What Should Montana Businesses Know About AI in 2026?

The federal numbers say the businesses most common in Montana held steady this year while bigger firms moved. That is the opening.

---

You have probably wondered whether your business is late to AI. If you employ fewer than twenty people in Montana, federal survey data released in May 2026 says firms your size held flat over the last six months while larger ones moved. That is not a scolding. It is a map, and it points somewhere useful.

This post is for the owner of a Montana construction outfit, a two-person real estate office, a family shop that has been running since the seventies. Not for a company with a marketing department. By the end you will know what the current data actually says, what Montana law says (it is the only state so far to write a right to compute into statute), and the specific order I would work in if the business were mine.

What do the numbers actually say about AI use at small businesses?

About one in five American businesses currently uses AI in some business function. The Census Bureau's Business Trends and Outlook Survey found that overall AI usage "hovered between 17% and 20%" from December 2025 through May 2026, with the national rate at 19.8% as of May 3, 2026. Firms with at least 250 employees came in at 37%, and firms with 100 to 249 employees at 32%. Worth noting what the survey actually asks: whether the business used AI in the past two weeks, so this is a current-use measure rather than a count of everyone who has ever tried it.

What matters for this audience is not the level, it is the direction, and it is worth separating the two carefully because most coverage does not. On level, the smallest firms are closer to average than you would guess. Reading the Census AI supplement published in April 2026, the Minneapolis Fed reports that "30 percent of U.S. firms with at least 250 employees said they use AI, while 17 percent of firms with fewer than 20 employees do." The Census Bureau's own biweekly series puts firms with four or fewer employees under 20%. A real gap, then, but not a chasm. Roughly one in six of your competitors is already doing something with this.

The direction is the part worth reading twice. Between December 2025 and May 2026, the Census Bureau reports, AI use rose among firms with at least 20 employees and did not change significantly among firms with fewer than 20. The larger firms are moving. The smaller ones are roughly where they were six months ago. Not because small owners are slower thinkers, but because nobody has handed them a version of this that fits a business run by four people who are already doing six jobs each.

Before going further, one piece of honesty about the headlines you have seen. The adoption number depends entirely on what question got asked, and the three numbers in circulation are not measuring the same thing at all. The Federal Reserve laid the three side by side in an April 2026 FEDS Note by Jeffrey S. Allen.

Same subject, three answers, and the spread between them is enormous. If someone tells you "78% of businesses are using AI," they are quoting a labor-force measure as if it were a firm count. A four-person shop in Bozeman and a bank with nine hundred employees both count as one firm in the first row and land very differently in the third. Knowing which number is which is most of what it takes to read this category honestly.

Even the firm-level question moved, and the size of the effect is the whole lesson. The Census Bureau notes that the survey was originally framed around AI use "in producing goods or services," and that it revised the wording in November 2025 to ask whether businesses were using AI "in any business function." The Minneapolis Fed put a number on what that change did: on the old, narrower question, the share was about 10% in late 2025, and with the broader wording it "jumped to 20 percent. That is double the rate of those who were using it for core production purposes at the end of 2025."

So half of the widely quoted one-in-five is businesses using AI somewhere around the edges rather than in the actual making and delivering of what they sell. Both are real. They are not the same thing, and a chart showing adoption climbing may be showing a question that got wider.

One more wrinkle in the same direction: the 37% figure above comes from the biweekly survey, while the April AI supplement puts large-firm use nearer 30%. Different instruments, different windows. The gap by size is real either way, which is rather the point about reading these numbers slowly.

Why does Montana's business mix change the math?

Montana is a small-business state to a degree that is genuinely unusual, and that concentration sits directly on top of the size class that has stopped moving. The state's business population is not a scaled-down version of the national one. It is a different shape.

The SBA Office of Advocacy published Montana congressional district profiles in February 2026, built on Census Statistics of U.S. Businesses data describing 2022, which is the most recent year available. Montana's District 1, in the western part of the state, led the entire nation in small business share of employment at 70.4%. District 2 came in at 64.2%. Small firms make up 97.3% of employers in District 1 and 95.9% in District 2.

Go one level finer and the picture gets sharper. Counting employers by headcount in each district:

Roughly 88% of Montana employers have fewer than twenty people on payroll. Two honest notes on that table. These are 2022 counts, the newest published, and the shape of a distribution moves slowly but the counts themselves are four years old. And the right-hand column adds the two districts together, so a firm operating in both is counted twice, which inflates the large-firm rows more than the small ones and means the small-business share is if anything understated here.

Construction alone accounts for roughly 6,200 of those very small firms across the two districts, the largest single category in both, though the SBA counts at the establishment level so a multi-location firm can show up in more than one industry. Montana also runs well above the country on working for yourself: 15.6% of private workers in District 1 and 14.1% in District 2 were self-employed in 2024, against a national rate of 9.9%.

Put the two data sets together. The size class that stopped moving is not a slice of Montana's economy. It is very nearly all of it. That is the local-first advantage, stated without any cheerleading: in a market this concentrated in very small firms, the competitor who would normally have gotten there first mostly has not.

That is a different situation than a business owner faces in Denver or Seattle, and it will not describe Montana permanently. The businesses that move first in a market like this get hard to catch, and the reason is boring and mechanical rather than dramatic: the work of becoming legible to these systems compounds slowly and cannot be bought back quickly.

What does Montana's Right to Compute Act mean for your business?

As of August 2026, Montana is the only state with a right to compute written into law, though similar bills have been introduced elsewhere. The 69th Legislature passed Senate Bill 212 in 2025, creating the Right to Compute Act with, in the bill's own words, an immediate effective date. Most Montana owners have never heard of it.

Here is the plain version. The Act's findings tie the ability to own and use computational tools to the Montana constitution's protections for property and free expression. Section 3 says that government actions restricting private ownership or use of computational resources for lawful purposes "must be limited to those demonstrably necessary and narrowly tailored to fulfill a compelling government interest." The definition of computational resources is deliberately broad, covering hardware, software, algorithms, networks, platforms, services, and machine learning.

The Act does impose one real obligation. Under Section 4, when a critical infrastructure facility is controlled in whole or in part by a critical artificial intelligence system, the deployer has to develop a risk management policy that considers a recognized framework, such as the AI Risk Management Framework from the National Institute of Standards and Technology. The statute lets that policy be written after deployment, accepts the ISO/IEC 4200 standard or another recognized framework in place of the NIST one, and treats a plan already prepared under federal requirements as compliance.

Do not skim the phrase "critical infrastructure facility," because it is broader than most people assume. The Act borrows the definition from an existing Montana statute, MCA 82-1-601, which enumerates twenty-two categories. Refineries and water treatment, yes, and also oil and gas wells and pipelines, mining facilities, trucking terminals and railroad yards, cable and broadband infrastructure, telecommunications, and dams. In a state with small oilfield operators, small trucking outfits, and rural broadband providers, "that means power plants" is the wrong summary.

What narrows the obligation is the second half of the test. The facility has to be controlled by a system that makes, or is a substantial factor in making, a consequential decision. Almost no small business is anywhere near that, which is the durable reason this section does not reach you.

The definitions add a second layer worth knowing. A technology that communicates in natural language to provide information, make referrals or recommendations, answer questions, or generate content is excluded from "critical artificial intelligence" entirely, provided it operates under an acceptable use policy prohibiting unlawful content. Note the condition. If you put a customer-facing assistant in place, write that acceptable use policy. It is about a page, and it is the thing the exclusion actually turns on.

Be clear about what this law is and is not. It binds Montana state and local government, not Washington: the Act says plainly that nothing in it preempts federal law. It is no shield against ordinary consumer protection, and the Act names deceiving or defrauding the public as exactly the kind of compelling interest that survives it. It is a standard courts apply when government restricts computing, not a right you invoke at the counter. What it means practically is narrower than the headline and still worth knowing: the direction of state policy here is settled, and it is settled toward the owner rather than against.

