Skip to content

AI-Assisted Innovation: Streamlining Documentation of Low-Code Workflow Tools

I was recently tasked with documenting the functions and features required for a new low-code workflow tool. This tool is intended to complement our existing toolset and duplicate some of its functionalities. My objective was to arrange the functions and features in a grid-like layout, categorising them based on frequency of need, importance, and whether they were critical.

Attempt 1: AI-Assisted Documentation

Our current tools have comprehensive documentation, but manually extracting the necessary information would be a laborious task. I considered leveraging AI for this task. Starting with the simplest approach, I used a single prompt directed at the online documentation. However, the results were subpar; the AI did not navigate through the pages as expected.

Attempt 2: Sitemap Strategy

Next, I created a sitemap of URLs and fed it to the AI. This resulted in an overload of URLs for the AI to process, leading to rate limiting by the website hosting the documentation.

Attempt 3: Targeted URL Approach

My third strategy involved a list of specific URLs. This approach began to yield better results and allowed me to refine my prompts. However, it was still too time-consuming due to ongoing rate-limiting issues, requiring me to segment the URLs and make extensive edits.

Attempt 4: Text Scraping Solution

For the fourth attempt, I wondered if feeding the AI text extracted separately from the URLs would avoid rate-limiting and allow for processing in larger batches. I used AI to develop a C# application that accepted a CSV file of URLs and output the data to another CSV file of text. The initial output had too much whitespace and was unwieldy. I then refined the AI’s task to produce text output, organised and titled by page, while also removing unnecessary whitespace.

The Final Steps: Data Formatting and Utilisation

With the text now appropriately separated and the prompt refined, I started to receive usable results. The next step was transferring this data to a Word document and formatting it correctly. Eventually, I documented the features, descriptions, and use cases. However, I needed them in an Excel table for scoring purposes. Feeding the formatted text back to the AI, I asked it to generate a CSV file for export and completion.

Conclusion: AI as an Efficiency Enhancer The process required significant effort and time, but I achieved the desired result. If I had done this manually, time constraints would have likely forced a compromise in quality. This experience reaffirmed my belief that AI doesn’t replace our best workers; instead, it enables them to produce high-quality work more efficiently.


Join the conversation

Your email address will not be published. Required fields are marked *