Vibe Coding: Chatting Your Way to a New App with ServiceNow Build Agent

[Earl Duque]

As developers, we’ve all had those late-night conversations about the future of our jobs. Will AI replace us? How do we stay relevant? On a recent live stream, Robert Ninness, an Outbound Product Manager on the ServiceNow AI team, and I, Earl Duque, sat down to tackle these questions while exploring ServiceNow's new Build Agent. The experience was a lot like "vibe coding": a free-flowing, conversational approach to building applications with an AI assistant.

The New Foundation for Development

Before we dive into the vibe coding, it’s important to understand the new developer stack that makes this all possible. Historically, building on ServiceNow meant working directly in the runtime, which is very different from traditional software development. The new ServiceNow IDE, ServiceNow Fluent, and ServiceNow SDK change all of that.

• ServiceNow IDE: This is your new development environment. It's file-based, just like what you'd see in VS Code, allowing for modern development practices like source control and CI/CD pipelines.

• ServiceNow Fluent: This is a declarative, TypeScript-based language designed specifically for ServiceNow metadata. It’s the code that the Build Agent understands and uses to define your tables, business rules, and UI pages. It's the key that allows an LLM to "talk" to the ServiceNow platform effectively.

• ServiceNow SDK: This provides the tools to build, test, and deploy your Fluent-based applications. It's an npm package, which means you can use it with the world of open-source JavaScript libraries.

The "Vibe Coding" Workflow

So, what does vibe coding look like in practice? Robert and I took a user suggestion from the live chat, a Visa Sponsorship application, and watched the Build Agent go to work. The process is a collaborative loop:

1. Prompt: You start with a simple, high-level request, like, "Create a visa sponsorship app."

2. Plan: The Build Agent, which is powered by Anthropic's Claude, plans out the app's structure. It decides on tables, roles, business rules, and UI pages.

3. Build: It uses its internal tools to write the Fluent and JavaScript/React code, creating the files and directories for your application.

4. Validate: The system runs a build, and if it finds any errors, the Build Agent fixes them on its own. It's like having a very excitable junior dev who learns from their mistakes.

5. Refine: Once the base app is built, you can give it new instructions: "Add a Kanban board," "create a user guide," or "fix the layout."

In our demo, we saw how the Build Agent created a visa app in minutes, complete with tables, roles, and a simple UI. We also saw an app that Robert had previously built: an Executive Briefing Center (EBC) app with a drag-and-drop Kanban board and React-based UI, all created with minimal effort. This process is a huge time-saver and lets you get to a working prototype in a fraction of the time.

The Bigger Picture: Vibe Learning and the Future of Dev

But the fun part is also using these AI tools to help us learn. We can use it to:

• Learn a new framework: By asking the agent to explain its code, you can quickly understand new concepts like ServiceNow Fluent.

• Debug more effectively: We saw how the agent could get stuck in "negative feedback loops," and it’s these moments when a developer's skills become invaluable. Knowing when to jump in and fix things manually is what separates a proficient developer from a prompt engineer.

• Level up your soft skills: The ability to clearly articulate a problem and provide precise instructions is the most critical skill for vibe coding. It forces you to think like a product manager or business analyst, not just a coder.

The truth is, AI isn't going to replace developers, but it will change our roles. It will handle the mundane, repetitive tasks, freeing us up to focus on the truly complex problems: architecture, integrations, and user experience.

If you want to see the future of development on ServiceNow, check out the full stream. And let us know what you think in the comments!