Bank of Georgia’s AI Employee is a group of AI agents that works as a member of the bank’s software development teams. An assigned task runs end-to-end, where the system gathers the context it needs, plans the work, writes and tests the code, and submits it for review. A human developer approves everything before it reaches production.
What separates it from a coding assistant is where the human sits. An assistant waits to be prompted at each step, while AI Employee is given a piece of work and comes back with a result. Several AI models are orchestrated according to what each does best, and the system draws on a technical knowledge base generated automatically from the bank’s own source code. When it finds it does not have enough information, it asks the question on the task itself and waits for the answer before continuing. Every stage is visible on a dashboard, so the work can be followed and audited as it happens.
The system was built in-house and runs in production, taking on a substantial share of the development work behind the bank’s digital products. It was designed as a modular platform, so different development teams can take it on for their own work and new capabilities can be added without rebuilding it. The aim is to increase what the bank’s engineering teams deliver without adding people in proportion, leaving developers to spend their time on architecture, complex logic and the judgement calls that need a person.