Is the window real, or is that just a comfortable story?

The window is real but it is narrower than the framing suggests, and the honest counterweight is in the same Federal Reserve note. Allen observes that in the newer survey data, adoption among the smallest firms is stronger than would be expected based on size alone. Small does not automatically mean absent.

There are three other things I would not paper over.

The first is that a Montana number does exist, and it is not the number this post is built on. BTOS publishes AI estimates by state, and Montana is in them: the Minneapolis Fed's Ninth District figure of roughly 18% is the mean of the published shares for Minnesota, Montana, North Dakota, and South Dakota, with Minnesota leading at 20%. So Montana sits near the national trend at the state level. What those state estimates are not is broken out by firm size within the state, which is the question that actually matters here. The argument in this post is therefore a composition argument: Montana holds an unusual concentration of the size class that stopped moving. That is an inference built from two solid federal data sets, not a measured statistic about small Montana firms, and it should be read that way.

The second is that this kind of advantage does not sit still. Adoption expectations in the same survey ran 20% to 23% for the following six months. A gap that exists in August is not guaranteed in March.

The third is the one I have to say as someone who builds these systems for a living: a tool bolted onto a business that has no process underneath it will not hold. It breaks quietly, usually within months, and the owner concludes the technology was the problem when the actual problem was that nothing was mapped before it got automated. Most businesses do not have an AI problem. They have a process problem, and the process problem is cheaper to fix.

What is the stranger's question test, and why start there?

Ask a general-purpose AI assistant the question one of your customers would ask, then write down exactly what it says about you. That is the whole test, it costs nothing, and you can do it before lunch. Call it the stranger's question test.

The question has to be a real one in your customer's words, not your company name. "Who does foundation repair near Livingston, Montana." "Best real estate broker for ranch property in Park County." Then read the answer as a stranger would.

This matters more than it did a year ago because people have genuinely changed how they ask. BrightLocal's Local Consumer Review Survey 2026, fielded with a panel of 1,002 US adult consumers, found use of generative AI tools for local recommendations rose from 6% the previous year to 45%, making it the third most popular source of business recommendations among those surveyed. Take the size of that jump with some salt, since a consumer panel of that size moves on question wording as well as on behavior. The direction is the part to trust.

The mechanism is worth understanding, because it tells you what to fix. These assistants answer from what they can read about you across the open web and structured business listings. When your information is thin, contradictory across sites, or simply absent, the system does not return an empty answer. It returns your competitor. A wrong phone number in three places is not a cosmetic problem anymore, it is a reason to be left out of a recommendation. The full version of why an assistant does not mention your business is its own subject, but the test above will tell you in ten minutes whether you have the problem.

Do this before you buy anything. An owner who has read that answer with their own eyes makes much better decisions about what to spend money on than one who is working from a sales pitch.

What would I do first, second, and third?

Work in this order, and resist the urge to skip to the third step because it is the interesting one.

1. Find out what the machine already says about you. The stranger's question test, plus a look at whether your name, address, phone, and hours actually match everywhere they appear. This is unglamorous and it is the step with the shortest path to being chosen.

2. Map one process before automating any of it. Pick the one that is costing you customers, usually the gap between someone reaching out and someone getting back to them. Write down what happens now, by hand, including the parts that only work because you remember to do them. What adapting actually means for a small business is mostly this step, and most owners underrate it.

3. Build the smallest system that closes that one gap, and learn to run it. One process. Not your whole business.

The reason for that order is that steps one and two are diagnosis and step three is treatment, and treatment applied without diagnosis is how businesses end up with four subscriptions and no working system.

For what step three looks like when it works: a fuel and petroleum company I built for has an assistant that answers customer questions and takes orders around the clock, structured and routed to the team, without anyone tied to a desk to catch them. A Montana broker's office runs its capture and follow-up on a system she owns and operates herself. Neither of those started with software. Both started with somebody writing down what actually happens when a customer calls.

