The only vendor in the room

A finance director has an audit in six weeks. A software company shows her a tool that automatically matches her transactions to bank statements, and nobody else sells anything like it. By the usual advice, that company is in the best spot a seller can be in.

She still has other ways through the audit. Her team can stretch the spreadsheet they already use. She can bring in a temp, or ask her accountant to take on more. She can also wait and look at software after the audit. Waiting adds no new cost this quarter, and it spares her from setting up a new system in her busiest weeks. If new software goes wrong now, she’s the one who picked it.

So the company really is the only one selling that tool, and it’s still one of five ways she could deal with the audit. Four of those five don’t involve buying anything from that company.

I’ve used the idea of being the only one in my own work. One of my tests asks, “What single noun do customers associate with you alone?” Over the last few weeks I’ve been working out why that word feels right inside a company and so rarely decides anything for a buyer.

Where the word comes from

In 1961 Rosser Reeves set out the Unique Selling Proposition. One of his conditions was that the claim must be one “the competition either cannot, or does not, offer.” In 2014 Peter Thiel told founders in Zero to One to aim for monopoly. Two years later the authors of Play Bigger reported that “category kings” held 76% of the market value in their categories. That figure measures market value, and the kings were picked after they had already won.

All three look at the market from the seller’s side and measure a company against its competitors. None of them starts with a buyer and what that buyer does today.

Thiel saw part of the problem. He wrote that companies without a real monopoly exaggerate how different they are by drawing their market very narrowly. His example was a British restaurant in Palo Alto. It may be the only one in town, and it still competes with every other place to eat there. A buyer draws the market wider again, around their problem and every way they could deal with it.

There’s a good reason to keep the idea inside the company. Choosing one thing to be known for helps a team decide what to build and what to turn down. A company that turns down a lot is also harder to copy. The trouble starts when the word moves from the investor deck to the buyer. Investors like it because it suggests the company can hold its prices. My guess, which I haven’t tested, is that some buyers read it as a sign that few people have tried the product yet.

What “only” can mean

When people say a company is the only one, they can mean at least five things. The first is about the market: nobody else sells this. The second is about the company. It has chosen one idea to organize around, and that idea guides what it builds and what it turns down. The third is about memory. When buyers think of a need, this company comes to mind first, or more strongly than others. The fourth is about the choice. The buyer considers this company and nobody else, or buys only from it. The fifth is about price. There’s no close substitute, so the company can charge more without losing many customers.

Most claims of being the only one start with the first meaning and assume the fourth. I don’t think that step holds. Economists use the word monopoly for a firm with no close substitutes and some power over price. Its buyers can still go without or solve the problem another way. Even Google, which a US court ruled in 2024 to be a monopolist in search, paid more than $26 billion in 2021 to be the default search engine on phones and browsers. It paid to be the option already in front of people.

Volvo is the usual example of a company that owns an idea in buyers’ minds, in its case, safety. In 2024, it sold a record 763,389 cars. That was under 1% of the roughly 89 million light vehicles sold worldwide. And when someone buys a Volvo because the dealer offered a good discount, the sale tells you little about safety. So a position can shape a decision without making it. One sale can’t prove or disprove a position, and neither can market share.

Everything the buyer can do

Here is what a buyer can do about a problem like the finance director’s. They can keep doing the work by hand, or build a better spreadsheet. Someone on the team can take it on, or they can hire a contractor for a few weeks. They can ask a supplier they already pay to add it, or buy a cheaper tool that covers most of it. They can change the process so the problem gets smaller, or put it off until the next budget. They can buy less of something they already have, or spend the money on a problem that feels more urgent.

And now they can ask an AI tool to help them build something, which I’ll come back to.

One item on that list is the vendor’s product. Most of the rest never appear on a seller’s list of competitors, because they are habits, people, budgets and timing.

Buyers also choose from short lists. Hauser and Wernerfelt reported in 1990 that people seriously consider a median of four shampoos, and two to five cars. In business buying, a Bain and Google study in Harvard Business Review in 2022 found that about 90% of buyers choose a vendor that was on their first shortlist. Google sponsored that work, so I treat it as a strong hint.

So a seller is competing for a place on a short list that already includes doing nothing new. A company can be the only one of its kind and still be left off the list, or sit on it next to a spreadsheet the team already trusts.

Why staying put makes sense

Carrying on as before is on every buyer’s list. Samuelson and Zeckhauser, who named status quo bias in 1988, gave three kinds of explanation for it. Some of it is rational, because change has costs and the result is uncertain. Some comes from perceptual errors, such as weighing losses more heavily than gains. Some comes from psychological commitment, such as not wanting to waste what was already spent. In the data they cited, only about 3% of Harvard employees changed health plans in a given year.

