Does AI Turn Developers into Managers? — Andrey Dmitriev

Andrey Dmitriev, co-founder of JUG Ru Group, begins with specialization: engineers once spoke to customers, developed software, tested it, and supported the result. Growing companies distributed that work across professions. Alexander Polomodov describes an experimentation system where statistics and domain knowledge mattered more than attachment to a single language. Agents now make it possible to reconnect stages of development, while a person remains accountable for the outcome.

Product engineers need to understand users’ needs and organize execution. Andrey compares managing agents with an employee lifecycle: bring them in, develop them, supervise them, and retire them. Alexander proposes checking whether the work should be automated at all. An agent can keep producing reports that nobody reads. Buying a subscription and demanding higher productivity does not replace a shared process, training, and clear checks.

Team knowledge becomes a foundation for delegation. Premature layoffs are dangerous when essential information exists only in employees’ heads. Alexander describes development where code can be regenerated from intent, architectural constraints, observable behavior, and reasons for decisions. Andrey highlights contradictory documentation and lost continuity during generation. Verification must cover security, architecture, and the continued relevance of the task as well as functionality.

The discussion expands to products and marketing for agents, personal interfaces, and software for one-off needs. Alexander describes a website for choosing a London neighborhood; Andrey describes a bot that gathers requirements and builds a website. The participants distinguish practical examples from uncertain predictions about future careers. Their closing priorities are critical thinking, domain knowledge, experience, and a question for managers themselves: what do they contribute to a changing process?