Nobody Wants AI. They Want the Outcome.

That gap is not a technology problem. The models work. The gap is a buying problem, and almost everyone in the deal is in on it: the board that asked for “an AI strategy,” the vendor that slapped “agentic” on last year’s chatbot, and the marketer who put AI in the headline because everyone else did.

For transparency, I co-founded a company that builds AI products, so take that stake into account. But in all my years selling to B2B buyers, I have never heard one ask for a technology for its own sake. They ask for a result.

What people buy when they say “AI”

Ask a B2B buyer what they want and listen past the first answer. A software firm wants a demo request answered in five minutes instead of two days. A behavioral health practice wants fewer empty chairs on Tuesday afternoons. A hotel group wants to know which guests will book direct next year, so it can stop paying a hefty commission to win them back. A real estate team wants to call the seller who is ready this week, not the one who will be ready next spring.

None of those people said “model.” None of them said “agent.” They named a result, a number, and a clock.

AI might be the fastest route to any of those results. It might also be a better form, a cleaner handoff, or a rule in the CRM that a junior admin could build on a Friday. The tool is a choice you make at the end. The outcome is the thing you buy.

The Noun Budget

Here is the trap I see over and over. I call it the Noun Budget.

A Noun Budget is a line item named after a technology instead of a result. “AI transformation.” “Agentic pilot.” “GenAI center of excellence.” Each one sounds like a plan. None of them has a number attached, and most have no single owner whose job depends on it.

Compare two project names:

• “AI customer service initiative.” Who owns it? What counts as done? When do you kill it?
• “Cut first response time on support tickets from 9 hours to 1 by March.” One owner. One number. One date. It might use AI. It might need two more people on the morning shift.

The second one can fail, and that is the point. A project that can fail can also win, because somebody will know which one happened. The first one can never fail, so it can never win. It just renews.

The research backs this up. When RAND interviewed 65 data scientists and engineers about why AI projects collapse, it found that more than 80% fail, twice the rate of normal IT projects. The top cause was not bad data or weak models. Leaders never made clear what problem the project was supposed to solve. Another top cause was teams chasing the newest tool instead of the user’s need. That is a Noun Budget, described by the people who had to build it.

Who keeps the Noun Budget alive?

It would be easy to blame one villain. There are three, and they feed each other.

• The board. IBM surveyed 2,000 CEOs in 2025. Sixty four percent admitted that fear of falling behind pushes them to invest in technology before they understand what it is worth. Only 25% of their AI projects had delivered the return they expected. That is not strategy. It is fear with a purchase order. When a board asks “what’s our AI plan?” the honest answer is often “what do you want to be true in a year?” Most teams answer the first question because it is safer.
• The vendors. Gartner predicts that more than 40% of agentic AI projects will be canceled by the end of 2027, due to rising costs and unclear business value. In the same report, Gartner put a number on the hype. Thousands of vendors claim to sell AI agents. Gartner counts about 130 that do. The rest are “agent washing,” which means putting a new label on old chatbots and automation. By May 2026, Gartner was warning supply chain buyers that relabeled automation raises “the risk of misaligned investments and long-term lock-in.” A vendor who sells you a noun never has to hit a number.
• The marketers. This one is on us. When every SaaS homepage leads with “AI-powered,” the phrase stops meaning anything. It becomes the new “cloud-based.” If your positioning says AI before it says what changes for the customer, you are sounding just like the company you are trying to beat. Buyers can tell. Most of them are already tired of it.

Put the three together, and you get the cycle. The board asks for AI. The vendor sells AI. The marketer says AI. Nobody writes down the number so that nobody can be wrong.

Why the best AI wins look boring

The AI projects that work tend to share one trait. You would not call them AI projects if you heard them described. They sound like “route every inbound lead to the right rep in under five minutes.” Or “flag the invoices that will be paid late before they are late.” Or “draft the first reply to every support ticket so a human can send it in one click.” The AI is in there, doing real work. But the project name is the outcome, and the owner gets judged on the outcome.

PwC’s chairman, Mohamed Kande, made the point in Davos this year. AI moved so fast that people “forgot that with the adoption of technology, you have to go to the basics.” He called it a matter of execution, not technology. Basics means a clear problem, a clean process, and someone who owns the result.

There is a lesson in what you stop doing, too. I wrote a while back that the real test of AI is whether you can name one thing you deleted because of it. A step, a report, or a meeting that no longer has to exist. That is an outcome. You can count it.

The pitch worth losing

Anyone who sells services has had this call. A prospect wants to “do something with AI in marketing.” You ask what number they want to move. Pipeline? Cost per lead? Time to first meeting? There is a pause, then a long answer that circles back to “we just know we need to be doing more with AI.”

That project will fail no matter who runs it. Not because the team is weak, but because nobody will ever be able to say whether it worked. If you can’t tell whether you won, you didn’t. You just spent.

Walking away from that deal is not a sales tactic. It is the most useful advice you can give for free. Come back with a number, and we can talk about how to hit it. Maybe with AI. Maybe with a better form.

Two tests you can run in five minutes

Pull up your current AI budget, or the AI line in your next quarter’s plan, and run these.
The Swap Test. Take the project name and replace “AI” with “software.” If it now sounds silly (“software transformation initiative”), you are funding a noun. A real project still makes sense after the swap, because the outcome carries the name.

The Three Blanks. Rewrite the project as one sentence: “[Owner] will move [number] by [date].” If you can’t fill all three blanks, don’t buy anything yet. You’re not ready, and no vendor can make you ready.

Then post the result. Tell me what your AI project is called now, and what it should be called. I’ll bet the second name is shorter and scarier, because it can fail.

What to do on Monday

You don’t need a new AI strategy. You need fewer nouns.

• Rename every AI line item as an outcome with an owner, a number, and a date. Kill the ones that can’t be renamed.
• Ask every vendor which of your numbers they will move and by how much. If they answer with features, they are selling you a noun.
• Strip “AI-powered” from your own headline unless the customer can feel the difference. Say what changes for them instead.
• Measure the thing, not the tool. If the number moves, nobody will care what built it. If it doesn’t, nobody should.

Nobody wants AI. They want the result AI keeps promising and has only started to deliver. The companies that win the next two years won’t be the ones that bought the most AI. They’ll be the ones that knew what they were buying it for.

Jeremy Mays is the founder of Transmyt, a marketing firm for B2B SaaS, health and wellness, real estate, and hospitality brands, and a cofounder of Paciva, which builds AI agents. Transmyt offers fractional CMO and digital strategy work. For more on measuring what matters, read You Are Not Data Driven. You Are Data Justified. and Good Strategy Does Not Fail. It Gets Killed Before It Can Work, or browse more Transmyt Insights.