Those explanations need different responses from a seller. Sunk cost is what has already been spent on the current way of working. It shouldn’t affect the decision, though it often does. Switching cost is the real future cost of changing: moving the data, training people, running two systems for a while, and the risk that something breaks. A buyer who weighs that carefully is making a sound decision.

Blame matters too. Ritov and Baron found in 1990 that many people would not give a child a vaccine that carried a small risk of its own. They held back even when that risk was lower than the disease risk. If the finance director buys the new tool and the audit goes badly, she chose it. If she keeps the spreadsheet and the audit goes badly, the spreadsheet was already there.

Herbert Simon had a word for how people choose under limits like these. He called it satisficing: looking for an option that is good enough, given the time and information available. For a buyer, good enough includes the effort of changing, who has to defend the choice, and what happens if it goes wrong.

None of this makes staying free. The manual work and the mistakes continue. Those costs are familiar and spread across the year, so they are easy to live with.

Why buyers rarely see an “only”

Most buyers don’t see much difference between brands. Romaniuk, Sharp and Ehrenberg looked at 17 markets in 2007. On average, about 10% of a brand’s current users called it unique, and 17% called it either different or unique. That study asked people directly, so it may miss associations they hold without putting them into words. In business buying, a 2013 CEB study with Google found that only about 14% of buyers saw enough difference between suppliers to pay for it. Google sponsored that work too.

New products have it harder. John Gourville of Harvard Business School offered a rule of thumb in 2006. Buyers overvalue what they already have by about three times, and sellers overvalue their new product by about three times, which makes a gap of about nine to one.

Some buyers do respond to being first. Everett Rogers’ model of how new ideas spread puts innovators and early adopters at about 16% of a market, by definition. A pitch built on being the only one may work with that group. Most of the rest want to see people like them go first.

For everyone else, a familiar name lowers risk. When Ted Roselius asked consumers in 1971 how they reduce purchase risk, brand loyalty and buying a major brand ranked first and second. The old line that nobody ever got fired for buying IBM says the same about business buying.

Buyers also make their own version of “only.”

Many companies cut their supplier lists to get one contract and one support line, usually with a supplier they already use.

The newest option

Many teams now use Claude or a similar tool for work a software subscription used to do. The question I keep hearing is simple. If Claude can do this in a few minutes or a few hours, why are we still paying for the subscription?

Some early numbers back this up. In late 2025, Retool surveyed 817 people who build software inside companies, including some Retool customers. Of those, 35% had replaced at least one SaaS tool with something they built, and 78% expected to build more in 2026. The tools replaced most often were workflow automations (35%), internal admin tools (33%) and business intelligence (29%).

MIT’s NANDA project reported in 2025 that workers at more than 90% of the companies it surveyed used personal AI tools for work. Only 40% of those companies had bought an official subscription. Anthropic’s own study of 100,000 Claude conversations estimated that the tasks would have taken about 90 minutes without AI, and that Claude cut the time by about 80%. That leaves out time spent checking the work, and the authors say it may be high.

In February 2026, Anthropic released plugins for Claude Cowork. That week, ABC News reported, Thomson Reuters and LegalZoom shares fell more than 15%, and RELX and FactSet had double-digit drops. Some of that was later recovered.

For the finance director, this adds one more way through the audit. Someone on her team could spend an afternoon with Claude and build a rough matching script. It would probably be worse than the vendor’s tool, and it might cost nothing extra.

What the subscription still offers

The same surveys show why many teams still pay. In Retool’s survey, 41% named security and compliance as a blocker to building, 26% named the work of maintaining what they build, and 22% named AI producing wrong code or data structures. Only 8% used AI-written code without changing it. MIT found that AI projects done with outside partners reached real use about two-thirds of the time, against about one-third for projects built in-house. That’s self-reported and doesn’t prove causation. Menlo Ventures, which invests in AI companies, found that US enterprises bought 76% of their AI use cases in 2025, up from 53% the year before.

Overall software spending hasn’t fallen. Zylo’s 2026 index found average SaaS spend per company up 8% on the year, with the number of apps flat at about 305. Spending on AI-native apps rose 108%. So far the money seems to be moving between vendors more than leaving.

Klarna shows how easily this gets misread. In August 2024, its chief executive said, “We just shut down Salesforce. Within a few weeks, we will shut down Workday.” In March 2025, he clarified, “So no, we did not replace SaaS with an LLM.” Klarna had consolidated its systems.

