Serhiі Tokarev: How to Transform Business Errors into Lessons with AI

The rise of artificial intelligence in the global market has brought about a profound transformation in how businesses operate, compete, and grow. AI is now capable of handling not only routine tasks but also genuinely complex processes that once required significant human effort and expertise. Many industry leaders have explored AI implementation in depth, discussing both its potential pitfalls and its remarkable opportunities to drive business growth. Serhiі Tokarev, an IT entrepreneur, co-founder and partner at Roosh Ventures, has shared three practical strategies for using AI to enhance businesses — and more specifically, to turn costly mistakes into meaningful lessons.

Failure Is Inevitable — Repetition Is the Real Problem

Serhiі Tokarev likens the adoption of AI to building a website. Just as the first version of a site is rarely perfect and requires continuous iteration, businesses should expect a learning curve when integrating new technologies. He emphasizes that failure and setbacks are an inherent part of both business and life. The true problem, he argues, is not that mistakes happen — it is when those same mistakes repeat themselves, leading to recurring failures that erode profits, morale, and momentum.

Tokarev believes that errors can be transformed into valuable learning experiences, and that AI provides some of the most powerful tools available today to facilitate this transformation. Rather than viewing artificial intelligence as a silver bullet, he frames it as a sophisticated system for pattern recognition, memory, and insight — one that helps organizations understand why things went wrong and how to course-correct.

He outlines three key approaches to effectively leverage AI for business success.

Identifying Cause-and-Effect Relationships

The first strategy focuses on one of AI’s most powerful strengths: its ability to identify complex, non-linear relationships that humans are likely to miss. Traditional business analysis tends to be linear. A decline in sales, for example, might typically be attributed to obvious factors like reduced consumer demand or pricing issues. A human analyst working with standard reporting tools may stop there.

AI, however, can go much further. Tokarev highlights that machine learning models can uncover surprising correlations — such as the influence of seasonal weather patterns, shifting cultural trends, hyper-local geographic preferences, or even the timing of competitor promotions. These are the kinds of hidden variables that rarely appear in a standard business review but can have an outsized impact on outcomes.

For business leaders, the practical implication is significant. By deploying AI to map cause-and-effect relationships across large datasets, companies gain a much clearer picture of what actually drives their results — not just what appears to drive them. This allows for more targeted interventions, smarter resource allocation, and fewer repeated mistakes rooted in misdiagnosed problems.

Building an “Institutional Memory”

While the first approach addresses immediate analytical challenges, Tokarev’s second strategy takes a longer view. It is about building what he calls an institutional memory — a structured, AI-powered record of a company’s historical successes and failures that can actively inform future decision-making.

Most organizations accumulate enormous amounts of data over time, but relatively few make systematic use of it. Reports get filed, post-mortems get written, and lessons learned are recorded — only to be forgotten when team members leave or priorities shift. AI changes this dynamic. By continuously analyzing past performance against specific parameters, AI systems can identify meaningful patterns across years of operational history and translate those patterns into tailored strategic recommendations.

Tokarev points to UPS’s ORION system as a compelling real-world example. ORION integrates AI directly into logistics operations, analyzing customer data, route maps, pickup schedules, and historical performance metrics to optimize delivery routes in real time. The results have been substantial: the system has allowed UPS to reduce travel distances by millions of kilometers each year, significantly improving both operational efficiency and environmental impact.

The lesson for other businesses is clear. When AI is given access to historical data and trained to recognize what worked and what did not, it becomes an organizational asset that compounds in value over time — functioning almost like an experienced senior advisor who never forgets a past project and never leaves the company.

Tuning Into Customers

Tokarev’s third strategy is particularly relevant for B2C businesses, though its principles apply broadly. Customer interactions — whether conducted through sales teams, automated chatbots, email threads, or online reviews — are rich with unstructured information that most companies barely scratch the surface of.

Every complaint, every five-star review, every abandoned shopping cart, and every support ticket contains a signal about what is working and what is not. The challenge has always been processing this data at scale. AI makes that possible. By analyzing inputs from all communication channels simultaneously, AI systems can surface actionable insights that help businesses refine their strategies, address friction points, and better meet customer expectations.

For companies that have experienced customer-facing failures — a product launch that underperformed, a service gap that triggered negative reviews, a communication breakdown that led to churn — AI-powered customer analysis provides a structured way to understand the root causes and avoid repeating the same missteps.

The Importance of a Balanced Approach

Despite his enthusiasm for AI’s potential, Tokarev is careful to emphasize that artificial intelligence is not a replacement for human oversight — and that blind reliance on algorithms can be just as dangerous as ignoring data entirely.

He references Zillow’s real estate division, Zillow Offers, as a cautionary example. The division relied entirely on AI-driven algorithms to predict housing prices and guide purchasing decisions. Without sufficient human intervention, the system failed to adapt to unexpected real-world disruptions — including the economic turbulence caused by COVID-19 and widespread labor shortages. The result was significant financial losses, mass layoffs, and the eventual closure of the division.

This example underscores a critical point: AI models are trained on historical data and can struggle to account for unprecedented conditions or black swan events. Human judgment, contextual awareness, and ethical reasoning remain essential complements to any AI-driven system.

• Use AI to surface patterns and insights, but apply human expertise to interpret and act on them.
• Build feedback loops that allow human teams to flag when AI recommendations seem misaligned with ground-level realities.
• Treat AI as a powerful tool, not an autonomous decision-maker, especially in high-stakes or rapidly changing environments.

Achieving the best outcomes, Tokarev argues, requires striking a deliberate balance between AI-driven insights and human judgment. While AI provides remarkable opportunities to learn from a company’s own mistakes, it is equally important to learn from the missteps of others — including well-resourced organizations like Zillow that moved too fast and trusted their models too completely.

The broader message from Tokarev is both practical and optimistic. Errors are unavoidable in business. What separates resilient, growing companies from those that stagnate is their ability to extract genuine learning from every failure — and AI, used wisely and in partnership with skilled human teams, is one of the most effective instruments available for doing exactly that.