Planning of AI in the Product Development Process

[samshdn]

We use various AI tools in our daily work. One type is web-based AI tools, such as ChatGPT.com and Grok.com. Another type is IDE plugins, which excel at providing coding assistance.

Another way is to use AI through APIs, which is the topic we are currently discussing.

Through an LLM gateway like requesty.com, we can programmatically read local data as context, connect to different LLMs, instruct the LLMs to perform tasks in batches, and present the LLM outputs in various formats.

For the data mapping task in the current project, we can explore using LLMs. In the short term, LLMs can assist in analyzing critical missing logistics data, and based on their outputs, we can improve data mapping rules. In the long term, we can consider transitioning from rule-based data mapping to using models to directly generate mapping status codes.

[jduan00]

Two potential future automation scenarios enabled by AI:

1. Automatically improve code quality: app log monitor --> LLM (trained with our code base) --> generate suggested fixes for developers.
2. Proactively enhance data mapping: An LLM can scan cached tracking data to identify any missing major status codes, and then send suggested fixes to the developer.
3. Scan the cached data for any requests that are taking an unusually long time to process. Then, send reports and suggest improvements to the team.