What tech stack does Bank of America use?

At the core of Bank of America's transaction processing sits one of the world's largest IBM mainframe environments, running COBOL and REXX for overnight batch processing, account ledgering, wire transfers, and payment settlement. These systems handle transaction volumes that would be technically and economically prohibitive to migrate wholesale to any public cloud — a reality BofA has embraced through deliberate mainframe modernization (adding z/OS Connect API layers over COBOL core logic) rather than replacement. IBM Developer for z/OS (IDz) and Eclipse are the documented tooling for mainframe developers; Maven, Gradle, and Jenkins appear in adjacent application development job postings.

The most strategically distinctive infrastructure decision BofA has made in the past decade is its choice to build a proprietary private cloud rather than standardizing on AWS or Azure. This initiative — detailed publicly by BofA and covered by CIO Dive — reduced the bank's server count from 200,000 to 70,000 and its data centers from 60 to 23, generating $2 billion in annual infrastructure savings versus the public cloud alternative. The bank built a software-defined network and containerized application runtime internally, giving it control over data sovereignty, latency, and regulatory compliance that public cloud contracts could not easily guarantee for a G-SIB operating under OCC, Fed, and FDIC oversight. As of 2025, BofA is selectively evaluating specific third-party public cloud workloads for non-critical or edge functions, including an 18-month IBM hybrid cloud collaboration — but the core banking infrastructure remains proprietary.

At the application layer, Java is the dominant language for modern banking services, APIs, and microservices. Python handles data science, risk modeling, regulatory capital calculations, and AI/ML workflows. DevOps tooling includes Jenkins for CI/CD, Ansible for configuration management, Bitbucket/Git for source control, and Maven/Gradle for build automation — all documented in publicly available BofA engineering job postings. z/OS Connect bridges the mainframe COBOL core to RESTful APIs, enabling modern applications to consume mainframe-resident account and transaction data without full replatforming.

Bank of America's mobile applications — used by more than 58 million active digital users — are built natively in Swift for iOS and Kotlin for Android, following industry-standard patterns for large-scale consumer financial apps. The web experience uses a JavaScript-based frontend framework (React-family signals appear consistently in BofA engineering and UX job postings). In 2025, BofA clients connected 30 billion times through digital channels — 16.6 billion logins and 13.3 billion proactive alerts — making this one of the highest-scale consumer financial interfaces in the world.

Erica, BofA's in-house AI virtual assistant, is a custom natural language understanding (NLU) and natural language processing (NLP) platform launched in 2018. It is not built on a third-party LLM API but on BofA's proprietary AI infrastructure. In 2025, 20.6 million clients interacted with Erica nearly 700 million times, and Erica now resolves 98% of client inquiries without further human interaction. The internal version — Erica for Employees — is used by more than 90% of BofA's 213,000 employees, has produced a documented 50% reduction in IT service calls, and is consistent with a ServiceNow ITSM integration for ticket deflection. BofA's $4 billion AI budget in 2025 funds further Erica development, new AI-powered meeting tools (an AI meeting summarizer launched for Merrill and Private Bank advisors in 2025), and agentic workflow exploration.

For data infrastructure, the bank operates IBM DB2 databases on the mainframe for transactional records and uses Apache Spark and Hadoop for big-data analytics, risk modeling, and regulatory reporting at petabyte scale. Salesforce CRM signals appear in BofA commercial banking and relationship management job postings, consistent with Salesforce's standard enterprise financial services deployment. BofA's Zelle integration (25 million active users, $556 billion in 2025 transactions) represents a significant fintech infrastructure partnership rather than an internal build.

Bank of America's most distinctive infrastructure posture — proprietary private cloud over AWS/Azure, mainframe-first for core banking, in-house AI over third-party LLMs — signals a strong build-versus-buy preference at the infrastructure and AI platform layer. Vendors pitching generic cloud migration, IaaS displacement, or off-the-shelf LLM integration for Erica are unlikely to find receptive buyers; BofA has already made, validated, and publicly defended those architectural choices and the $2B annual savings they generate. The most productive infrastructure vendor angle is interoperability: tooling that works with BofA's existing private cloud and IBM mainframe environment rather than displacing it.

The categories with the most active vendor evaluation cycles in 2025–2026 are AI safety and governance (BofA is deploying AI at 90%+ workforce penetration and needs observability and risk tooling), LLM fine-tuning and inference infrastructure for enterprise-specific models, modern data lakehouse platforms (the transition from legacy Hadoop/DB2 architectures toward Apache Iceberg-style platforms is a live conversation across G-SIB banks), and agentic workflow orchestration for the 'Erica for Employees' next generation. CTIO Hari Gopalkrishnan has been publicly explicit about these priorities in Banking Dive and CTO Magazine interviews.

For GTM and CRM tooling, the commercial banking and relationship management lines of business are the most receptive — BofA's corporate bankers and Merrill financial advisors operate in structured CRM and revenue intelligence workflows that modern sales intelligence and AI coaching tools can augment. Any vendor must be prepared for BofA's rigorous InfoSec review (NIST-aligned, with Fed and OCC oversight), Vendor Risk Management assessment, and procurement process — typically extending deal cycles to 6–18 months. A BofA POC approval is the most important milestone; production contracts follow naturally from a validated pilot at an institution with this level of procurement sophistication.