Opinion
The AI tune is changing: From a wave of hype to business realism
The AI revolution is not an overnight upheaval, but a calculated evolution. In the coming years, companies will not win simply because they adopted AI; they will win because they knew how to manage it optimally.
About two years ago, when the GenAI revolution burst into our lives, the promise was grandiose: billions in savings, miraculous efficiency gains, and rapid workforce replacement. Organizations rushed to jump on the trend, launch pilots, and experiment with the new technology. Now, in 2026, we find ourselves in the midst of a wave of sobering reality. Initial enthusiasm is giving way to tough questions about ballooning cloud bills and return on investment (ROI).
Let us be clear from the start: the AI revolution is not a burst bubble. It is here to stay and is already delivering immense value to organizations that implement it correctly. What we are witnessing is not a technological failure, but a natural maturation process, just as happened with the internet, mobile, and cloud before it. This is a calculated evolution, not a chaotic revolution. For organizations to get the most out of it, they must understand that the rules have changed: we have moved from the phase of unchecked experimentation to smart management grounded in financial discipline.
The Expensive Lesson of Poor Planning
The primary challenge surprising organizations today is the inability to predict infrastructure costs. Unlike traditional fixed-price software licenses, the AI landscape relies on a consumption-based pricing model calculated by tokens and model execution. Without proper planning, projects burn through annual budgets in a matter of months. According to business reports, this even happened to a giant like Uber, which burned through its annual AI budget in just four months.
A few data points illustrate the scale of this phenomenon:
A recent report by Zylo shows that the average organizational spend on AI surged by 108% over the past year, reaching $1.2 million, with 78% of IT managers reporting surprise billing.
Only 11% of organizations can predict their AI costs with reasonable accuracy in advance.
This lack of financial control led research firm Gartner to predict that roughly 30% of GenAI projects will be abandoned after the pilot phase due to budget overruns and an inability to prove business value.
The primary source of waste lies in selecting unjustifiably expensive infrastructure, for example, sending simple queries to costly frontier models instead of Small Language Models (SLMs) that are hundreds of times cheaper.
CFOs Take the Reins
While tech and R&D leaders led the conversation early on with nearly open-ended budgets, CFOs now run the show. The era of the CTO driving AI out of technological enthusiasm is giving way to the era of the CFO managing it with financial discipline. This trend is particularly striking among enterprise organizations that already manage their cloud infrastructure professionally: according to a report by the FinOps Foundation, the percentage of organizations allocating a dedicated, separate AI budget leaped from just 31% in 2024 to over 90% today. Virtually everyone engaged in professional cloud cost management now incorporates AI into their workflows. This is not a step backward, but a sign of maturity: a technology transitioning from experimentation to critical infrastructure demands risk management and budgetary discipline, just like any other major business investment.
Organizations that started with the technology rather than the business use case learned that it is difficult to extract value from a product created simply because it was novel. Economic justification must be established upfront. Conversely, studies show that companies implementing AI correctly, with business planning and financial discipline, report an average revenue increase of approximately 16%.
The Myth of Losing Junior Talent
Another widely discussed topic is the future of employment and the fear that AI will replace workers, particularly junior staff. In reality, the pendulum swung too far and too fast. Companies that rushed to downsize staff are now discovering the true cost of losing organizational memory and institutional knowledge.
In this context, Gartner predicts that by 2027, roughly 50% of companies that laid off customer service representatives in favor of AI will rehire human workers. A notable example of this course correction is tech giant IBM, which announced it is tripling its entry-level hiring. As their Chief Human Resources Officer explained, without investing in entry-level employees, the talent pipeline and organizational well will simply dry up. AI is not designed to replace junior employees, but to make them significantly more productive. The role is not disappearing, it is evolving, shifting from writing basic code to managing, routing, and overseeing AI tools.
So What Should Executives Do Tomorrow Morning?
Start with the business: Define the business problem and expected ROI before kicking off a new project.
Optimize before expanding: Small Language Models (SLMs), intelligent query routing, and license pooling can save tens of percents in costs.
Implement AI FinOps: Continuous monitoring of cloud and token consumption will prevent "invoice shock" at the end of the month.
The AI revolution is not an overnight upheaval, but a calculated evolution. In the coming years, companies will not win simply because they adopted AI; they will win because they knew how to manage it optimally.
Ishai Ram is Senior Vice President at Sela.
