Dmitry Chudinov, Russian Post: 70% of AI Success Is Process Restructuring, Not Racing for the Model

Dmitry Chudinov, Deputy General Director of Russian Post for Digital Transformation, spoke on June 30, 2026 at the CNews FORUM "Cases: Experience of IT Leaders" conference with a thesis that diverges from the familiar logic of the AI market: 70% of the success in implementing neural networks is not the choice of the right model, but the restructuring of business processes to fit it.

Why are working AI projects so rare?

A year ago, about 50% of Russian companies declared the implementation of artificial intelligence. Today, it's almost 100% — AI has firmly entered corporate strategies, budget cycles, and reports to boards of directors. The technology is discussed at every industry conference, budgets are allocated for it, and teams are formed.

But the growth in declarations has not led to a growth in actually working projects. According to Chudinov's observation, functioning AI implementations remain extremely rare. The market is filled with pilots, demo stands, and "AI on slides" — but not with systems that work steadily in production and deliver measurable operational results.

The key reason is a misplaced priority. Companies start with choosing the "right" model: a domestic LLM, a foreign neural network, their own development, or a fine-tuned solution on open weights. Technical choice concentrates resources, vendor negotiations, and management attention. The question of how real work processes will actually change — regulations, distribution of responsibility, decision-making points — takes a back seat or is not raised at all. Companies imitate transformation instead of implementing it: they buy technology, report on "AI implementation" — and return to the familiar operational scheme.

What does "70% — these are processes" mean?

Russian Post formulates a different approach: operational transformation first, then technical choice. This means that before choosing a vendor or model, a company must answer basic questions: what exactly changes, how is the effect measured, and who is responsible for the result?

According to Chudinov's assessment, 70% of an AI project's success is determined by organizational work: a reassessment of who and how makes decisions, which steps are automated completely, and where AI is integrated into the work cycle as a full participant. It is at this stage that it becomes clear what data is needed, who is responsible for it, and how the human's role in the process loop will change.

The remaining 30% — technological choice: model, infrastructure, integrations. This is an important stage, but it should follow operational design, not precede it.

For Russian Post with its federal scale — hundreds of thousands of employees, a logistics network across the country — this is not an abstract principle, but a practical requirement. Each AI implementation should deliver a measurable result: speed, accuracy, reduction in operating costs. Without prior process restructuring, the model is embedded in the old scheme and reproduces its inefficiency.

What this means

The gap between "almost 100% of companies talk about AI" and "only a few get real results" — this is not a failure of technology. This is a failure of management approach: organizations buy a tool without changing the environment in which it should work.

Chudinov's thesis is simple and inconvenient at the same time: for AI to work, a company must change itself — before the neural network is launched into production. Technology amplifies what exists: if the process is built — AI accelerates it; if not — it exposes chaos. This is precisely why AI transformation is a management project, not an IT procurement.