Gartner: only 13% of CIOs see significant AI value. The missing ingredient is project discipline

Most AI pilots are like gym memberships. They begin with enthusiasm in January, they are paid for faithfully every month and very few people can say what they have achieved. The difference is that a gym membership costs thirty dollars, and an AI initiative can cost enough to make the finance team develop a facial twitch.

On October 19, at its IT Symposium/Xpo in Orlando, Gartner put numbers on that feeling. The research firm released results from its 2027 CIO and Technology Executive Agenda survey: 2,002 CIOs and technology executives, surveyed from May to July 2026. The headline is blunt. Only 13% of them say their enterprise has realized significant or transformational value from AI so far. Gartner analyst Alicia Mullery summed it up with a line that deserves a poster: "Volume is not value."

I am not a CIO, and I do not run a Fortune 500 company. I help startups, clinics and builders run projects. But this survey is about project discipline wearing a very expensive suit, and that is territory I know well. Here is what it says, what it does not say and what a small team can take from it.

What the survey found

These are the figures as Gartner and the coverage of its release report them.

• 13% of CIOs say their enterprise has realized significant or transformational value from AI so far. The other 87% are somewhere between "some value," "not yet" and "ask me again next quarter."
• 31% have rules for stopping an unsuccessful AI pilot before it moves into production.
• 17% have criteria for abandoning low-value AI that is already in production.
• Gartner uses the word "vanguards" for the CIOs whose organizations deliver better business outcomes and whose CEOs perceive higher AI value. Having guidelines to keep unsuccessful pilots from reaching production makes a CIO 2.3 times more likely to be a vanguard, and having criteria for abandoning low-value AI in production makes it 2.6 times more likely.

Coverage of the release is careful to say that this shows an association. It does not prove that stopping rules cause better returns. Maybe disciplined organizations are simply better at everything, including writing stopping rules. I find that plausible, and also not very comforting: it means discipline is the rare ingredient.

The boring truth behind the numbers

Read those four bullets again, and you will notice that none of them is about technology. Not about the model, the vendor, the chip or the prompt. They are about deciding what success means, measuring it and having the courage to stop. These are project management basics, the same ones that apply to a building, a clinic opening or a software launch.

• A business case. What problem are we solving, which number should move, what will it cost and what could go wrong? I walk through one in the first part of my PMBOK series, including a question that belongs in every proposal: what would make us stop?
• A baseline and a measure. If you do not know how long the task took before, you cannot know whether the machine saved time. This is what benefits realization is about.
• A stage gate. A point where the project must earn the right to continue. A pilot that has no end date and no exit rule is not a pilot. It is a pet.
• A risk list. AI projects fail for old reasons (an unclear goal, no owner) and new ones (confident wrong answers, private data in the wrong place). A risk register covers both.
• An owner. One person who answers for the result, not "the AI team."

The stopping rule is the one that makes people uncomfortable, for good reasons. Stopping means admitting that an idea did not work, that money was spent and that somebody's favorite project is finished. Nobody enjoys that meeting. That is exactly why a rule written in advance, when everyone is calm and optimistic, is so much easier than an argument held in the middle of a failing pilot.

Four habits for a small team

You do not need Gartner's budget to use this lesson. Here are four habits for any team, from two people to two hundred. They are my reading of the survey, not Gartner's recommendations.

1. Write the stop rule before you start. One line: "If after 30 days the draft replies still need heavy rewriting, or if we see more than three serious errors, we pause and review." Put a date next to it.
2. Record the baseline. How many minutes does the task take today? How many errors? How many complaints? Write it down on day one, because it will not be recoverable on day thirty.
3. Give it one owner and one review rhythm. A person, a weekly sample review and a mistake log. Not a committee.
4. Decide keep, fix or drop at 30 and 90 days. The benefits tracker is made for this. Dropping is a result too: it frees time and money for the next idea.

If you want a step-by-step way to choose a first, safe AI project in a small business, I wrote AI for small business: what to automate first, and I am starting a weekly series on the same theme, AI for small business, step by step. For the people side of change, see change management and the people side of projects.

What this survey does not tell us

A few honest caveats, because a news post should not claim more than its sources.

• It surveys CIOs of large organizations. Small businesses have different problems and often faster feedback. A bakery can tell in a week whether a reminder system works.
• It measures perception. "Significant value" is what the executives report, not an audited figure.
• It shows an association. Organizations with stopping rules may differ in many other ways.
• I have not read the full research, which sits behind Gartner's paywall. I am working from the public announcement and from news coverage of it.

Still, even with those caveats, the direction is useful. If the people who manage the largest technology budgets in the world are struggling with the same basics we all struggle with, the right move for a small team is not to be intimidated. It is to copy the basics: a number, an owner, a review and a stop rule.

What I am watching

Two questions. First, whether the share of organizations with stopping rules rises in the next survey, since a rule that is written and used is a sign of a maturing practice. Second, whether the value gap closes because the technology improves, because organizations get more disciplined, or both. My bet is on discipline, but I would happily lose it to better software.

If you would like a hand turning a vague AI idea into a project with a number, an owner and a stop rule, that is what I do with Project Focus, and you can write to me through the contact page.

Does your team have a rule for when to stop an experiment? Tell me in the suggestion box on the blog.

Sources

• Gartner press release, Oct 19, 2026
• eWeek: Only 13% of CIOs see major AI value
• IT Brew: Few CIOs seeing transformational value from AI
• AIwire: Gartner on AI value, Oct 20, 2026

Facts come from the sources above. The reading of those facts is mine.