English | EN 
Over the past nine months, one of our development teams had built an application for our client to support requirements gathering for their own customer projects. Work on what started as a straightforward web and native form application on Microsoft’s Power Apps platform had begun with promise. But as requirements and expectations expanded, the application’s scope grew beyond what our team could support on the platform. Time was ticking on the engagement, and progress had stalled. With only six weeks remaining, our client wanted to know how we were going to meet our obligations and complete a production-worthy product suitable for public release.
The solution: toss out the original application, assemble a new, three-developer team and rebuild it from scratch as a traditional enterprise application. Implementing a new version from the ground up would allow us to create a performant, scalable product that provided the basis for building the advanced features the customer wanted in the future.
How could we deliver in the allotted timeframe though? Comparable applications typically take multiple months with larger teams to complete. Sure, we were starting with an existing application and requirements we could reference for our new build, but the challenge remained daunting.
Our development team was able to meet our commitment through comprehensive use of agentic artificial intelligence tailored to write application code. The technology we chose was Claude Code, a command line tool from Anthropic. Claude Code is designed to leverage large language models and integrate directly with the custom code bases its users create to complete a variety of software development tasks. From architecture and feature design to implementing application code to conducting code reviews and writing unit tests, Claude Code provided us with a tool to rapidly build an enterprise application with quality comparable to traditionally built applications in a fraction of the time and headcount.
Two of the developers had prior experience using the tool to build enterprise software, though on a smaller scale. Equipped with Claude Code subscriptions and a mandate to use the Agentic AI platform to build the entire application, “Vibe Coding”-style, the team wrote little to no code, configuration, tests or other development artifacts manually ourselves. Instead, by providing to Claude’s models detailed instructions and examples of what we sought to achieve, AI agents conducted the low-level code creation. The human developers served in roles more akin to development managers, guiding their junior resources on what to build and correcting them when they made mistakes.
In approximately six weeks, our team was able to reach feature parity with the original application, which had taken nine months to develop. The application, built with React Native, supports iOS, Android and web versions; includes dozens of input fields; implements sophisticated role management and transition workflows and supports both online and offline use. It is not just a demo or proof of concept. It is a real enterprise application intended for business users to support their day-to-day work.
The results we achieved using an Agentic AI-assisted development process were astounding and clearly demonstrated the promise the technology offers in enhancing time efficiency in software development. I would caution that it is still early days for this technology and the practices that enable its effective use. I credit much of our success to the knowledge and skill of our team’s developers. Our prior experience building software the old-fashioned way served us well. We have the intuition about when the outputs Claude provides are on the mark and when the model goes astray that is essential. Having humans in the loop who possess the right technical skills to guide development is indispensable to achieving high-quality results in a fraction of the time it would have taken using a traditional process.
In future posts, I will dive into the specific tips and tricks, Dos and Don’ts, that fostered our success, as well as strategies and patterns we are investigating to further enhance our Agentic AI-assisted development process. As 2026 begins, the promise of leveraging these new tools to build high-quality software more rapidly than ever before is unfolding. I am excited to see what the new year brings and how we can continue to push forward, innovate and grow.