So a subscription has a good answer to the question. It can show who fixes the tool when it breaks, who keeps it secure, who updates it when the rules change, and who picks up the phone in audit week. A script built in an afternoon has none of that, and many buyers know it. The vendor’s job is to make that answer easy to see before the buyer asks.

When building gets cheap

Cheap building changes what a product’s features say about the company behind them. Michael Spence in 1973 and Amotz Zahavi in 1975 described the same idea, in economics and in biology. A signal is believable when it’s expensive to fake. If a team can build a feature in an afternoon, having that feature tells a buyer little about the vendor’s commitment. What stays expensive is being accountable when something breaks, years in the market, data nobody else has, and the work a company chooses not to do.

The homemade tool also has its own pull. Norton, Mochon and Ariely found in 2012 that people would pay about 63% more for an IKEA box they had built than others would pay for the same box. People valued their own origami at about 23 cents, while others valued it at about 5 cents. The effect disappeared when people didn’t finish the build. If that carries over to AI-built tools, the people who built a working tool will defend it even if it’s worse than the vendor’s. A failed build won’t get that loyalty. My inference is that a vendor’s best chances come before the team finishes, or after their tool breaks.

Price moves too. In value-based pricing, as Thomas Nagle describes it, buyers judge price against the cost of their next-best alternative, plus the value of whatever is different. If that alternative is an afternoon with Claude, the reference point drops for every seller. Meanwhile, Tropic, which helps companies buy software, reported vendors asking for AI-tied price rises of 20% to 37% at renewal, up from the usual 3% to 9%. After negotiation, the rises came to about 12%. My guess is that these two pressures meet at renewal.

Who decides now

AI also changes who decides. Here is what I think is happening, and how sure I am of each part. The buyer is splitting into more people. MIT’s gap between personal AI use at more than 90% of companies and official subscriptions at 40% is one sign. Retool’s finding that 60% of builders had built something outside IT oversight is another. My inference from those numbers is that one employee can now replace a tool without anyone signing a purchase order. A seller who only talks to procurement may never meet the person who made the decision.

Waiting looks more sensible than it used to. Dixit and Pindyck showed in 1994 that when a decision is hard to reverse and the future is uncertain, waiting has a value of its own. If AI tools get noticeably better every few months, a buyer who puts off a three-year contract may be making a good call.

Switching may get cheaper. If AI can help move data and rewrite the links between systems, lock-in weakens for every vendor, including the newest ones.

The career risk may be changing direction. Buying the known vendor used to be the safe choice for the person signing. In some companies, building something with AI may now be the move that gets noticed. Retool’s respondents gave reasons such as speed (31%) and existing tools falling short (25%), and those say nothing about careers.

The shortlist may be written by an AI assistant. A buyer who asks an assistant what to use may get a list of vendors and an offer to build it instead. You could test this by asking the same buying questions across several models and recording which brands come up and how often the answer is to build it.

What I’m changing

I’m keeping part of the idea. A company should choose one thing to be known for and organize around it, including what it refuses to do. Carpenter and Nakamoto showed in 1989 that the first brand in a category can shape what buyers come to prefer. Later brands then find it hard to win share, even when switching costs little. It’s a real advantage in memory, and it only pays off if buyers adopt the category at all.

What changes is how I describe it and how I test it. A business can commit to one clear position while its customers keep several credible alternatives. The work is to earn that meaning through consistent choices, and to find out when it matters to the buyer. Some buyers will hold it strongly, and it will matter most in the situations where it’s relevant.

That leads to five changes. First, I’m removing “alone” from my test. The better question is which idea buyers link with you more strongly than with anyone else. Second, measure that among buyers, including people who don’t buy from you. Count how often you’re named first for the need. Then show buyers the same product under your name and under a rival’s, and see which one they rate higher on the idea. Third, start sales conversations with what the buyer does today and what would make them stop. Fourth, record deals lost to “we built it ourselves,” “we kept the spreadsheet,” and “we decided to wait,” next to deals lost to competitors. Fifth, if you sell software, show what’s expensive to fake: who fixes it, how long you’ve done this, what data you hold and what you won’t do.

The finance director will probably look at the tool again after the audit. By then someone on her team may have built a rough version with Claude. It will be slower than the vendor’s and will need checking every month, and the team will probably like it more than it deserves. When the vendor calls back, that’s what it will be up against: a script in a shared folder with her team’s names on it.