If you run a small business, you have probably heard over the last year that AI is "changing everything". You have also probably noticed that most of the people telling you this are trying to sell you something via a social media thread and you aren't quite sure what to believe because to you, AI seems like a mystery. It is hard to tell what is actually happening with AI today from what somebody wants you to believe is happening. That can lead to a lot of misinformation and FOMO that your business is falling behind without real AI adoption.
To help sort through the noise, we sifted through many credible sources and discovered some interesting truths along the way. How did AI get better this year, and where does that leave me and my small business?
Did AI really get better this year?
In terms of raw computing and reasoning power, yes. In July of 2025, Google DeepMind entered a version of its Gemini model into the International Mathematical Olympiad, widely considered the hardest high school math competition in the world. It solved five of the six problems and earned a gold medal score inside the same four and a half hour limit the students get with no original data to train from. Those results were officially graded and certified by the IMO's own coordinators, using the same criteria they apply to the students.
What makes it more impressive is the year before at the 2024 Olympiad, Gemini needed the problems translated out of plain English by hand first, and then needed two to three days of computing to work through them. Within one year, the AI model was able to cut its reasoning time from three days to four and a half hours and finish with a score rivaling the smartest mathematics students in the world.
This same pattern of improvement showed up in the laboratory as well. In May of 2026, Nature.com published work on a Google system called Co-Scientist that generated new ideas for repurposing existing drugs against acute myeloid leukemia, a type of blood cancer. Those ideas were then confirmed by real experiments in a lab.
In December of 2025, Nature published enzymes designed from scratch by an AI model that worked orders of magnitude better than previously designed enzymes of their kind. And in January of 2026, the Lancet published the first randomized controlled trial of AI-assisted mammogram screening, which found the AI-supported readings caught more cancers, 80.5 percent versus 73.8 percent, without producing any more false alarms than standard screening. These research and laboratory findings are rapidly escalating the scientific community's ability to produce groundbreaking discoveries which wouldn't have been possible just one year ago.
Did AI actually become cheaper?
Amazingly, with the extreme growth of AI data centers, compute price came down significantly. Anthropic's top model (Opus 4.5) went from 15 dollars per million input tokens and 75 per million output tokens, down to 5 and 25, inside a single year. That is a two thirds cut on the flagship. A token is roughly a word fragment that is internalized by the AI model in order to understand direction and intent, and it is the unit by which these AI models bill their customers.
Even more interesting, Anthropic reported that Opus 4.5, set to a medium effort level, matched Sonnet 4.5's best score on a standard coding test while using 76 percent fewer output tokens to get there. The price per token came down, AND the number of tokens needed to finish the same job came down at the same time. The real cost of completing a task dropped in terms of dollar amounts and compute power. This is why AI tools that felt like expensive novelties eighteen months ago are now being built inside of software you already use (like Microsoft Copilot or Google Gemini) with free basic tiers for all users.
How many small businesses are using AI?
Not many. The Census Bureau found AI use hovered between 17 and 20 percent of American businesses from December of 2025 through May of 2026. This is essentially a flat trend line through the most eventful year this technology has ever had.
With all of these massive improvements and cheap rates, you would assume businesses rushed in, but they did not.
| Business size | Currently using AI |
|---|---|
| 4 or fewer employees | Under 20 percent |
| 250 or more employees | 37 percent |
| All US businesses | 17 to 20 percent |
The Census Bureau's own summary says that use increased among firms with at least 20 employees, but did not change significantly among firms with fewer than 20.
The AI Models got better and cheaper, but the businesses that started using them were mostly the ones that already had staff to spare and time to learn. If AI adoption has felt like some phenomenon happening to someone else, you wouldn't be alone in thinking that, and you are not behind your peers. Very few small businesses are taking the time to learn about AI and utilize it to speed up their processes, create better quality products, or improve their marketing.
Why do so many owners push back on AI adoption?
The Small Business Administration's research office looked at why small firms were not planning to adopt AI. For businesses under five employees, roughly 82% gave the same reason: lack of relevance. They said, in effect, that these AI models and tools do not apply to a business like mine. Lack of knowledge about AI came in at 6.7% and privacy concerns at 6.3%.
The Census Bureau noticed a similar trend from the other direction. They asked businesses whether they used AI "in producing goods or services", but when they sat down and interviewed owners, they found people answering no who turned out to be using AI regularly, just not in a way that felt like producing anything. The examples of AI use that the Census recorded were hiring new employees, project management, accounting, research and development, and built in assistants for software the business already used.
