Leading a Team with AI — Dmitry Tverdokhlebov

Dmitry Tverdokhlebov, founder of Omnius.team, explains how a small team brings generative AI into large organizations. In his experience, attempts to reproduce cloud model capabilities entirely in-house often delay projects. The agency starts with external information, brings a useful result into the company, shows it to decision makers, and earns support for the next stage. Integration still requires documentation, interfaces, and implementation support. This approach builds trust; it does not establish that a standalone pilot has already transformed the core business.

Examples include tax legislation monitoring, help with Telegram advertising moderation, and finding substitutions between construction materials. In the last case, sixteen agents examine different constraints while additional critics challenge their conclusions. The comparison between a month of work by two or three analysts and 30–40 minutes of system work is the guest’s estimate; he does not yet quantify revenue from new markets. Leaving corporate employment broadened his access to tools and experience across clients. A separate agricultural research project illustrates his interest in problems beyond conventional corporate assistants.

Inside Omnius.team, nine independent specialists are connected by an assistant that retains context from meetings, email, calendars, and agreements. Tasks reflect discussions. Rules evolve through mistakes: private salary conversations must stay out of public channels, and the bot should not interrupt without reason. Dmitry considers the system far from his intended level of maturity. The speakers draw a parallel with software engineering: correct both a code defect and the conditions that produced it, while preserving requirements and decision rationale outside the regenerable implementation. Engineering understanding remains necessary to verify the result.

When hiring, Dmitry looks for personal experience experimenting with AI, clear communication, critical thinking, and motivation to build useful things. Talant.club is developing a model for smaller orders: the platform creates a specification, vetted teams propose solutions, AI checks conformity, and the client selects the preferred result. The winner receives half the budget; the remainder is distributed after a platform operating fee. The community has five people at the time of the conversation. The final recommendation to managers is to solve a small problem with AI personally, giving subsequent management decisions a foundation in practical experience.