From 2 days to 45 minutes: Bradesco’s agentic AI transformation | BRK272
Felipe Fernandes dos Reis and Bruno Borges present Bradesco’s approach to agentic AI across the engineering lifecycle, from assessing legacy services to planning and carrying out modernization. The session connects the bank’s operating model and cloud strategy with its use of AI-assisted developer workflows.
Modernization at Bradesco’s scale
Bradesco is putting agentic AI to work across 5,126 services, building on an AI foundation behind more than 600 initiatives serving 74 million customers. The session highlights a modernization task reduced from two days to 45 minutes, while retaining architecture, security, and quality gates.
Azure AI Foundry and GitHub Copilot support the assessment, planning, and modernization work. The discussion also addresses legacy migration challenges and unsuccessful early cloud migration attempts, providing context for the bank’s current approach.
Topics and chapter guide
- 00:00 — Introduction: AI modernization and GitHub Copilot initiatives.
- 00:48 — AI transformation and operating models: Bradesco’s AI transformation overview and changes to its IT operating model.
- 03:04 — Technology landscape: The bank’s technology environment and Kubernetes modernization strategy.
- 05:38 — Business and developer AI platforms: BIA, Bridge, and BIA Tech in business and developer workflows.
- 08:55 — Cloud architecture and infrastructure: Resilience patterns and AI-driven infrastructure management.
- 10:09 — Legacy migration challenges: Lessons from failed early cloud migration attempts.
- 13:09 — Agent-assisted discovery and planning: Using AI agents for discovery, dependency mapping, and migration wave planning.
- 18:31 — Java application modernization: Modernizing Java applications with GitHub Copilot and AI agents.
- 27:36 — Target architecture and migration strategy: Target cloud architecture, migration strategy, and modernization lessons learned.
- 37:07 — Agent tooling and Q&A: Azure Catalyst, MCP servers, hosted agents, Copilot Fusion, and audience questions.
Session resource
This Microsoft Ignite 2026 session is BRK272, an advanced (300-level) breakout in the Agent Platform track, presented in English (US).