A lot of clients and colleagues come to me with a question that has become a full-fledged pattern.
Can you build me a private ChatGPT that is secure, smart, and uses all of my own business data?
I love building systems and in most cases, off-the-shelf parts rarely fit requirements such as this. However, I’m continually amazed at what can be built almost instantly and for so much less effort when this requirement raises its head.
CustomGPT can do this in a few minutes. Sure, there’s a cost, but time is a cost, software testing is a cost. When I run the numbers, it’s difficult for me to compete with this platform but not because it compresses time and simplifies the effort to stand up a private ChatGPT instance. There are many other reasons.
CustomGPT carefully implemented an API that allows you to fully automate the flow of business information into the chat solution. It can even stream data in real-time to enhance the capabilities of your chat agent. The API makes it possible for your chatbot instance to always contain the latest business information.
I use it to pump data from sheets, Firebase, and docs into GPT chatbots every few minutes. The automation and scripting elements of Workspaces and Firebase allow me to create seamless awareness of business content and all without anyone needing to update the chatbot.
This scenario is killer.
Imagine a user asks your CustomGPT chatbot about a topic that it doesn’t currently know about. It can send an event notification to a Google Apps Script function indicating that this new topic is of interest by a user. The apps script webhook it fires uses Workspaces search to find all data and content related to the topic and pushes it into the CustomGPT chatbot instance. Seconds later, the chatbot is able to answer the question.
This means that it is possible to start with an empty chatbot instance and allow the system to magically build the AI corpus in real time.
Wow! Just Wow! This is better than GPT Plugins or GPT Functions. I love building AI solutions for clients who need to bake AI into their business processes.
Yep - this is a winner because this company thinks carefully about integration, real-time computing, and all the other aspects of LLM conversations so I can focus on delivering AI value to clients.
Oh, BTW - MIT has decided to implement CustomGPT.
It’s due to features like this (one of many!) that the brilliant minds at MIT chose to integrate CustomGPT into their systems, harnessing the power of generative AI technology (See sneak preview showcasing CustomGPT at the massive MIT AI Imagination conference)
And if you happen to be struggling with prompt development and management, check out Promptology, a free resource I created to help me build smart and often complex GPT prompts.