Why Most Companies Are Leaving AI (And What It’s Doing to AI Culture)

Most companies are giving up on intelligence and this is what is happening to the artificial intelligence culture.

I looked at twelve research reports and studies and the pattern is clear: artificial intelligence is not failing because the technology is weak it is failing because companies are not built to sustain intelligence.

When artificial intelligence exploded in 2023 executives everywhere rushed to fund pilots. They were doing things like chat-bots, copilots, auto-summarizes and artificial intelligence agents. Now the excitement is over.

Five percent of companies are actually making money from artificial intelligence per the Boston Consulting Groups global study of over one thousand two hundred and fifty firms. This is not a story about some projects that failed. This is a retreat from artificial intelligence and it is creating a deep and lasting scar on the artificial intelligence culture inside companies.

Below I break down why companies are walking away from intelligence the real root causes behind the failures how this is reshaping the artificial intelligence culture and what the five percent of successful companies are doing differently.

One the big picture is that artificial intelligence is in the trough of disillusionment. Gartner officially placed intelligence and synthetic data into the trough of disillusionment in its 2025 Hype Cycle. That is the phase where hype peaks, expectations crash and projects get skeptical. At thirty percent of artificial intelligence projects will be abandoned after proof of concept due to poor data quality inadequate risk controls, escalating costs and unclear business value.

This is not a moment when the technology’s bad it is a moment when we rushed in without foundations. Two why companies are leaving intelligence the top seven reasons are: vague goals and a “just do something with artificial intelligence” mentality, weak data and broken process foundations, misalignment with business goals, wrong metrics, measuring accuracy instead of impact skills gaps and unmanaged change cost explosion and unclear return on investment and security compliance and ethical risks.

Most companies adopted intelligence because it was trendy or because investors pressured them not because they had a clear problem to solve. The Massachusetts Institute of Technology's report finds that most budgets are concentrated in sales and marketing pilots and the return on investment is lowest there. Back office automation actually produces the returns. Few prioritize it.

Many leaders say “let us experiment with intelligence” instead of “here is the specific business problem the metric we will move and the workflow we will redesign”. Artificial intelligence does not fail in the lab it fails in the company when it collides with goals and organizational inertia. You cannot layer intelligence on top of broken workflows and garbage data and expect magic.

Over eighty percent of intelligence projects fail because they are misaligned with business goals. The common pattern is that tech teams build models. Business teams do not see the value and no one owns the outcome. The Harvard Business Reviews report emphasizes that “most artificial intelligence initiatives fail not because the models are weak but because companies are not built to sustain them”.

Without incentives, redesigned decision processes and an artificial intelligence ready culture even advanced pilots will not become capabilities. Companies obsess over model accuracy, precision, recall and benchmark scores. They forget to track did revenue increase did cycle time drop did headcount or outsourcing costs decrease and did customer satisfaction improve.

The Harvard Business Review and others note that “it is similar to measuring how sophisticated your oven is than whether customers enjoy the food it produces”. If you are not measuring business outcomes you will never prove return on investment. You will kill the project. The risk is not a lack of experimentation difficulty translating ambition into outcomes.

Change management is a determinant of success. It is one of the most under invested areas. Most organizations deploy intelligence into workflows without preparing the people expected to work alongside it. The Massachusetts Institute of Technology's report echoes this saying “ninety five percent of intelligence projects are being worked by the wrong people”.

The problem is not the intelligence it is the wrong team composition, lack of domain expertise and no training on how to use artificial intelligence effectively. Executives expected wins. Instead they got rising compute costs, complex integrations, customization headaches and ongoing maintenance. The top obstacles are cost, data privacy and security risks.

When pilots do not show return on investment within six to twelve months budgets get. Projects die. Artificial intelligence introduces risks, such as data security, algorithmic bias and privacy violations. When a single high profile incident happens trust evaporates.

Three, the impact on intelligence culture what happens when companies leave artificial intelligence. When artificial intelligence projects fail at this scale it does not just kill budgets it rewires culture. Inside companies the narrative shifts from “artificial intelligence will transform everything” to “artificial intelligence was a hype cycle let us go back to traditional analytics”.

When artificial intelligence fails companies shut down intelligence projects replace them with traditional analytics and redirect budgets away from artificial intelligence toward information technology infrastructure. This creates a cultural pendulum swing from over optimism to skepticism. New research from the Harvard Business Review describes “ intelligence brain fry” which is mental fatigue from excessive use or oversight of artificial intelligence tools.