In November of 2025 the Census Bureau rewrote the question to ask about "any of its business functions." The answers shifted enough that they had to start a brand new data series, but the sentiment around using AI to help improve core business functions did not change.
Business owners have pushed back on AI adoption over the past year not because of its compute power, cost, or capabilities, but because they lack the creativity to utilize it to fit their specific business operation. This is creating a massive opportunity in the market for any small business to leapfrog the competition if business owners would dedicate even a minimal amount of time and resources into AI adoption.
What does this mean for your marketing?
Among the businesses that did adopt AI, the Census Bureau found the number one use is sales and marketing at 52%, ahead of strategy and business development at 45% and IT at 41%.
The big picture is a technology that got much better and much cheaper, adopted mostly by companies with more than twenty employees, and has typically been pointed at one or two functions for a company. There are endless applications for AI adoption, but here at TKBS, we can help you navigate your marketing and also use AI to help you get ahead of the competition.
Overall adoption is flat for a reason, and the reason is because nobody has explained what AI adoption looks like for you in a business your size. The New York Federal Reserve found that 37 percent of workers say their employer offers no AI tools at all, another 11 percent say their employer forbids them, and only about 16 percent of employers offer any training. Nearly everybody is starting at the beginning, but the good news is it is faster and cheaper than ever to start utilizing AI in your small business.
Many business owners ask us, "how do I keep up with the rapid advancements in AI? What can I do to make my business more efficient with AI?" We always say, don't immediately go out and buy a flagship model subscription. Having a subscription to Claude or GPT won't solve your problems overnight. First, walk through your business strategy and figure out where your marketing gaps are to increase your revenue, then decide whether a tool belongs in that gap. A business with no reliable way to capture a lead does not get fixed by adding AI to it. It gets a faster version of the same leak.
That is why TKBS, Turnkey Business Solutions here in Michigan, builds systems instead of selling services. We can help you navigate the rapid improvements in AI and take the first step in improving your business marketing. Once we understand where you are with your business, we can create custom software solutions that can utilize AI to impact your business results and modernize your business processes.
Frequently asked questions
Did AI really get better this year, or is that marketing?
It genuinely got better. The clearest proof is the 2025 International Mathematical Olympiad, where a Google DeepMind model earned a gold medal score of 35 points and had those results officially graded and certified by the competition's own coordinators. A year earlier the same lab's systems needed problems hand-translated and multiple days of computing.
How many small businesses are actually using AI?
Fewer than 20 percent of US businesses with four or fewer employees, according to the Census Bureau, compared with 37 percent of businesses with at least 250 employees. Overall business use held between 17 and 20 percent from December 2025 through May 2026.
Why are so few small businesses using it?
Mostly because it does not feel relevant to them. The Small Business Administration found about 82 percent of businesses under five employees cited relevance as the reason they were not planning to adopt, far ahead of lack of knowledge at 6.7 percent and privacy at 6.3 percent.
Am I behind if I have not started using AI?
No. The Census Bureau found adoption among businesses with fewer than 20 employees did not change significantly over the last year, so most of your peers have not moved either. The New York Federal Reserve also found 37 percent of workers say their employer offers no AI tools at all.
What do most businesses use AI for?
Sales and marketing, at 52 percent among businesses that use it, ahead of strategy and business development at 45 percent and IT at 41 percent.
Can I trust what AI tells me?
Not without checking. An audit by the European Broadcasting Union and the BBC covering more than 3,000 responses found 45 percent contained at least one significant issue, most often involving sources. Treat AI output as a starting point that needs verification, not as an answer.
Why do AI adoption numbers seem to contradict each other?
Because different surveys ask the question differently. The Census Bureau found owners answering no to whether they used AI, then describing using it for hiring or accounting in the same interview. It rewrote the question in November 2025 and started a new data series, because the old and new answers no longer meant the same thing.
Where should a small business actually start?
Start with where your marketing is leaking, not with a tool. If leads come in and nothing captures them, or nothing follows up, adding AI speeds up the leak instead of fixing it. Get the system right first, then use AI to make parts of it faster.
Find Out Where Your Marketing Actually Leaks
We will not sell you an AI strategy, because that is not a real thing. What we will do is look at how customers actually find you, what happens when they do, and where the gaps are, then show you the real numbers behind it.