Summary 

During the migration of one of our client’s applications, Programmers partnered with the client’s engineering team to develop custom Skills in Devin (AI coding assistant) and Windsurf (AI-powered IDE). These Skills were used to map complex architectures, enrich the client’s technical wiki with standardized documentation, and generate consistent QA testing guides. 

The result: the QA cycle for a full application release was reduced from three sprints to one and a half sprints—a 50% reduction—while also accelerating planning for the company’s Azure migration. 

Context 

In software engineering environments with distributed applications, multiple messaging services, and continuously evolving systems, living technical documentation becomes a strategic asset—especially when teams are preparing for major initiatives such as a migration to Microsoft Azure. 

That was exactly the situation faced by one of our clients. As they planned the migration of a business-critical application to Azure, Programmers worked alongside their team to develop custom Skills in Devin and Windsurf that could map architectures, document service interactions, and standardize QA testing procedures. 

These outputs were automatically published into the client’s wiki, creating a more complete and up-to-date knowledge base. 

What began as a migration initiative quickly became a strong example of how generative AI can accelerate delivery, reduce risk, and strengthen software engineering practices. 

Challenges 

The engineering team faced three major challenges simultaneously. 

Strengthening Institutional Knowledge 

Much of the application’s architectural knowledge lived in the heads of individual engineers. There was no single source of truth where teams could reliably understand how the system worked. 

Creating a structured and searchable knowledge repository became essential for continuity, scalability, and onboarding. 

Preparing for the Azure Migration 

Without a clear understanding of service dependencies and interactions, estimating migration scope—and particularly testing scope—was difficult and unreliable. 

Standardizing Testing for Messaging Services 

Many messaging services lacked formal testing documentation. Every QA cycle required engineers to rediscover how services should be tested and which messages needed to be sent. 

In short, the team needed a scalable and repeatable way to transform technical knowledge into actionable documentation. 

Our Approach 

Programmers helped drive the initiative by providing both structure and implementation support. 

The solution consisted of two interconnected artifacts developed in parallel: 

• Architectural documentation for the client’s wiki 

• Standardized QA Testing Guides 

Both were generated and maintained with AI assistance. 

1. Architectural Documentation with Devin

For each application domain, the team created a dedicated repository. 

Within Devin, custom Skills were built using two separate knowledge bases: 

• One for generating architectural documentation from scratch 

• Another for keeping documentation synchronized as services evolved 

Each repository included contextual prompts tailored to that specific domain. 

Using repository documentation and source materials, Devin automatically generated: 

• Mermaid architecture diagrams 

• Service interaction and workflow mappings 

• Cross-links between wiki pages and QA Testing Guides 

The content was then published directly to the client’s wiki, significantly enriching an existing knowledge base that previously lacked architectural depth. 

2. QA Testing Guides with Windsurf

For testing documentation, the team developed a custom Skill in Windsurf that can also be adapted to other AI platforms. 

For every NServiceBus messaging service, the generated guide includes: 

• What the service does 

• How to test it end-to-end 

• Which messages should be sent and in what sequence 

All guides follow a standardized structure, dramatically reducing the learning curve for QA engineers working with new services. 

Technology Stack 

Connecting Documentation and Testing 

One of the most valuable aspects of the initiative was linking architectural documentation and testing guides into a single knowledge system. 

From any architecture page in the wiki, engineers can access the corresponding testing guide with a single click. 

This removes friction between understanding a system and validating it. 

While the Skills were developed specifically for Devin and Windsurf, the overall approach can be replicated in any LLM or AI coding assistant with equivalent repository access. 

Results 

The initiative delivered measurable and repeatable outcomes. 

50% Reduction in QA Cycle Time 

The testing effort for a full application release dropped from an estimated three sprints to one and a half sprints. 

Single Source of Architectural Truth 

Teams now have a reliable location within the client’s wiki to understand service structures and dependencies. 

Faster QA Onboarding 

New QA engineers can ramp up quickly using standardized Testing Guides as their starting point. 

More Predictable Azure Migration Planning 

The architectural documentation provided greater visibility into service dependencies, allowing for more accurate migration and testing estimates. 

Repeatable Framework 

The model can now be extended to other application domains, increasing impact without increasing effort proportionally. 

What’s Next 

The initiative continues to evolve. 

Current priorities include: 

• Expanding the approach to additional application domains 

• Exploring automated documentation updates when repositories change 

• Sharing the Windsurf Testing Guide Skill across other Programmers teams and client engagements 

FAQ 

What was delivered in this project? 

A set of custom Skills in Devin and Windsurf that automatically generate architectural documentation for the client’s technical wiki and standardized QA Testing Guides for messaging services. 

Which AI tools were used? 

Devin was used to generate architectural documentation, while Windsurf generated QA Testing Guides. Architecture diagrams were created using Mermaid. 

Do I need Devin or Windsurf to implement this approach? 

No. The Skills were developed within those platforms, but the methodology can be replicated using any LLM or coding assistant with equivalent repository access. The value comes from the architecture, prompts, knowledge bases, and integration strategy—not the specific tool. 

How much did QA time improve? 

The QA cycle for a full application release decreased from three sprints to one and a half sprints, representing a 50% reduction. 

Can this approach be applied elsewhere? 

Yes. The framework was intentionally designed to be reusable across applications and clients by creating dedicated repositories and adapting contextual prompts to each environment. 

Conclusion 

This project demonstrates how we apply generative AI strategically alongside our clients, not as a showcase, but as a practical engineering accelerator. 

By combining custom Skills in Devin and Windsurf with a thoughtful repository strategy, the team transformed technical documentation—historically a bottleneck—into a living, scalable, and measurable asset within the client’s wiki. 

Beyond saving engineering time, the initiative elevated the client’s software engineering capabilities and laid the groundwork for larger efforts such as Azure migrations and broader AI-enabled knowledge management. 

This is just one example of how we use AI to solve real engineering challenges with methodology, scale, and measurable business outcomes. 

If your organization wants to: 

• Turn technical knowledge into a strategic asset 

• Accelerate QA and delivery cycles 

• Plan Azure migrations with greater confidence 

• Embed generative AI into engineering workflows 

Let’s talk. We’ll help design the next evolution of your engineering organization.