Everyone is talking about AI adoption. The numbers look convincing: 88–89% of companies use it in at least one function. But behind that number might be a girl making a greeting card in a neural network: formally, she’s engaged, and in the survey, she «uses AI.» The question isn’t whether they are using it, but how deeply. Next is an attempt to follow the data and understand where the hype ends and the real work begins.
Expectations vs. Reality

AI in Marketing: 90% Use It, But Implementation Depth Remains Low. Data from the Presentation

AI in Customer Service: The Bot Responds, But Doesn’t Solve the Problem. Data from the Presentation, 50% — Scripted Dialogue.
Implementation
88–89% of companies use AI, but most utilize it in only one or three functions. Sources: McKinsey, Stanford HAI, Deloitte; U.S. Census Bureau
AI Adoption: Global — 88–89%, Russia — 65%. Sources: McKinsey, Stanford HAI, Deloitte (2024–2025); «AI Implementation in Company Business Processes, ” 2025
If everything is so superficial in practice, where does all the money come from? In the first quarter of 2026, the global venture market set a record—$297–300 billion—and 80–81% went into AI. But big investments don’t always equal big benefits; sometimes it’s consumer remorse, just on an industry-wide scale.
Market
80–81% of global venture capital is flowing into AI—a record $297–$300 billion per quarter. Q1 2026
61% of new unicorns in 2025 will be AI startups. Data from the presentation: 2015–2025
Let’s step away from money and corporate reports. Let’s see what people are actually doing with AI every day—at work and in their personal lives.
Staff
Shadow AI is overtaking corporate AI: 34–48% versus 25–36%. Sources: TrustedTech and Censuswide (March 2026), Thomson Reuters Institute, IBM, Deloitte
AI is generating text at work, but it is increasingly creating multimedia: +8.5 percentage points. Comparison of request categories: June 2026 to July 2024
Outside of work, AI became a consultant: practical advice rose by 35%. Comparison of query categories: June 2026 to July 2024
Inquiries
37 million queries in Copilot: technology, work, and health are the top three topics. Source: Microsoft
ChatGPT: Tutoring and «how-to» instructions—nearly 19% of messages. Source: OpenAI, approximately 1.1 million messages, May 2024–June 2025
Yandex PromptHub: IT and Neural Networks — 16,4% of 3,483 prompts. Source: Yandex PromptHub
Consumers
74% trust AI to choose products—but 71% worry about privacy. Sources: Deloitte, Microsoft, Ascend2
And here is the main temptation—to chase the trendiest model. But the agent data suggests otherwise: you can converse with AI using plain language, and a complex solution is not always better than a simple one.
Maturity
How Companies Are Adopting AI: Three Levels of Maturity. Three Perspectives: R. Hoffmann, I. Mollick, K. Kozyrkova
Agents
The main AI revolution is talking to it in simple language. Examples: GLM 5.2 Agent, Claude Code
AI Agents: The simple task is turned into a sequence of actions. The 2026 Trend
OAuth: the agent gains access to services via authorization. Available actions depend on the API and permissions.
Economics




