How we build tools with the OpenAI API

This date is about to get a little more technical. We built a tool that uses the technology behind ChatGPT directly through the OpenAI API.

The idea came from a simple problem: tech teams continuously release updates and new features, but often do not have enough time to document them properly.

As a result, Customer Success, support and marketing lack important information. Users can also struggle to understand how new features work.

Our approach is automated documentation in both video and text form. The content can then be used for FAQs, marketing material or chatbots.

A crawler goes through the website or web platform and stores the HTML code and a screenshot for every page.

This data is saved in a database and connected to ChatGPT. The system analyses possible user inputs and creates descriptions of the different use-cases.

Based on this information, the tool generates a Playwright script that automatically reproduces the relevant user interactions.

While it runs, a video is recorded. The video is stored centrally together with the screenshots and HTML.

The result is video and text documentation of the most important software use-cases that can be generated automatically and reviewed later.

We also use ChatGPT or Google Gemini in day-to-day project work. One approach is to create a separate chat for each client, with access for relevant employees or freelancers depending on the project.

Relevant documents can be added after discussing it with the client and anonymised where necessary. This gives future prompts access to the specific project context.

AI can also handle simpler tasks such as compressing images, removing individual pages from PDFs or converting files into different formats.