What should you skip?

Skip anything sold to you on the promise that it runs without you. That promise is the tell.

Skip the full-suite platform purchase when you have one broken process. You will pay for capability you never turn on, which is the most common way small businesses waste money in this category.

Skip the tools that require you to hand over your customer list, your content, and your history with no clear way to get them back out. Whatever you build should be something you own rather than something you rent, because the alternative is running your operation at the mercy of somebody else's pricing decisions.

And skip the timeline anxiety. The data does not support panic and it does not support waiting either. It supports starting small, this month, on one thing.

Frequently Asked Questions

How many small businesses actually use AI right now?

The Census Bureau's Business Trends and Outlook Survey put overall business AI use between 17% and 20% from December 2025 through May 2026, at 19.8% as of May 3, 2026. Reading the April 2026 AI supplement, the Minneapolis Fed reports that 17% of firms with fewer than 20 employees use AI, against 30% of firms with at least 250 employees.

Why do different sources report wildly different AI adoption rates?

Because they measure different things. A Federal Reserve note from April 2026 compared three: about 18% of firms used AI at the end of 2025, about 41% of workers personally used generative AI for work, and 78% of the labor force worked at a firm that had adopted AI somewhere. Firm counts and labor-force shares are not interchangeable.

Is Montana actually behind other states on AI?

No. BTOS publishes state estimates and Montana is in them, feeding a Ninth District average of roughly 18%, close to the national rate. Those state figures are not broken out by firm size, so they cannot say what very small Montana firms are doing. What is measurable is that roughly 88% of Montana employers have fewer than twenty people, and that size class is where national adoption stopped moving.

What is the Montana Right to Compute Act?

Senate Bill 212, passed by Montana's 69th Legislature in 2025 with an immediate effective date, is the nation's first right to compute law. It requires that government restrictions on privately owning or using computational resources for lawful purposes be demonstrably necessary and narrowly tailored to a compelling government interest.

Does the Right to Compute Act require my business to file anything?

Almost certainly not. The Act requires a risk management policy, never a filing, and only where a critical infrastructure facility is controlled by an AI system that makes or substantially factors into consequential decisions. Montana defines critical infrastructure across twenty-two categories, from refineries to broadband, but that consequential-decision test keeps the duty away from ordinary businesses.

What should a Montana business do first?

Run the stranger's question test. Ask a general-purpose AI assistant the question a customer would ask about your kind of business in your town, and read the answer as a stranger would. It costs nothing, takes about ten minutes, and tells you whether you have a visibility problem before you spend anything.

Why does being found by AI assistants matter for a local business?

Because the way people ask has changed. BrightLocal's Local Consumer Review Survey 2026, fielded among 1,002 US adult consumers, found that use of generative AI tools for local recommendations rose from 6% the previous year to 45%, making it the third most popular source of business recommendations among those surveyed.

Is it too late to start?

The data says the opposite. Adoption among firms with fewer than twenty employees did not change significantly between December 2025 and May 2026, so the ground most Montana businesses stand on is still open. Expectations for the following six months ran 20% to 23%, which is why starting small now beats planning big later.

---

Pick the one process that is costing you customers this month, and write down what happens now, by hand, before you look at a single tool. That document is worth more than any subscription you could buy this week.

If you want help figuring out where this actually fits in your business, Bennin Systems can map the system with you before you spend money on tools.

---

Stacy Bennin is the founder of Bennin Systems, where she builds the automated systems small businesses need but rarely have time to set up themselves: lead capture and follow-up that runs on its own, chatbots that answer questions and take orders around the clock, custom websites that act as an employee, and the back-office workflows that keep an operation from running on memory and sticky notes. Located in Montana, she works with businesses and real estate professionals anywhere in the United States. She is also a licensed Montana real estate broker affiliated with Legacy Lands Real Estate. Reach her at benninsystems.com.

Bennin Systems, Paradise Valley, Montana. (406) 224-3267. benninsystems.com