When artificial intelligence pilots fail repeatedly business leaders blame tech saying “your artificial intelligence does not work” and tech teams blame business saying “you did not give us goals or good data”. Middle management gets stuck in the middle. This erodes trust. Makes future artificial intelligence initiatives even harder to start.

After high profile failures leaders become risk averse saying “let us wait until artificial intelligence is more proven”. We will revisit artificial intelligence next year”. This creates a culture of hesitation, where no one wants to sponsor intelligence projects and people hide artificial intelligence usage instead of formalizing it. Innovation slows down.

The top six percent of companies are pulling away they are three times more likely to use intelligence for transformation, redesign processes, not just automate and treat artificial intelligence as a growth driver, not a cost play. This creates a two tier culture, intelligence ready companies that are evolving workflows, up-skilling and measurable return on investment and artificial intelligence trying companies that are stuck in pilots, skeptical and burnout prone.

Four what the five percent of companies do differently. The Boston Consulting Groups study of over one thousand two hundred and fifty companies across sixty eight countries finds that five percent are actually making money from intelligence. Seventy percent of intelligence success is people and processes and only twenty percent is technology.

The winners have a business problem a focused use case. They invest eighty percent or more in reshaping key functions. They redesign workflows end to end. They have a time to value of nine to twelve months. They have the mix of skills and domain experts. They use specialized vendor led solutions. They have intelligence literacy programs and they measure business key performance indicators, such as revenue, cost, time and satisfaction.

They have change management as a core part of the program. In words artificial intelligence does not fail because the technology is flawed it fails when it is misapplied.

Five, the path rebuilding an artificial intelligence culture. If you want to avoid becoming part of the ninety five percent failure stat here is what needs to change: start with problems, not tools define the business problem identify the metric you will move and then apply intelligence to it. Then choose intelligence versus automation versus process redesign.

1. First fix data and process foundations. Clean your data. Break down silos. Standardize naming conventions. Make sure your existing workflows actually work before you use intelligence on them.
2. Second measure business outcomes, not model scores. Track the time saved. Track the cost reduced. Track the revenue increased. Track the error rate decreased. Track the customer satisfaction and employee satisfaction. Do not just track the accuracy. Do not just track the F1 score. Do not just track the benchmark rankings.
3. Third invest heavily in people. Change management. Up-skill least twenty five to fifty percent of your workforce. Create intelligence centers of excellence. Train managers on how to lead with intelligence. Make artificial intelligence literacy a baseline skill.
4. Fourth design for load, not efficiency. To prevent intelligence brain fry integrate artificial intelligence into shared workflows, not individual chaos. Give guidance on how artificial intelligence fits into work. Make time for managers to answer intelligence questions. Value work life balance explicitly.
5. Fifth treat intelligence, as rewiring, not rollout. High performers treat intelligence as rewiring. Redesign decision processes. Update incentives. Change how work gets assigned and measured. Embed intelligence into culture not tools.
6. Six, the line. Companies are not leaving intelligence because the technology is broken. They are leaving because they rushed in without foundations. They measured the things. They did not align people, processes and incentives.

People did not think change management was a deal.

They made something called Artificial Intelligence that was too much for people to handle. It caused burnout.

This led to a lot of problems like skepticism and fear of trying things. Now people think Artificial Intelligence is not as great as they thought which will make it harder for companies to use Artificial Intelligence in the future.

For the companies that do it right Artificial Intelligence is really useful.

• It can change the way people work and make them more productive.
• It can also help people work better together and not get too tired.
• The companies that will do well are not the ones with a lot of models.
• They will be the ones that know what problems they need to solve.
• They will be the ones with data.
• They will be the ones with processes.
• They will be the ones with incentives.
• They will be the ones that care about people and use Artificial Intelligence in a way that helps them.

If you are working on Artificial Intelligence in your company do not try to be like the companies that fail.

Try to be like the companies that really change the way they do business.

If you liked this follow me for information, about Artificial Intelligence and how developers can be more productive.

In my article I will explain how I design Artificial Intelligence projects that actually work and get bigger over time.

Why Most Companies Are Leaving AI (And What It’s Doing to AI Culture) was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.