Microsoft supported Bradesco, one of the main banks in the country, in the creation of BIA (acronym for Bradesco Artificial Intelligence) Tech – an Artificial Intelligence (AI) co-pilot custom-made for developers’ daily needs. The financial institution has a robust innovation strategy and utilized all its expertise in artificial intelligence to create a development-focused solution with code sharing through GitHub Copilot and Azure OpenAI Service. The initiatives helped modernize the bank’s internal processes and led to efficiency gains of up to 40% in writing code snippets.
Advancing digital transformation in services used daily by millions of Brazilians is not an easy task. For this, Bradesco uses the learnings and innovative DNA that gave rise to BIA – an AI assistant launched in 2016 that revolutionized consumer service in Brazil – in its innovation strategy. Since the launch of its AI, which already has generative AI capabilities and enables rapid and accurate document generation, there have been nearly 2.5 billion interactions with customers and a wide range of added services, ranging from sending PIX and paying bills to personalized recommendations for Products, Services, and Investments, as well as alerts in case of suspected fraud in transactions. Currently, BIA with generative AI is available to 40,000 employees and over 500,000 customers.
“Bradesco has always invested in innovation and was one of the pioneers in using artificial intelligence in financial service assistance. We have accumulated great experiences with BIA serving our clients efficiently and later on with BIA Agencies assisting our frontline managers to deliver better services and consult internal norms. That is why we wanted to once again harness the full potential of Generative AI to help our developers be more efficient and have time to be more creative and consultative,” explains Nelson Borges, Bradesco’s IT Superintendent.
A BIA (Tech) for developers
With the successful implementation of BIA for both branch employees and external clients, Bradesco takes another step forward in AI adoption with the creation of BIA Tech, which leverages the full power of Microsoft Azure cloud and large language models from Azure OpenAI Service to address specific challenges of the IT team. The first of these was to support in writing instructions for development projects.
In order for a new feature to be introduced, the Bradesco development team needs to write ‘stories’ with instructions on what needs to be done by developers. With the great analytical power of Generative AI, BIA Tech has become an important tool to improve the quality of these instructions, breaking each story into smaller guidance parts and highlighting missing elements to make the information clearer. Additionally, the team created a plugin to allow developers to converse in natural language with BIA Tech integrated into existing tools and develop the story in an assisted manner.
“Everything being planned is based on feedback from Bradesco’s developer teams. Bia Tech already has over 1.5 thousand loyal users (heavy users), and the teams have been able to see the benefits of working with an AI co-pilot supporting developers’ day-to-day activities,” Nelson highlights.
According to Bradesco’s technical team, BIA Tech not only relieved the more bureaucratic work of a project but also increased the quality of stories produced by about 46 percentage points, with an average rating of 8.5 out of 10. This success led the bank’s IT team to expand the capabilities of the AI assistant, which can now help assess the risks of a code change, identify potential vulnerabilities, assist with metrics, among other tasks. The expectation is that in the future, BIA Tech will be expanded to cover all stages of the software development life cycle.
Still according to Nelson, “the use of BIA Tech to evaluate incident solutions brought significant gains such as a 23% reduction in the analysis time of an incident involving the root cause identification stage and the definition of corrective solutions, with a 90% accuracy rate in the suggested solutions”.
BIA Tech with Github
Bradesco is also investing in AI solutions combined with automation that can complement and further enhance the efficiency of internal processes, covering the entire development chain. In practice, this is reflected in the acceleration of “Time-To-Market”, the period a product takes from conception to being available for sale in the market, reducing the time required for provisioning and configuring a structure.
The combined solutions of BIA Tech and Github Actions and Github Copilot have promoted automation in code management processes, benefiting process efficiency, team collaboration, and improved management quality through telemetry and data analysis – contributing to the creation of bank products.
Currently, GitHub Copilot suggests over 10,000 lines of code per day, with great acceptance from developers. These numbers reflect a significant improvement in software development efficiency, with time savings ranging from 30 to 40% in a code snippet.
The goal is to empower employees and provide tools for them to replace repetitive tasks with activities that explore their creative and cognitive potential, ensuring the bank remains at the forefront of technological innovation. The implementation of GitHub, GitHub Copilot, and Bia Tech at Bradesco paves the way for exploring new technologies and innovations. The bank is also experimenting with other AI and automation solutions that can complement and further enhance its internal processes.
“What we have noticed is that the use of AI within the IT solutions cycle can bring significant gains in productivity to engineers, either in solution creation or operation. Our focus for 2025 will be to expand and broaden use cases through open, collaborative, and modern AI architectures focusing on multi-agents.”, concludes Nelson.
“Generative AI presents unprecedented potential for financial institutions. It supports teams in their daily tasks, streamlining repetitive processes, assisting technical teams, and facilitating access to internal regulations. This technology is also a great ally in the creation of assertive solutions, fueled by data, which will meet the specific needs of the bank’s diverse client profiles,” comments Joana Moura, Vice President for Corporate Clients at Microsoft Brazil.
