Sber taught AI to create business processes from a simple text query

The neural network independently assembles the automation scheme and writes the necessary software components, reducing weeks of work to a few hours

Sber has taught artificial intelligence to convert ordinary text queries into ready-made business process automation schemes. Now, an employee only needs to explain the task in their own words, and the GigaARPA platform will independently prepare the technical basis for its execution. According to developers, creating the first prototype can take just a few hours instead of several weeks.

Image source: ChatGPT

For example, you can ask AI to automate application verification, document processing, or data transfer between programs. The neural network will break down the task into individual actions, select the necessary tools, and write the missing code. The result will be a scheme combining software robots and AI agents based on GigaChat.

However, it will not be necessary to fully trust the neural network with the work. The employee will be able to check the proposed actions, make changes, and test the process before launching. The system also saves a history of changes and allows for the cancellation of erroneous decisions.

Sber expects that the technology will simplify the automation of daily tasks and reduce companies' dependence on developers. The platform can be connected to existing corporate systems, and it is not necessary to write program code independently to create the first prototypes.

Read more on the topic: