What if every development task could benefit from what the organization had already learned?
For a leading global insurer, that question became increasingly important as development teams looked to build on years of engineering standards, architectural principles and development practices accumulated across the organization.
One application in particular had become a valuable reference point. Over time, it brought together many of the standards, implementation patterns and design decisions that teams wanted future projects to follow. Supporting documentation explained how applications should be built, while repositories showed those standards in action.
The challenge was helping teams find and apply that guidance when they needed it. Developers needed access to standards and implementation patterns, while business, operations and IT teams often had to navigate multiple sources to locate the information most relevant to their work.
The organization saw an opportunity to change that. If those answers already existed within the organization, teams should be able to access them more easily as they worked.
When Experience Becomes Hard to Transfer
Finding the right answer often meant moving between documentation, repositories and application code, piecing together information from multiple sources. Even experienced engineers could struggle to identify the examples most relevant to the task in front of them.
For new team members, the challenge was even greater. Before contributing to new projects, they first had to understand the structure of the application, learn the organization’s development practices and determine which standards applied to different components. Often, this meant a manual search or relying on experienced colleagues who understood the application and the decisions behind it.
As development teams grew and new projects emerged, the challenge became even more pronounced. Though standards and working examples already existed, applying them to specific tasks still took considerable time and effort.
The insurer needed a simpler way to connect documented standards with the code that demonstrated them.
Creating a Single Source of Guidance
To address this challenge, Persistent began with the documentation that had guided the development of the reference application. The material contained detailed principles, standards and implementation guidance, but its size and structure made it difficult to navigate efficiently.
The team converted the content into structured Markdown and organized it within a Docusaurus-based knowledge environment. The Knowledge Base brought together product information, business workflows, architecture guidance, technical documentation, testing assets and support knowledge from multiple sources, including source code, business and technical documents, manuals and SME inputs.
At the same time, the application’s repositories were enriched with AI-generated explanations and organized around technologies, frameworks and application components.
The result was a centralized knowledge experience that could serve multiple teams across the organization. Business teams could quickly access product and process information, Operations teams could retrieve troubleshooting and support guidance, while IT teams could explore requirements, architecture, dependencies, APIs, integrations and implementation patterns. Instead of searching across disconnected sources or relying heavily on SMEs, users could access relevant information through a single experience tailored to the context of their work.
Supporting Developers Where Work Happens
Once the documentation and code had been organized, Persistent connected them to the insurer’s enterprise AI platform through an orchestration framework. When application-specific questions were submitted, the solution routed those queries to the most relevant source and returned contextual guidance or code examples.
Developers could ask questions in plain language and receive answers grounded in the insurer’s own standards, documentation and repositories. Rather than navigating multiple systems, they could review implementation approaches, understand dependencies and access relevant guidance within the same experience.
The AI Agent was designed to support teams throughout the software development lifecycle. Beyond helping users locate information, it could assist with requirements analysis, design and planning activities, code reviews, vulnerability identification and unit test generation. Because those capabilities were grounded in the organization’s own documentation and code, teams could apply established standards and practices more consistently across projects.
Extending the Solution Beyond Information Retrieval
As the project evolved, the focus expanded beyond helping teams find information. Through multiple MVP phases, Persistent refined the solution based on user feedback and real-world usage.
One requirement that quickly emerged was support for practical development activities. Teams not only needed access to standards and examples but also ways to apply them while building new functionality.
The solution also enabled better alignment across frontend, backend and testing teams. With shared access to business rules, implementation examples, APIs, testing assets and troubleshooting knowledge, teams could work from a common understanding of requirements, reducing dependency on manual handoffs and repeated knowledge transfers.
Creating a Repeatable Model
The solution was built using loosely coupled components, making it easier to bring additional applications into the framework over time. As new applications are added, their documentation, standards and implementation patterns can become part of the same experience, allowing future teams to build on proven approaches rather than starting from scratch.
What began with a single reference application can now be extended across the organization, creating a growing body of guidance that future teams can draw upon as new projects take shape.
Building on What Has Already Been Learned
The insurer had already done the hard work. Over the years, its teams had defined engineering standards, refined architectural principles and demonstrated how those ideas could be put into practice through a successful application.
Persistent helped bring those standards, implementation guides and code examples together within the organization’s existing AI-assisted development environment, making them easier to access at the moment they were needed.
As a result, teams building new applications no longer had to spend as much time searching for examples, interpreting standards or tracing decisions through documentation and repositories. Developers could access the information they needed through a single conversational experience rather than navigating multiple systems.
The impact was particularly noticeable for developers who needed to understand organizational standards and implementation patterns before contributing to new projects. According to the project team, the solution helped reduce developer search and onboarding effort by as much as 70%. Onboarding itself was estimated to shrink from roughly three months to approximately 15 days by giving new team members a more direct way to find guidance, understand application design decisions and access proven implementation examples.
Creating a Foundation for Future Applications
While the immediate benefit was easier access to information, the longer-term value lies in how the model can evolve over time.
As additional applications are brought into the framework, their documentation, standards and implementation patterns can become part of the same experience. Teams no longer need to rely solely on institutional memory or spend time rediscovering approaches that have already proved successful elsewhere. Instead, they can build on a growing body of guidance shaped by projects that came before them.
That shift is ultimately what the insurer set out to achieve. Instead of treating every application as a separate body of work, it created a way for future teams to build on what previous teams had already learned.
