How Axos Bank is making agentic systems its competitive advantage

Challenge

Winning over the development team

Axos Bank is the first fully digital bank in the United States and a pioneer in the branchless banking model.

When Kevin Hearn joined the company in 2020, Axos Bank had already adopted OutSystems for its rapid development capabilities. But the platform’s full potential to modernize systems and improve developer productivity remained largely untapped.

"I inherited an IT team composed of Angular and C# developers that had typical preconceived notions about OutSystems," he recalled.

Kevin, however, saw things differently. When he managed a program for recent technology graduates that trained them in various technologies through rotation cycles, he noticed that the OutSystems track yielded the best results. "Developers trained in OutSystems looked at problems in a way we hadn’t before, pursuing experimental ideas and concepts," he shared.

That’s when the tables turned.

Solution

Leading with conviction

Focusing on developer buy-in first, Kevin made a few changes. "I moved a few developers that were still skeptical about OutSystems to different groups and kept those who were genuinely interested in learning the platform," he explained. This approach was crucial to assembling the right team.

The next stage was winning over the business. After seeing the potential of OutSystems to improve clunky processes, Kevin decided to tackle projects with high technical debt that hindered operations. "We had several processes in on-premise platforms like Microsoft SharePoint and Nintex K2. So, we partnered with business stakeholders to understand how we could improve them," he shared.

With just two developers and an average two-month turnaround per project, the OutSystems team successfully modernized 12 critical business processes. "It was extremely fast and cost-effective, and it made the business realize the platform’s power," Kevin concluded. Today, Axos has fully sunsetted the K2 platform.

A structured approach to agentic AI

As the team’s confidence grew, Kevin set his sights on a new frontier: agentic AI. But he rejected the idea of ungoverned AI adoption, as tool sprawl and fragmented systems posed security risks the bank couldn’t afford.

"Give AI to 300 people without clear direction, prompts, or guardrails, and you’ll get 300 different results, while potentially exposing intellectual property and rapidly driving up related costs," Kevin added.

To avoid this, Kevin and his team established a top-down strategy that started by defining the authorized tools, then providing developers with pre-configured workflows to automate daily activities, and finally, allowing the team to build and deploy autonomous agents.

Using Agent Workbench, Kevin’s team deployed their first agent in just two weeks, followed by three additional agents within a month, including:

• Intelligent Platform Monitoring Agent: This agent automatically ingests and categorizes application logs into a real-time dashboard, enabling informed decision-making and replacing a previously manual, error-prone process. In just three days, Axos delivered a second iteration adding observability platform integration, automatic alerting, and AI-proposed code changes to improve performance and reduce tech debt.
• Legacy Modernization and Reverse Engineering Agent: This agent connects to Azure DevOps repositories, and extracts and analyzes code at a feature level. It autonomously generates requirements documents, system architecture diagrams, and API specifications that are fed directly into development environments to rapidly rebuild applications.
• Business Analyst Agent: This agent enables users to ask questions about existing logic and functionality of an application or feature. It is very useful when evaluating legacy applications for modernization, where comprehensive documentation or resident knowledge within the business or developer community is inadequate.
• Dev-Mate Agent: This agent connects to Azure DevOps and enables developers to ask questions about their assigned work items, whether at the individual task level, or at a higher Features and Epics levels. It provides greater context for developers about how their work fits into the larger initiatives, who else is contributing to the same area, and other related information.

Building on this momentum, Kevin has encouraged the team to adopt the newest AI capabilities of OutSystems, positioning Axos Bank to be an early adopter of Agentic Systems Engineering. The team is now exploring the next generation of Mentor to automatically identify and resolve technical issues, as well as auto-generate documentation that helps business users navigate new tools more effectively.

Results

Faster time to market at 40% of traditional development cycle

In the first six years since adopting OutSystems, Axos Bank has launched 32 apps, countless enhancements, and four AI agents.

The bank is now delivering applications at a much lower cost than traditional development. Kevin explained: "We ran several competitions where I pitted a traditional Angular developer against an OutSystems one. The development cost in OutSystems was about 12-20% of the cost in Angular, considering speed and resources." The result was a 66% improvement in developer productivity with just a third of the resources in OutSystems projects.

Additionally, by retiring legacy workflow platforms, Axos is saving $150K annually while increasing business productivity and reducing errors.

The four AI agents the team built with Agent Workbench have also delivered immediate and measurable results. By transitioning from manual oversight to an autonomous agentic model, Axos Bank:

• Reduced overall platform technical debt by 10% in the first 60 days.
• Improved the performance of critical application services by up to 50%.
• Compressed a three-year modernization roadmap into a single year.

These agents have also changed how the team operates. Axos’ Intelligent Platform Monitoring Agent provides real-time alerts to development teams and network operations centers, identifying issues before end users are ever affected. It also eliminates manual filtering and reviewing of logs, freeing support personnel from hours of daily manual analysis.

These results validate Axos Bank’s shift toward Agentic AI, proving that courageous leadership and the right technology can turn a development team into a powerful engine for change.