π – Step 1 – Read Project Files

Before finalizing the requirements, I put a good few of prompts to the AI. Initially, my approach was scattered, but gradually, I learned to provide more logical and structured requirements based on the feedback from the AI.

Sometimes, the AI just deviates from the intended path. When this happens, it’s crucial to correct the course and ensure that the AI understands your requirements by providing a final version accompanied by output examples. This allows our human eyes to easily spot any differences or discrepancies.

Here’s how I approached it for ensure correct understanding of the requirements for the AI:

• Iterative Prompting: Begin with broad prompts and refine them iteratively. This helps in honing the AI’s understanding of what you need.
• Logic and Structure: You may not sound very logic at the start, by keep iterating from the first step, your asking will be more logical and structured, which will push AI to give you more details and better suggestions.
• Feedback Loop: Ask AI for examples to establish a feedback loop, ensuring that the AI’s output aligns with your expectations.
• Comparative Analysis: With output examples, perform a comparative analysis to quickly identify inconsistencies or misinterpretations.

By following these steps, I was able to achieve more precise and useful outputs from the AI, enhancing both efficiency and accuracy in project management.

If you are curious about what happened, here are those inputs:

Prompt 1: I need to build a program running on the cloud, but before doing so, i need to run it locally on my mac, here is a requirement, I have a list of files sitting in different folders, with same file format, .mpp files, which are ms project files, each folder has a date partten, if it says 8.1, it means august 1, if it say 08.01 or 8.01, it also means August 1. I need you to write a program that will go into each folder and be able to turn all those .mpp files into jason files like a massive data file in json format. and once you can detect new files in new folders, you will gather those files and turn them into json format, i will use the json file for analysis purpoese, if you understand my requirement, please list my requirement in details one by one

AI now asked

✅ Before I move on to coding, can you confirm:

• Do you want one big JSON file containing all projects, or one JSON per folder (with a master index maybe)?
• Do you have a preferred library for reading .mpp files? (There are Python libraries like mpxj via JPype or pymsp, but support may vary.)

Would you like me to suggest the best Python library for reading .mpp files on macOS next?

Prompt 2: one big JSON file, mpxj is fine. can you give me a json example so that both of us are on the same page. Here is what i have in the .mpp file. each file contains a lot of projects inside, each project is a summay task, under each summary task, there is a few keywords to indicate certain development actions in sequence, HLD, DDS, Coding, QA, UAT, each sub-task is assigned with resource names, and each name has its own format, if it a name contains QA, it means this resource is a QA, if a name contains Dev, then it is a developer, if a name contains PDM, then it is a product manager.

So, AI generated a json file for me to review, and I want to add more requirements as below

Prompt 3: each task has their own start and finish date, please also include those in it

So, AI updated again. Now I realized the folder structure as below

Prompt 4: the folder structure is like this: ProjectFiles/2025/8.August, this folder path means under the root folder "ProjectFiles", we have a year folder, that is called 2025 and so on, the folder_date should be buillt by looking at the folder path, do you understand?

AI now asked

Do you want the folder_date to always be YYYY-MM (year + month only),
or do you want me to go deeper (e.g., if there’s a day subfolder like 8.1 or 8.01, then YYYY-MM-DD)?

Prompt 5: under 8.August folder, you will see a few different folder, if you see 8.1 or 8.01, that means August 1, do you understand, that is the exact folder's date

AI now asked

Do you also want me to store the raw folder path (like "folder_path": "ProjectFiles/2025/8.August/8.01") in the JSON for traceability, or should I just keep the normalized folder_date?

Prompt 6: usually there are four teams project files, if the file name contains Phoenix, then this file comes from team "Phoenix", if the file name contains Universal, then this file comes from team "Universal and APEX" team, if the file name contains xmen, then this file comes from team "Xmen", if the file name contains Terminal, then this file comes from team "Terminal" team, each team should have their own json file, but across different folders, the same team's files should be in one json file, am i clear about it?

AI now asked

Do you want these JSON files to be continuously updated (e.g., if new .mpp files are added later, the program appends them into the existing team JSON file), or should it regenerate from scratch each time you run it?

Prompt 7: I want these JSON files to be continuously updated, do not run from scratch because it will waste a lot of time

AI now asked

Do you want me to implement the continuous update by:

1. Keeping all processed file paths in one index file (like processed_files.json), OR
2. Storing processed file info inside each team’s JSON file (so each team JSON knows what files are already included)?

Option 1 keeps things cleaner and easier, but option 2 keeps everything self-contained per team. Which do you prefer?

Prompt 8: use option 1 please, but in the main json data file, at the very top, because we know how many files that we have processed, which means we know the date range, and the number of files, could you list those at the top using optione 1

AI now asked

Do you want me to make the summary date range strictly based on the folder_date (from the folder path like 2025-08-01), or should it use the task start/finish dates extracted from .mpp files?

Prompt 9: strictly based on the folder_date please. can you also add the summary info into the team.json file as well so that it is easier to access this info when needed?

AI now asked

When updating the summaries (earliest/latest date, total files), do you want me to recalculate from scratch each run (to stay 100% accurate), or should I just incrementally update based on the new files only?

Prompt 10: just incrementally update based on the new files only

---

There are more and more questions and answers back and forth, like prerequisites, database schema, python program structure etc. This process continued for a total of about two hours. I kept refreshing the AI’s understanding of our current progress by asking if it grasped our requirements, and requested examples to illustrate its comprehension. Through this method, I was able to clarify my own needs. This brainstorming session proved to be extremely useful in helping me understand what I actually required.