If you worry someone may monitor your phone, computer, or email, use a device and account only you control. Press Esc or the Quick exit button to leave this site at once.
Articles · Using AI in Your Business

Before you use Claude to start a business, ask it whether the business will work

Most new businesses don’t fail because a lawyer botched the startup paperwork. We lawyers come later in the process. New businesses fail over money: too few customers pay enough, often enough, to cover the bills. About one in five closes in its first year. About half are gone within five.

By Erin Conroy · November 11, 2026 · About an 8-minute read · Part 1 of a series on using AI to start a business

The short version
  • Most new businesses fail over money, not paperwork.
  • One evening with an AI assistant can tell you whether your idea can pay.
  • Skip the reading and grab the five-step test.
1 in 5
new businesses close in the first year
1 in 2
are gone within five years

Source: U.S. Bureau of Labor Statistics business survival data.

Here’s the truth: the magic of a new business isn’t in the attorney paperwork. Sure, we handle the process of standing your company up as a going concern, but the legal paperwork only makes you street-legal: the title, the plates, the insurance card. The guts and the engine make the business, and those come from you.

MN 1 2
1

The paperwork. Title, plates, insurance card. It makes you street-legal.

2

The engine. Customers, cash, and you. It makes the business go.

This article shows you how an AI assistant such as Claude can help you find out, in one evening, whether your idea can make money. Do it before you sign formation documents or commit to a $100,000 line of credit. I’ll give you the questions to ask, with prompts you can copy.

Viability comes before anything else

Forming an LLC in Minnesota costs $155. A company vehicle, a lease, inventory, a bookkeeper, or a year of your life costs far more. Plan the business before you pour your time, your talent, and a lot of treasure into it, only to discover the market was already full.

The legal steps protect a business from legal exposure and keep it square with federal, state, and county rules. Before you get there, answer the first question: is this a good business idea? Then break it down further:

  • Does my market hold enough customers, and enough room, to support it?
  • Will customers buy my product or my service?
  • If nobody around here does this, why not? There may be a good reason no one has tried.

Claude can’t answer those questions for you. It can help you ask them in an organized way and spot the holes you can’t see (or don’t want to see).

The agreeability trap

One warning before you start: AI tends to be agreeable. Ask “Is this a good idea?” and you will usually hear yes. These tools are built to be helpful, and helpful slides into flattering faster than you’d think. To stay out of the agreeability trap, tell your AI to disagree with you, push back on you, and show you where you’re going wrong.

Here’s a habit I borrowed from my day job. When I prepare for a hearing, I often pull the judge’s earlier public opinions and the public transcripts from the same cases. I read both and write down what seemed to matter most to the court. Does this judge care more about procedure or substance? Only after I’ve committed my own read to paper do I run the same documents through Claude, and I don’t tell it what I concluded. I let it reach its own read. Then we compare notes. Where we differ, one of us missed something, and I want to know which one.

The lesson carries straight over to your business idea. Form your own answer first, write it down, and keep it to yourself. Make the AI do its own work before it sees yours. It can’t flatter an opinion it hasn’t seen.

An AI can’t flatter an opinion it hasn’t seen.

How I use it

I drink a lot of coffee. I don’t mean a cup or two in the morning. I mean probably-a-problem amounts of coffee. I’m also an early riser, so I spend my wake-up time either outside listening to the birds or in my chair, catching up on the news and toggling between Claude and my latest business idea.

Before you ask, I already know the question: “But wait, aren’t you a lawyer?”

Why, yes, I am. But I didn’t start out wanting to be a lawyer. I wanted to be a designer and an art professor, and these days I feed that creative hunger with new business ideas. I think about branding, packaging, and clever marketing. My husband has heard more sales pitches than the investors on Shark Tank.

Not long ago, I decided our rural Minnesota front yard would make the most fantastic Halloween haunted attraction. Too bad for my husband, he agreed, and that launched me into an entire business plan: how to lay out the yard, a theme for each year, and a business model. I even had a staffing idea I still think is brilliant: cast the local theater students as the actors, with a new theme for them every year.

Okay, but aren’t there a lot of haunted farms, pumpkin patches, and corn mazes around? This is Minnesota, after all. How would I know whether the idea was worth pursuing? Where would I even start? I couldn’t exactly walk up to the other businesses and ask for their numbers. That would be rude, and not very neighborly.

Enter Claude. Or ChatGPT. Pick your AI flavor; there are plenty of good ones. I use Claude because it was the best on the market when I got serious about AI. I have no idea whether that’s still true, but I’m in deep now, so I’ve hitched my proverbial wagon to it. Come on in, the water’s fine.

Here are the starting questions I asked about the haunted attraction:

  • How many attractions like this already operate within 50 miles?
  • How far away is each one? Map out the radius.
  • If we sit close to a competitor, can we design our five acres to be different enough that we don’t compete?
  • Can our gravel road handle the traffic? What rules does the county have for a seasonal business like this?
50 miles 25 miles Us
The first question

Who already does this within 50 miles, and how close is the nearest one? Draw the circle before you draw up a business plan.

Illustration only, not real data.

In the end, I concluded the concept was viable and we could proceed. We also had a full year to convince the neighborhood we wouldn’t ruin their lives with traffic, outbuildings, theater kids, and loud screaming.

