Itaú Unibanco cuts mainframe migration time 70% with AI agents

Latin America's largest bank is using AI agents to dismantle decades of mainframe code, signalling a new phase in banking's cloud transition.

By Disrupts Editorial Team Mon, 2 November 2026 AI level 2 of 5

AI level 2 of 5: drafted by our AI editorial assistant from the company's announcement; fact-checked, edited and signed off by Mark Walker, Editorial Director. What the levels mean

Two rows of dark grey server racks with glowing green and blue LED indicator lights line a brightly illuminated data center hallway, stretching towards a distant white wall beneath rectangular ceiling lights.

Itaú Unibanco, Latin America's largest bank by assets, has cut the time required to migrate its mainframe batch-processing systems to the cloud by up to 70% in an initial project phase, working alongside German digital-transformation firm GFT Technologies. Project delivery accelerated by up to 3.3 times over the same period. The results place the collaboration at the centre of one of the most consequential technology transitions under way in global financial services: the race to retire legacy infrastructure before it becomes an insurmountable barrier to AI adoption.

The bank's batch systems handle the daily mechanics of modern finance: overnight closings, settlements, and reconciliations. These are functions built over decades on mainframe architectures that predate cloud computing, and migrating them without disrupting live operations is widely regarded as the hardest category of modernisation work a bank can undertake. Itaú launched its broader cloud migration plan in 2018; the GFT engagement represents a significant acceleration of that programme.

AI agents as engineering force-multipliers

The technical approach centres on GFT's Batch2Cloud tool, integrated into its Wynxx generative AI platform, combined with Itaú's own Code Understanding solution developed by in-house technology unit Zup Innovation. AI agents handle the repetitive and analytically intensive tasks: scanning legacy codebases, identifying modernisation opportunities, estimating migration complexity, and generating and validating new code. Human engineers retain responsibility for architecture, governance, and all consequential technical decisions.

"Modernising critical systems is one of the biggest challenges organisations must address to turn AI ambition into real business value," said Marco Santos, Global CEO at GFT Technologies. "What we have achieved with Itaú shows how AI can become a native part of software engineering, combining automation with deep human expertise to accelerate complex transformations while preserving the governance, quality, and reliability that financial institutions require."

The governance emphasis is not incidental. Banking regulators in Brazil and across the G20 have become increasingly attentive to the risk profile of cloud migration programmes, particularly where operational continuity of systemically important payment and settlement infrastructure is concerned. Itaú's explicit framing of AI agents as subordinate to human architectural oversight is as much a regulatory communication as a technical one.

A sector-wide constraint becoming a strategic opportunity

The Itaú-GFT results land against a striking backdrop. In a recent GFT survey of 945 chief information and technology officers, 84% reported that legacy system limitations had forced them to cancel at least one AI project. That figure reframes what has often been treated as a modernisation cost centre, mainframe migration, into a direct constraint on AI return on investment across the entire financial sector.

For the broader technology industry, this is a significant signal. Cloud hyperscalers (AWS, Microsoft Azure, Google Cloud), which compete aggressively for large financial-services workloads, have a direct commercial interest in the pace at which mainframe-era systems are decommissioned. Accelerating that migration with AI agents also compresses the professional-services revenue cycle, creating pressure on the traditional systems-integration model that firms such as Accenture, IBM, and Capgemini have built around multi-year legacy transformation engagements.

The scalability claim GFT makes for Batch2Cloud is the most commercially consequential element of the announcement. If the 70% time-reduction figure holds across other institutions and code environments, a proposition that carries meaningful execution risk, it positions the tool as infrastructure-layer software for a migration wave that still has years to run. Santander, BBVA, HSBC, and dozens of regional banks across emerging markets face structurally identical legacy burdens. The capital implications extend further still: faster cloud migration compresses the window in which banks must carry parallel mainframe and cloud operating costs, directly affecting return on equity calculations that institutional investors in the sector watch closely.

Itaú's willingness to integrate AI agents into core settlement infrastructure will be studied carefully by peer institutions. The question of whether AI-assisted migration can be replicated at scale without introducing new classes of operational risk is one regulators and risk officers will be asking long after the project's efficiency metrics have been published.