Artem Lyashanov: Why Most AI Investments Fail to Pay Off and What to Do About It in 2026

Corporate artificial intelligence (AI) has moved past the euphoria stage. Now comes the reckoning, when boards of directors and investors demand not slide decks full of percentages but real impact on profit. According to the IBM Institute for Business Value and Oxford Economics, only about 25% of AI initiatives today meet executives' profitability expectations. Expert Artem Lyashanov explains why this is not a crisis of the technology but a predictable result of the wrong approach to implementing it, and what sets apart the companies that are already in the black.

Key takeaways

• The return on pilot AI projects is almost always higher than the return after scaling across the entire company;
• Only 25% of AI initiatives deliver the expected profit, but this is not a reason to stop investing — it is a signal to rethink how those investments are structured;
• 64% of AI budgets have already shifted toward core business functions;
• Chaotic, unsystematic AI adoption is declining;
• The next stage is agentic AI, and companies without a systematic foundation risk failing to keep up with this transition.

Argument 1: the gap between pilot and scale is a pattern

Many executives still judge the effectiveness of AI by the results of pilot projects. This is a methodological error. A pilot operates under controlled conditions, on a limited amount of data, with a dedicated team and minimal integration costs. Scaling to the entire enterprise removes all of these greenhouse conditions at once.

That is exactly why, according to Artem Lyashanov, impressive pilot-stage returns are almost never sustained in the move to full-scale production, and a drop below the company's cost of capital (the Capex hurdle rate) should be seen not as a failure of the technology but as a signal that the project assessment was incomplete from the start.

Conclusion: if a company decides to expand its AI budget based on pilot data alone, it systematically inflates expectations and sooner or later will face disappointment from the board.

Argument 2: 25% successful initiatives is not a market failure

The 25% figure often sounds like a verdict on the technology. In reality, it says something else: the market is gradually weeding out projects built on hype rather than business logic. Artem Lyashanov stresses that the growing pressure on technology leaders — to deliver not innovation for innovation's sake but a measurable financial result — is a healthy process, not a sign of an industry crisis.

Conclusion: companies should not abandon AI investment because of the low average ROI across the market; instead, they should honestly assess whether their own projects are truly tied to operating profit or exist as a stand-alone cost item.

Argument 3: Budgets shifting to the core

64% of AI budgets today are directed at core business functions, and this, in Artem Lyashanov's view, explains both the slowdown in ROI growth and its future potential. Working with core business processes requires deep coordination between departments and high-quality foundational data — precisely what is hardest to build quickly.

The core of the business delivers a slow start, but many times greater scaling potential with proper implementation discipline.

Conclusion: if a company's ROI is falling right now while budgets are growing, this is not necessarily a warning sign. The company may simply have moved from easy wins to harder and more valuable work.

Argument 4: A declining share of chaotic adoption

The share of companies adopting AI without a system has fallen noticeably over the past year. This is perhaps the most underrated signal in all of the statistics. It reveals not so much current profitability as the future trajectory of the market.

Companies that are building a coherent, company-wide AI strategy based on verified data today are laying the foundation for ROI that will show up not this quarter, but over the next two to three years.

Bottom line

Artem Lyashanov believes the next stage in the technology's evolution will be the shift to agentic AI — models in which algorithms stop being tools and become full participants in business processes. Companies that are investing in systematic approaches today, rather than in the number of pilots, will be ready for this transition much faster than their competitors.

The main argument: a low average ROI across the market is a reason to rethink how exactly these investments are structured. The difference between a profitable technology stack and a wasted budget today is determined not by the size of the investment but by the discipline with which it is applied.

Source of statistical data: joint study by the IBM Institute for Business Value and Oxford Economics, 2026. This material is for informational purposes only.