The main gap between companies today comes down to their ability to rebuild workflows. According to BCG research, only 5% of businesses have systematically integrated artificial intelligence (AI) across all functions, while another 35% are scaling AI but admit they could move faster. The remaining 60% of companies see almost no impact from this technology on costs or revenue.
Serhii Tokarev, an investor and founder of the Tokarev Foundation, believes that the problem is not the technology itself. Some companies are genuinely rebuilding their work around AI, while others are trying to buy “ready-made autonomy” without gradually changing their processes.
The most common mistake is evaluating artificial intelligence by the results of individual tasks. For example, if an employee uses AI to prepare documents ten times faster, but a lawyer spends more time reviewing them, overall productivity barely changes. A business gains real value only when the entire workflow improves, not just one part of it.
AI Should Analyse but Not Always Act Independently
When analysing a business, AI can become the first analyst. It helps review financing history, reconstruct the competitive landscape, find important data, and flag information that requires closer examination. This technology saves time on collecting basic facts, while a person can focus on what matters most: understanding whether the company has a chance to create long-term value.
“I never ask a model whether I should invest in a startup. This technology analyses data faster than an entire team, but the final decision and responsibility remain with a human,” says Serhii Tokarev.
In his view, AI can move from analysis to action only when clear boundaries are defined for it. Each new level of autonomy must be supported by reliability, checks, and measurable quality.
Where AI Delivers the Best ROI
Companies should not expect artificial intelligence to immediately automate entire areas of work, such as finance, customer support, or legal operations. Projects where companies simply try to “replace analysts” usually fail because reviewing the results becomes almost as complex as doing the work manually.
“AI works well where large volumes of information need to be analysed, contracts need to be reviewed, or invoices need to be processed. In such cases, the model takes on complex preparatory work, while a person evaluates a short, structured result,” says Serhii Tokarev.
When calculating ROI, companies cannot simply compare the cost of AI tokens with an employee’s salary. What matters is the cost of an accepted result that meets the company’s requirements for quality, speed, and risk. This includes spending on the model, repeated requests, integrations, monitoring, human review, and rework in case of errors.
That is why some businesses do not see productivity growth, even though they already spend more on AI than they previously spent on people’s salaries. Such companies optimise the cost of tokens, not the cost of results.
“In our portfolio companies, AI does not replace operations or investment teams. It complements and expands them. The value is not in a single model, but in how all these systems work together,” adds founder of Tokarev Foundation.
