I believe that small businesses do get more value and can build much faster than the corporations can.

Let’s translate everything from McKinsey’s version to mine.

Firstly, that map of the business translates directly to your offers, prices, clients, processes, and your tone, all written down once. Since your business isn’t some large corporation, you can build all this with AI in about a day. This is simply providing AI with documents or written accounts of all these different parts of your business, and working together to create a real, in-depth understanding of how the business works.

Their calculators, which are primarily used for analytics and keeping the business up to date, are the same as any KPIs and analytical data that you might be tracking. From social media engagement and likes all the way to tracking your revenue, AI can look at and calculate direct numbers to give you better strategy. While that’s necessary for large corporations to make sure they’re not falling behind anywhere, smaller businesses typically neglect tracking their KPIs. So using AI to track them for you not only gets you actually tracking them, but gives you the direct benefit of tracking them.

The AI workers are the same in both versions of the AIOS. Having AI agents available to do work for you benefits both large and small businesses. AI agents and AI subscriptions are relatively cheap across the board, and proportionately the spending is the same as well.

Finally, McKinsey’s front door, or how they’re actually using the AI, is much simpler for a small business. This is still just a chat on your computer, or even your phone, but this chat is linked to those AI workers who have the capability to get work done. A large corporation has to make sure this works for everybody in the corporation, where a small business only has to make it work for a small team.

I can give you a direct example from somebody I’ve helped. My publisher client was at the “using” stage a while ago. She was using ChatGPT for emails and some blog posts, but nothing actually touched the work itself. So together we rebuilt the project work around her 200-plus-task list for each project. We started with one process, redesigned it for AI, and if it were to stop working, she would immediately see the difference. I helped her build it, and in the case study we did together, she counts around 25 to 30+ hours a month back in time saved.

Then there’s the speed it took to see results. McKinsey says first value takes around 12 weeks at enterprise scale. Smaller firms like my client can move much faster and can see results in about two weeks of dedicated work. And this is all with a good boundary as well. We’re keeping a person in charge of approving what goes out. AI is just drafting and doing the legwork. All the judgment calls, pricing, and client conversations stay the owner’s.

My recommendation to you, if you’re a small business owner, is to pick the one process that eats the most hours but doesn’t necessarily require you. Today, I want you to write down how it runs end to end. And then tomorrow, work with AI to rebuild it so AI can take it over. Finding that first process is what my AI Opportunity Assessment does.

It’s one recorded call with me, about an hour, for the owner of the business. I map where your week goes and find your top three AI opportunities, so you leave knowing which process to rebuild first. It’s free because the call becomes content, and you approve everything before it goes out. I only take two a month, and you can apply for the free AI Opportunity Assessment here.

And if you’d rather just talk it through first, I do a free discovery call.

Frequently asked questions

Have you actually adopted AI? Or are you just using it? Using AI means opening ChatGPT, typing a question and pasting the answer, which saves a few minutes at most. Adopting AI means building it into an actual process in your business, like an agent that reads your inbox every morning and drafts replies for you to review. Nearly 9 in 10 organizations use AI, but only 37% see any profit impact, and only about 6% of the companies McKinsey surveyed get 5% or more of their profit from it.

Why are businesses under 20 people falling behind? US Census data from December 2025 to May 2026 shows AI use rising in firms with 20 or more employees and barely moving in firms under 20. McKinsey found large companies scaling AI agents went from 27% to 40% in about a year, while smaller organizations stayed around 22%. The gap had nearly closed once the tools got cheap, and it’s reopening now on actually doing something with AI.

Why does most AI spending go nowhere? By some estimates more than 80% of AI projects fail, and in one 2025 survey 42% of companies scrapped most of their AI projects. RAND found the top causes were focusing on the tool instead of the business problem, or picking the wrong problem entirely. About three in four of the companies getting the most from AI redesigned their workflows around it, versus about one in four of everyone else.

Is it too late to start? No. Measured by real adoption, with only about 6% of companies getting real profit, business is still at the very start of the adoption curve, and the crossing into the mainstream is estimated at 12 to 24 months away. You can wait for the tools to get better, but you can’t wait on the groundwork: writing down your processes, your data and how your team works, because AI only works as well as what’s under it.

What is McKinsey telling big companies to build? An AI operating system: one AI layer that knows how the whole company works, answers anyone’s question, recommends decisions, carries out the approved ones and gets better with every result. It sits on top of what the company already owns, with a written map of the business, calculators for exact numbers, AI agents and a front door inside the tools people already use. McKinsey reports early builds getting planning done up to 10 times faster and finding first value in about 12 weeks.

What does that look like in a 10-person business? The same pieces, smaller: your offers, prices, clients, processes and tone written down once, your numbers tracked for you, AI agents doing the work, and a chat on your computer or phone connected to them. One publishing client rebuilt her project work around her own task list and counts around 25 to 30 plus hours a month back, with a person still approving everything that goes out. Start with the one process that eats the most hours, write down how it runs end to end, then rebuild it with AI.

Photo by GreenForce Staffing on Unsplash