A Fast-Moving AI Startup’s Platform Bets

Chatbase, a platform for building custom AI chat agents trained on company data, has scaled to over 500,000 monthly visitors and $4M in annual recurring revenue within 18 months. From the outset, the company prioritized development speed to compete in a crowded AI market. Its core stack choices—Vercel, Next.js, and the AI SDK—were made to minimize infrastructure overhead and accelerate feature delivery.

Why the Stack Mattered for Iteration

Chatbase’s product is an AI playground where each chat instance can be configured with different models and settings. This flexibility was a key driver behind its LLM Compare Mode, which lets users test various language models for their retrieval-augmented generation (RAG) chatbots by tweaking parameters such as model selection, temperature, and system messages. The AI SDK’s model-agnostic design made it straightforward to implement rapid model swapping, giving users the ability to fine-tune chatbot behavior for specific use cases.

Rendering and Management Efficiency

Chatbase runs both its marketing site and the main application from a single repository. The team uses static rendering for content-heavy pages to maximize SEO and load speed, while the app layer combines server-side and client-side rendering to deliver personalization without sacrificing performance. Next.js’s App Router provides granular control over which components render statically or dynamically.

Runtime and Workflow Protections

  • Security: Vercel’s Firewall shields AI applications from costly DDoS attacks, an essential layer of protection for a service that handles live customer-facing agents.
  • Skew Protection: Server and client code stay in sync during production, preventing version mismatch issues that can disrupt end users.
  • Built-in performance: Native Next.js caching and streaming capabilities are used to return fast responses from AI-driven routes.

Deployment and Growth Workflow

The team follows a trunk-based development model: pull requests merge continuously into the main branch, which is directly linked to production. New features are built on separate branches, where developers use Preview Deployments for testing and collaboration before merging.

This architecture delivers tangible user benefits: static content loads quickly, and dynamic sections render without friction. Vercel’s infrastructure also holds up during traffic spikes—a common occurrence for an AI startup when a feature launch or social media post drives sudden interest. Features like Instant Rollbacks and the option to disable auto-assigned production domains add further control, making releases more reliable and reducing stress for the engineering team.

The net effect for Chatbase is an environment where the team can focus on product innovation rather than infrastructure management, a critical advantage when moving quickly in the fast-paced AI sector.