AI Agents Uncover Hidden Business Process Logic

Artificial intelligence identifies hidden business processes that are informally supported by employees. AI agents effectively handle tasks that are not covered in official regulations.

Hidden Process Logic

In poorly documented workflows, there is often an employee who prevents failures. They know who to call and where to find missing files. Such work is usually absent from the regulations.

Developing AI agents quickly brings this hidden logic to light. Anton Rubenchik, CIO of Sibur, noted this at TNF. The company is experimenting with agent autonomy, focusing on process improvement rather than the number of implementations.

Kolya Agent Example

At Sibur, a procurement agent named Kolya checked supplier documents. In its first week, it failed on archives up to 380 MB. Human buyers had spent years manually sorting through heavy PDFs and duplicates.

After launching the agent, rules had to be rewritten: a single file format, one iteration for additional requests, and a 24-hour response time. Work quality increased from 41% to 65%. The model was not the main issue; improvements were made to the process itself.