Paf built 85 custom GPTs and says they do the work of 12 people

The gaming company Paf didn’t just dabble with ChatGPT Enterprise. They went all in, and the numbers coming out of their engineering department are the kind that make CFOs sit up straighter. After testing models including LLAMA and Claude, Paf’s team found GPT-4 was 25% more accurate without costing more. That was enough to roll it out company-wide, but it’s what the 100-person engineering team did next that’s genuinely instructive. They built a chain of over 85 custom GPTs, each handling a discrete step in the development workflow—from converting Swagger API definitions into TypeScript endpoints to generating React components that conform to Paf’s specific style guidelines.

Frontend developer Krista Koivisto uses ChatGPT around 20 times a day, and the tailored approach is deliberate. By chaining focused models together instead of throwing everything at a general-purpose window, the team found they could curb hallucinations and generate boilerplate code that’s actually deployable with minimal tweaking. The GPTs have names that map directly to real tasks: Swagger GPT, TypeScript GPT, GraphQL Nexus GPT, Relay GPT, and React GPT. Each one knows Paf’s internal standards and reuses existing helper functions, which is where the productivity surge really comes from.

The impact extends beyond seasoned developers. Paf integrated ChatGPT Enterprise into its grit:lab coding academy, where 65 aspiring developers now learn with AI assistance baked into the curriculum from day one. DevOps engineer Kim Gripenberg observed that junior developers and students are progressing years faster, thinking at a higher, systems-architecture level rather than getting stuck on syntax. It’s a fundamentally different approach to training—one that produces developers who see the whole application, not just the function they’re writing.

Chief Technology Officer Fredrik Wiklund estimates ChatGPT is handling the equivalent workload of 12 full-time employees. With 70% of Paf’s 315 employees now actively using ChatGPT Enterprise across finance, HR, marketing, and support, the company is betting its entire operational tempo on AI integration. Wiklund’s blunt framing—“Either you are on the train, or you are back at the station, watching it leave”—reflects a leadership team that views this as an existential shift, not a pilot program. The next year will see them push for full API integration across all processes, aiming for a development velocity that punches well above their weight class.