AI Products: First Figure Out Whether You Need AI

Daniel Levinishnikov begins with T-Bank's AI organization and products replacing help from a branch or call center. His career strategy was to build management skills in consulting, learn product iteration in a startup, understand scale at Alibaba, and combine those modes in a large company.

An AI product is defined by a useful customer job, not by a model. The team asks whether AI is needed instead of a heuristic, then tests data, quality, scalability, and economics. Internal products are judged by cost and support workload; external ones by acquisition, retention, lifetime value, and transactions.

Frequent jobs will remain in interfaces, while the long tail moves to agents, so products must serve machine customers too. In a large company, prototypes face integration, security, and error-cost constraints. The portfolio runs like venture capital: fund a round, inspect traction at explicit gates, and close weak bets.

The leader need not be the strongest narrow specialist. The job is to gather experts, maintain strategic range, and create conditions for good work. A predictable experiment process reduces avoidable uncertainty even when the domain remains unclear. The practical advice is to learn adjacent professions and combine product, technical, and management skills.