Paf Increases Developer Productivity via ChatGPT Enterprise and Custom GPTs

Paf, an international gaming company, has integrated ChatGPT Enterprise across its organization to automate routine development tasks and accelerate employee training. By deploying a suite of specialized custom GPTs, the company has achieved a level of productivity that its Chief Technology Officer, Fredrik Wiklund, estimates is equivalent to the work of 12 full-time employees.

Model Selection and Evaluation

Paf selected GPT-4 after conducting head-to-head accuracy and cost comparisons against LLAMA and Claude. The evaluation revealed that GPT-4 was 25% more accurate than its competitors without incurring higher costs.

Streamlining Development with Specialized Custom GPTs

Paf's engineering team of 100 developers utilizes over 85 custom GPTs to reduce manual effort and minimize hallucinations by avoiding model overloading. Rather than relying on a general-purpose model, developers chain together a suite of specialized coding GPTs to generate standardized application flows and APIs:

  • Swagger GPT: Converts Swagger JSON API definitions into TypeScript service endpoint definitions based on Paf’s coding standards.
  • TypeScript GPT: Writes backend service code using endpoint definitions and existing session validation functions.
  • GraphQL Nexus GPT: generates GraphQL Nexus schemas and integrates helper functions for frontend interaction.
  • Relay GPT: Creates React Relay hooks using GraphQL Nexus schemas for backend communication.
  • React GPT: Writes React components following Paf’s React and TypeScript style guidelines and core component library.

Frontend developer Krista Koivisto notes that these focused GPTs allow the team to "automatically generate functioning boilerplate implementations with far less effort."

AI-Augmented Developer Training at grit:lab

Paf has integrated ChatGPT Enterprise into the grit:lab coding academy to train 65 aspiring developers. This approach shifts the focus from syntax and basic coding errors to high-level systems architecture. Students use the AI for:

  • Understanding new programming concepts
  • Efficiently debugging code errors
  • Debugging code errors efficiently
  • Learning syntax and structure across different languages
  • Rapidly generating test data

DevOps engineer Kim Gripenberg states that junior developers and grit:lab students progress "years faster" with AI assistance, enabling them to think at a higher, systematic level from the start of their training.

Organizational Impact and Future Integration

Beyond the engineering team, 70% of Paf employees across finance, HR, marketing, and customer support actively use ChatGPT Enterprise. The company plans to fully integrate ChatGPT Enterprise and the OpenAI API into all internal processes over the next year.

Chief Technology Officer Fredrik Wiklund envisions a future where GPTs handle more of the writing, testing, and deploying of software, allowing human developers to focus exclusively on higher-level systems work.

Sources