92% of US developers already use AI tools to write code every day. 41% of all global code is generated by artificial intelligence. These numbers are not a forecast for the future, but the current reality of 2026. Startups are launched in weeks instead of months, product teams halve the development cycle, and engineers spend more time on architecture than writing lines of code. In Russia, the situation is different: large and medium-sized businesses are still writing code the old fashioned way, resisting new tools and losing time. The gap between those who implement AI and those who wait is growing every month. The question is no longer, whether you need to switch to AI development, but how much time is left for those who do not.
Vibe coding in the product team: from routine to design
Andrei Karpaty, a machine learning researcher, coined the term “vibecoding” to describe an approach in which a developer formulates a problem in natural language and AI generates code. In theory, it sounds like an experiment. In practice it is already a workflow in product teams around the world.
In companies that have switched to vibe coding, the share of hand-written code is 5-10%. The rest is generated by Claude Opus, Gemini and other AI tools. The team spends most of its time describing scenarios, passing context, setting up automations, and developing skills for AI agents. The result is a 50-70% acceleration of development.
The routine part of the developer’s work has finally disappeared. Instead of writing typical functions and template code, engineers focused on designing the architecture, thinking over integrations and developing interaction scenarios. This is a fundamental shift: the developer ceases to be a performer and becomes an architect.
At the same time, acceleration should not reduce the quality of the product. Mature teams build a system of three levels: alpha releases, beta releases and full releases. Rollout rules do not change with the advent of AI – the distribution of time changes. The faster the code is created, the more resources are spent on testing. QA engineers and analysts are more loaded than before.
DORA State of DevOps Report 2024 data confirms that the quality problem exists in the industry: with the growth of AI adaptation, the stability of releases decreases by 7.2%. Teams that do not rebuild test processes to the new speed get more bugs and rollbacks. Vibe coding speeds up code creation, but without strengthening quality control, this acceleration works against the product.
Russia vs West: Why vibe coding is stalling
For startups, vibe coding takes root quickly. Small teams launch MVP in a matter of days, the number of new projects is growing. In medium and large businesses, the picture is the opposite: developers resist, managers do not believe in technology, companies lose their competitive advantage.
In the West, the approach is different. In Europe, requirements for developers are changing: companies increasingly expect engineers to work with AI tools. This affects not so much salaries as demand: specialists who know how to work with agent development gain an advantage in the labor market.
Technology has ceased to be an option – it has become a requirement. Those who do not master new tools get less or lose their jobs.
There is no such pressure in Russia yet. This creates the illusion that the transition can be postponed. In fact, the gap is accumulating: Western teams release the product in two weeks, Russian teams – in two months. At the same level of functionality, the one who enters the market slower loses.
Among the CEOs and technical directors of Russian product companies, skepticism about AI agents and vibe coding is still widespread. Meanwhile, the speed of companies entering the market can grow 3-10 times thanks to AI tools. For Russian product teams, this means that competition will become even more intense – not only with domestic players, but also with global products that are developed faster and cheaper.
AI in team communications: assistant, not participant
Many product teams already use AI to generate meeting summaries and highlight tasks. The next step seems obvious – an AI agent that independently allocates tasks, monitors progress and initiates refinements. However, development will go differently.
AI will not become a full member of the team. Its role is a powerful assistant that works in real time. Product teams are already connecting agents to Telegram, Slack and internal portals. The scheme looks like this: before the meeting, the participant asks the agent for context – he collects information from correspondence, trackers and documentation, generates a summary of open questions and decisions made. During the meeting, the agent logs the conversation, and at the same time retrieves data from corporate systems on request: an estimate of labor costs, task status, and discussion history. The output is a ready-made protocol with reference to tasks, and not just a transcript of the conversation.
According to Adobe AI and Digital Trends 2026, 63% of companies expect agent AI to free up employees for strategic and creative work. In practice, this will not happen. AI really removes the routine – extracting information, formatting reports, preparing documents. However, the time freed up does not become free. The speed of companies entering the market is growing 3-10 times, which means that strategic decisions need to be made more and faster. AI takes away the routine and immediately gives this time back in the form of an increased burden on thinking and decision-making.
Technological sovereignty and the future of SaaS in Russia
From January 1, 2026, foreign software is completely prohibited in Russia at critical information infrastructure facilities. The regulatory shift is massive, but its implications are mixed.
Generative AI is developing more slowly in Russia than in the West. The ban on the use of Western tools will increase the backlog. At the same time, the business will continue to find ways to circumvent restrictions – to work without Claude, Gemini or GPT in 2026 means losing speed and quality. Government structures will be forced to work with available tools, and the gap between the public and private sectors will grow.
The volume of the Russian IT market in 2026 is approaching 4.5 trillion rubles. At the same time, organizations doubt the stability of domestic solutions. The reason is the product creation model. The state pours subsidies into IT development, but products are created without direct interaction with business. The launch cycle takes at least five years, and without feedback from real users, the product is divorced from the needs of the market.
The result is predictable: government-funded products come out with low quality, without market adaptation and without a sustainable business model. The product owner develops functionality based on his own ideas, and not from the tasks of the end user. At the output there are solutions that do not withstand competition.
What a leader should do: AI transformation starts with yourself
Harvard Business School talks about the need for Change Fitness – the ability of an organization to adapt to continuous technological change. In the Russian context, this term takes on a practical meaning: teams that do not master AI at an above average level will fly out of the market.
Employees who sabotage the use of AI tools, agent development and prompting become ballast for the team. Reaching people to the right level is expensive and long – it can take years. At the current pace of competition, years are an unaffordable luxury. The entire market has turned red, and teams with minimal AI literacy cannot resist.
At the same time, the forecast for 2-3 years is unexpected: the staff of IT specialists will grow, not shrink. Now it seems that AI is replacing developers, but in reality, the more AI tools a company uses, the more engineers are needed to support infrastructure, integrations and quality. The following specialists will become more in demand:
- analysts and security professionals;
- QA engineers and testers;
- architecture and integration designers;
- developers capable of managing AI agents and controlling generation quality.
Main tip to the leader: transformation starts with a personal example. If the leader does not use AI tools every day, does not show the team how they work, and does not invest in training, no transformation will occur. An order from above without a demonstration doesn’t work. Only the habit that the leader forms in the team through his own practice works.
Three things a manager should do now:
- Start using AI tools in person every day and show results to the team.
- Rebuild testing processes – if development accelerates, QA should accelerate proportionally.
- Stop waiting for technology to “ripen.” They are already mature. The gap between those who implement and those who wait is growing every month – it will be more and more difficult to close it.
By Ruslan Gatiyatov, founder of the Glabix platform, an expert in the field of IT team management