Later, I had Claude run the numbers. What will the overhead run? What permits will we need? Eventually it laid out an entire business plan, complete with assignments, due dates, barriers to entry, and legal problems to solve (traffic! neighbors! zoning!).

Full disclosure: we are never going to build it. My neighbors can relax. Building it was never the point. The point is that one morning of questions took a daydream and told me whether it could work, what it would cost, and what stood in the way. That hour is the cheapest step in starting a business, and most people skip it.

One hour of questions is the cheapest step in starting a business. Most people skip it.

Run it on your own idea

You don’t need to copy my questions. I packaged the whole session into a download, the Business Idea Test. It comes two ways: a skill file you can add to Claude, and a plain prompt you can paste into any AI assistant. Either one walks you through five tests, one at a time:

  1. Describe it honestly. It interviews you about the idea, your money, your time, and what you don’t know how to do. Think of AI as the world’s most literal assistant: quick, eager, and clueless about anything you leave out. Give it the facts you’d rather skip.
  2. Argue against it. It plays a skeptical lender and lists the five most likely reasons the business fails. This is how you climb out of the agreeability trap.
  3. Run the numbers. It builds a break-even table: how many sales cover your costs, and how many also pay you.
  4. Design a cheap test. Three ways to find real paying customers for under $200 in under 30 days, with the passing score set before you start. A test you grade after the fact isn’t a test. It’s a participation trophy.
  5. List the legal questions. Not answers. Questions, sorted by who can answer them: your city, the state, an insurance agent, a lawyer.
Download

The Business Idea Test

One guided session, five tests. Pick the version that fits the AI assistant you use.

Skill file for Claude (.zip) Plain prompt for any AI (.txt)

The skill file adds to Claude as a skill. If your plan doesn’t offer skills, or you use another assistant, open the plain prompt and paste it into a new chat.

Meet Jordan

Here’s what the test looks like in practice. Jordan is a fictional 24-year-old in Isanti County. He wants to start a mobile car-detailing business: drive to the customer, clean the car in the driveway, and charge $150 for a full detail. He has $4,000 in savings, a pickup, and his weekends.

The first test turned up a problem right away: where does the water come from at a customer’s house in January? I’d add a second one, because at our house the water is so hard you could slice cheese with it. If Jordan hauls his own water in a tank, does that change his license, his permits, or his insurance? Onto the list they go.

Then came the numbers:

Price per full detail$150
Supplies, fuel, and card fees per detail−$25
Left over from each detail$125
Fixed monthly costs (insurance, equipment payment, phone, ads)$900
Details per month to break even ($900 ÷ $125)8
Fixed costs plus $3,000 a month in pay for Jordan$3,900
Details per month to also pay Jordan ($3,900 ÷ $125)32, about 8 a week
8 details a month: the bills get paid
32 details a month: Jordan gets paid, too

Each square is one $150 detail. Each row of eight is one week.

Eight details a week means more than one a day, six days a week, all year. Including January. In Minnesota. That changes the plan. Jordan can raise his prices, add a cheaper service customers buy more often, or treat the business as weekend income for the first year. Any of those can work. The point is to see it before he spends the $4,000.

One habit to keep: check the AI’s arithmetic yourself. These tools make math mistakes and state them with total confidence.

What Claude can’t do for you

  • It doesn’t know your local market. Claude may not know your town, your competitors, or current prices, and it can be confidently wrong. Check local facts against a map and a phone call, and talk to real customers. Call your local chamber of commerce or economic development office, too, and ask what data they will share about your market and other startups in the area.
  • It can’t promise success. Good analysis lowers your risk. It doesn’t erase it.
  • Garbage in, garbage out. AI won’t make your business succeed, and it won’t keep hardship away. It can help you forecast, and a forecast is only as honest as the numbers you feed it. Fudge the inputs and you’ll get a very confident wrong answer.
  • It isn’t your lawyer or your accountant. Use it to prepare questions and understand terms. Make legal and tax decisions with a professional who knows your facts.
  • Protect private information. Don’t paste Social Security numbers, bank or card numbers, passwords, or another person’s private details into any AI tool.

Human help at no cost

Minnesota’s Small Business Development Centers offer one-on-one business consulting at no cost, with more than 30 offices around the state. Bring your Claude work to that meeting. A consultant who sees your break-even table and your test results can help you far faster than one starting from scratch.

Start your engines

If your numbers hold up and real people will pay, it’s time to build the business properly: the structure, the state filings, taxes, licenses, and insurance. Our guide, Starting a business under 40 in Minnesota, walks through each step. The next article in this series will cover how to use Claude to prepare for your first meeting with a business lawyer, so you spend that hour on decisions instead of definitions.

Ready to form your business? Rum River Law forms Minnesota and North Dakota businesses for a flat fee quoted before the work starts, by phone, by video, or in person. Every formation includes a Company Book with your records, renewal dates, and license deadlines in one place.

Claude is an AI assistant made by Anthropic. Rum River Law has no affiliation with Anthropic, and the same approach works with other AI assistants. AI tools change often, so features may differ from what this article describes. Jordan is a fictional example, and the numbers are illustrations, not market data. Business survival figures come from U.S. Bureau of Labor Statistics data. This article is general information as of November 2026 and not legal, tax, or financial advice.

Tested the idea? Let’s build it right.

A short call tells you what your business needs now, what can wait, and what it will cost. Call or email to schedule by phone, by video, or in person.

Call 763-310-8339