How do you build on more than 200 years of banking expertise to power a digital-first future? For KeyBank, an American financial institution that operates approximately 950 branches and 1,200 ATMs across 15 states, technology has long been a strategic driver, connecting relationship-driven banking with modern digital experiences. Recognizing that agility, security, and scale would define the next chapter of financial services, KeyBank set out to further strengthen its technology foundation to support faster innovation.

“Developers spent significant time managing infrastructure and working around long lead times,” recalls Azim Shaik, KeyBank’s lead enterprise architect. “That friction limited experimentation and slowed the pace of innovation, which is a real challenge in an AI-driven world.”

Starting in 2018, KeyBank shifted its data warehouse and fintech acquisitions to Google Cloud to reduce costs, bolster security, and establish a flexible foundation for growth. In the past, traditional infrastructure setup times often constrained developers and delayed new features. To eliminate these hurdles, KeyBank adopted Google Cloud VMware Engine to rapidly migrate workloads to the cloud, while deploying Google Kubernetes Engine (GKE) and Compute Engine to allow developers to build, test, and modernize applications in parallel.

“Moving to Google Kubernetes Engine and Google Cloud VMware Engine gives us elasticity, speed, and flexibility,” Shaik says. “We've reduced infrastructure provisioning down to minutes—a more than 90% improvement. Teams can self-service compliant environments in minutes instead of weeks. That dramatically increases developer productivity and enables faster experimentation.”

As the bank scaled its cloud presence, a new bottleneck emerged. While the infrastructure was modern, the millions of lines of code running on it often carried the weight of decades of technical debt. This invisible friction meant that even the most talented developers were tethered to manual, repetitive tasks—like migrating legacy Java frameworks or conducting grueling, day-long code reviews, slowing the pace of innovation. To truly unlock the power of the cloud, KeyBank needed a way to accelerate the development life cycle itself.

“With 200 years of banking history, we have millions of lines of legacy code. Coming up with a new feature, adding it up to an existing codebase takes a lot of manual steps,” Shaik explains. “We now have AI assistance available from Google Cloud.” 

Building on the bank’s existing foundation with Google Cloud, KeyBank decided to adopt Gemini for developers to support its engineering teams and empower them to focus more on creativity rather than routine maintenance. By integrating Gemini and AI developer tools directly into daily workflows, KeyBank has made multimodal capabilities available to help tackle everything from routine documentation to high-stakes migrations. 

For instance, legacy platform code often requires niche expertise, hindering ongoing maintenance and modernization efforts. Now, developers can analyze decades-old logic and create thorough documentation to ensure systems remain future-ready. If an issue occurs, teams can now quickly pinpoint the break and execute a fix, even using Gemini to generate the precise code needed to resolve it. 

“One of our engineering teams built a framework, which handled our legacy process of moving code bases at KeyBank, which usually takes 600 plus hours. We are now able to do the same thing in less than 13 hours,” Shaik adds.

Beyond speed, Gemini has also significantly improved the quality and clarity of the development process. Previously, reviewing large code branches could take a full day, but now teams can summarize changes, eliminating the noise and confusion of complex releases. The ability to scan and understand specific code areas has also helped translate technical complexity for non-technical stakeholders, providing high-level summaries about how a function works before they head into meetings.

“Teams can now deliver more outcomes without adding headcount,” Shaik emphasizes. “Shifting effort from repetitive tasks to high-value work is a huge win for KeyBank. Teams spend less time being blocked and more time solving problems together.”

Safely scaling AI across a financial institution demands an unwavering commitment to security, compliance, and trust. KeyBank needed to preserve its standing as a centuries-old institution while simultaneously evolving their codebase and scaling AI safely. Speed couldn’t supersede risk, meaning governance and security needed to be considered from the start. 

“In a highly regulated industry like ours, the biggest barriers are governance, data control, and explainability,” Shaik says. “In financial services, AI must be trustworthy and auditable before it can scale.”

To solve this, KeyBank created a retrieval-augmented generation (RAG) engine, which can retrieve real-time information from internal sources and ground models directly in its proprietary data.

“We built a retrieval-augmented generation engine on the Gemini platform, supported by a dedicated governance layer,” Shaik explains. “Grounding models in our enterprise data makes AI outputs more accurate and context aware. It turns AI into something teams trust and rely on in real business workflows.”

Operating on a unified platform enables KeyBank to establish clear policies to reduce risk, providing consistent guardrails and standardized controls for data access, audit trails, and model behavior. Through its RAG architecture, the bank is also able to ground Gemini models directly in its institutional knowledge, ensuring that every response accurately reflects current business realities and operational contexts. 

With this foundation in place, KeyBank has created a governed, repeatable framework to pursue future innovation. This safe path is allowing teams to rethink long-standing problems across every area of the business, opening up new possibilities for using AI to drive faster execution, smarter decisions, and greater efficiency that can better serve its customers for years to come.