The global race to develop and deploy artificial intelligence is moving fast on many people.
Nvidia has become one of the most valuable companies in the world, on the back of increasing chip demand. According to Gartner, worldwide AI spending is expected to reach $2.5 trillion by 2026. Wall Street has declared AI one of the defining investment topics of the decade.
And yet, for many companies, the returns are not showing. A landmark MIT study found that 95% of organizations saw zero measurable return on AI investment, despite spending between $30 billion and $40 billion on AI initiatives.
The tools work. Models are eligible. The problem, according to experts who work within these organizations, is almost never technology. These are the people, the culture and the surrounding systems. Here’s what’s really going on.
Many executives treat AI deployment like a software rollout. Buy the equipment, install the system, train the staff. done
This approach fails at scale. Axialent, a leadership consulting firm that works with large organizations on change, has studied this pattern closely. The company argues that companies consistently underestimate the human side of AI adoption, focusing on the technology while ignoring how people are actually changing the way they work.
“AI is being embraced by people, not servers,” Axial CEO Osis Ramirez told TheStreet. “If people don’t change the way they work, the technology will simply sit there.”
Related: Bill Ackman Makes Bold AI Bet
Even when AI productivity tools are fully available, employees frequently use them only for small, surface-level tasks. Deep workflows, decisions, and judgment requests remain unchanged. The technology is available. There is no change.
This pattern is the same. Budgets are going towards models and infrastructure, while the hard work of changing the way people work gets little attention. AI is delegated to technical teams even when the main decisions are strategic. And when experiments fail, as they often do, most organizations are reluctant to push through.
Management hierarchy and incentive systems were developed long before the advent of AI, giving employees little reason to adopt new workflows when performance metrics remain tied to old practices.
Sales teams may receive AI-generated forecasts that challenge traditional quotas, but if compensation systems don’t change, those insights are completely ignored.
Most employers are using AI as a somewhat intelligent search engine rather than a tool that fundamentally changes how work is done.
Organizations that invest heavily in AI models without addressing culture are only seeing tools used for small tasks, with no measurable impact on business outcomes.
Companies that see real results from AI don’t necessarily have the most advanced models. This is what reconfigured how people work around these models.
Other AI Stocks:
This means changing mindsets, rewriting motivational structures, and holding leadership accountable for accepting, not just hiring. Research points to culture as a make-or-break factor when organizations adopt AI. Create a culture, and follow through. Leave it, and the technology gathers dust.
Ramirez said it clearly. “The majority of employers use AI primarily as a somewhat intelligent search engine. The technology is there, but the way people work hasn’t really changed. Companies that invest heavily in human resources rather than technology are seeing stronger results.”
Even when companies successfully deploy AI and drive real-world applications, a new and largely unseen problem emerges: They can’t charge for it properly.
Traditional software pricing is built around subscriptions, seats, and licenses. AI services work differently. Rates are tied to tokens processed, API calls made, or modules executed. Most billing systems were never designed to track this type of spending.
Sayo, a revenue management platform that works with SaaS companies on this exact challenge, has seen the results firsthand. CEO Aries Agmon told TheStreet that the pattern is the same.
“Most SaaS billing systems are designed with predictable subscriptions in mind,” he said. “AI Leads to Misconsumption.”
Companies that invest heavily in human AI adoption, not just technology, see the strongest results, one expert says. Termmee/Getty Images ·Termmee/Getty Images
The result is a revenue stream. Agmon recalled a CFO whose billing system only recorded usage on the day of the billing cycle. If a customer goes up in price in the middle of the month and drops back before the billing date, the spike is completely gone.
As that CFO said: “I was just collecting what was due on the billing cycle. I missed the spike. I missed that money.”
Billing systems built for flat subscriptions can’t track the erratic, consumption patterns that AI products create, creating gaps between usage and invoiced revenue.
Finance teams resort to exporting usage data in spreadsheets, manually reconciling across platforms, and generating invoices by hand—workarounds that are breaking down as AI adoption scales.
Companies that fail to get usage right struggle to realize the value of their product, making informed pricing decisions nearly impossible.
Income leaks compound over time: Small gaps per customer in a large base can represent hundreds of thousands of dollars in annual lost revenue.
Companies that go forward treat monetization as a core product design decision, not an afterthought for the finance team to sort out. Those who move fast enough to build a billing infrastructure capable of accurately tracking AI consumption will have a structural advantage over those who are still reconciled with spreadsheets.
Numerical stacks are concrete. SaaS companies typically lose between 0.25% and 2% of their annual recurring revenue to billing gaps alone. For a company with $20 million to $50 million in ARR, this translates to $250,000 to $600,000 in lost revenue per year.
History consistently shows that access to technology rarely determines which companies win. The advantage goes to organizations that quickly adapt their internal systems to new tools.
In the era of AI, this principle is sharper than ever. MIT’s own research found that the companies that succeed with AI aren’t the ones with the most advanced models. They are the ones who pick a pain point, execute well, and integrate AI deeply into existing workflows instead of running discrete experiments.
Those who skip these steps may find that implementing AI is the easy part. Working in a real organization, and getting paid for it accurately, proves to be an entirely different challenge.
Related: AI is the buzz—but here are the real opportunities
This story was originally published by The Street on March 7, 2026, where it first appeared in the Technology section. Add TheStreet as a Favorite Source by clicking here.
