Making a Custom MYGPT for organizational purpose

GPT builders
1

Background – We are an organization wanting to make a private customGPT that can be used by our employees + departments in the effort to minimize research time. We are General Contracting Company so we work with many outside entities (clients, client reps, vendors, engineers, architects).
Basically we want to create an AI assistant who can parse company wide data and provide answers to questions anyone might have.

We have acquired a ChatGPT Teams subscription.

Use Cases – (These are very crude examples…)

  1. Compare the AR and AP of project ____ and consolidate what we can pay out to vendors provided we have received payment from clients.
  2. List all open projects running between March 2023 and December 2024 that made a loss.
  3. Make a MS project schedule for this project with based on past projects in the prior year.
  4. List all projects completed at ____ address where ____ was the project manager.
  5. Give me the schedules of the last 10 projects we completed with a certain client.
  6. How many items are late across all projects.

Dilemma –

  1. How to provide real-time data on network drives?
    We have network drives where all our information is stored in folder structures. A lot of info concerning projects was put into a spreadsheet and uploaded into the knowledgebase of the customGPT. But the knowledge base does not allow more than 5 uploads and the drives are updated in Realtime. Need some way to have customGPT access the network drives and query them to answer questions.

  2. How to provide users with permissions and access.
    The customGPT was made but giving others admin access is problematic. Anyone in the workspace can access it but it seems only the creator is able to get to the configure settings page.

Any and All suggestions welcome. Thank you.

1 Like
2

This may help.

1 Like
4

1- You need to implement your own OpenAI base assistant API, connect it to your own backend and enforce your permissions and access rules on your own backend.
2- You need to use RAGs to connect all your data to ChatGPT assistant or CustomGPT
3- As next steps a little fine tuning on GPT models you use for assistant will help further with more expert tasks such as working with MS project files and stuff.
Please check out AlgorandGPT and DID-GPT to see such vastly connected custom GPTs querying and using terabytes of data at will.

3 Likes
6

Yes i believe this is the proper route. I just learned of RAG systems yesterday, Llama Index to be precise. Also i was looking at Microsoft and there is something called Azure Open AI
There is a 15 min video explaining it all towards the bottom.
I believe this is a better suited solution.

1 Like
7

Yes 100% will work, but will be pricey to build and run. But it would be cutting edge. :honeybee::rabbit::heart: