Sberbank Develops a Platform That Lets Anyone Build Their Own AI Agents
Russia’s largest bank, and the largest in Central and Eastern Europe, Sberbank has launched a new digital solution designed to help anyone learn to work with artificial intelligence and populate the digital world with their own AI agents. GigaPlatform AI Studio is a free service that can be used for personal projects as well as business tasks.
The platform’s main advantage is that users can try different GigaChat models, configure voice features, work with documents and build a functional AI agent prototype – all without any programming skills.
Everything You Need in One Place
GigaPlatform AI Studio offers a wide range of features. Users can transcribe lectures, convert text to speech, get advice or ask follow-up questions on virtually any topic. The platform provides access to all current versions of Sber’s AI models, including its flagship GigaChat 3.5 Reasoning.
In the future, developers are expected to be able to obtain their own API access keys for GigaChat text models directly through the platform and test them in their projects for free. The team behind the service has brought GigaChat models, tools for configuring them, voice and document features, and a visual builder that assembles agents from ready-made blocks into a single interface. With everything users need in one place, the time from idea to working prototype has dropped significantly – from weeks to hours. Testing an AI hypothesis typically requires programmers and considerable time: each technology needs custom code, tools must be tested in different environments, and previews often aren’t available. GigaPlatform AI Studio lets users assess an idea’s viability on their own. The service is free to use, with sign-in through Sber ID.
It Can Record a Podcast and Conduct Research
How does GigaPlatform AI Studio work? Users get simultaneous access to the entire GigaChat model family and the tools needed to configure them. Users who want to dig deeper can submit the same prompt to several models at once, compare the completeness and accuracy of their responses, adjust generation settings and test the results in a chat. Users can also connect their own documents or databases to a model, allowing the virtual assistant to base its answers on those materials. It can serve as an HR adviser, help a sales team or handle customer support.
For more complex research, Deep Research analyzes public sources, compares information and compiles a detailed report with source links. Users can specify which sources to include or exclude, then save the finished report as a PDF, HTML or Markdown file. They can also share an AI conversation with colleagues or use it with other AI services.
In Sber’s new service, models can recognize speech and turn text into spoken audio with pauses, stress and intonation. Special markup makes it possible to create conversations featuring different voices, so users can produce a full podcast. Building an AI agent in GigaPlatform AI Studio requires no code: users simply connect blocks in the visual builder, including a model, knowledge base, voice settings and dialogue scenarios, then test and launch the agent. These agents can take on routine tasks, from working with spreadsheets and generating images and speech to checking email and much more. The built-in AI assistant explains what is happening on screen and checks and analyzes uploaded files. To evaluate a document, users upload it and select the “Analyze” command. The assistant then produces a summary and answers questions about the file.
The platform also includes another useful tool: a free meeting assistant. It detects when a call begins, offers to record and transcribe it, and summarizes the discussion. Meeting information is not sent to third parties or uploaded to the cloud.
From Testing Ideas to Building Your First AI Agent
Sber’s platform is designed for people who want to learn how to work with AI models without extensive technical training and use them for routine tasks such as text-to-speech conversion, audio transcription and information searches. It also makes it easier to test an idea quickly, experiment with prompts and settings before deployment, and build a first agent from a ready-made template. Platforms such as GigaPlatform AI Studio lower the barrier to using generative AI for everyday users, developers and small businesses. Previously, these technologies required machine-learning experts and people who could write code. Now, users can get started without that expertise.
For individual users, the service can help create personal digital assistants, process documents and work with information. Businesses can test ideas more easily and launch their own digital services.
At the national level, the growth of platforms like this could boost adoption of Russian AI solutions and create an ecosystem of users around domestic generative AI models. The platform, developed by Russian IT specialists, reflects a global shift from experimenting with large language models to using AI agents at scale to handle routine work. In the near term, GigaPlatform AI Studio could be adopted in Russia’s government and social services, education and small-business sector. It could eventually become an export-ready IT product for countries across the Commonwealth of Independent States (CIS), Asia, the Middle East and Africa that are looking for alternatives to Western software.
