No-Code GPT Plugins?
It's not often that you need to scale AI access to your business content, but when you do, there's a really simple pathway to near instant conversational AI.
I’ve written at length concerning make vs. buy decisions that many organizations face concerning the accessibility of documents and other files through conversational AI. Most of my thoughts and experiences suggest the buy-side is winning. There are exceptions, but the advent of ChatGPT and the drunken stupor it created, leaves me wondering how soon before we sober up.
Chatbots that lean on LLMs have demonstrated profound abilities to help everyday workers access the inaccessible or unfindable. They’re shaved minutes and hours off tasks. Sure, they’re not perfect, but neither are we.
The road to AGI is scary for many businesses, and rightly so. There are new technologies, new types of APIs, and new risks to be considered. It’s a tenuous path with lots of uncertainty.
But all businesses know this.
Information workers spend more than 2 hours daily probing for information
to do their jobs.
AGI is beginning to put a dent in this productivity crisis, and chatbots are at the forefront of that fight. However, unsurprisingly, false claims concerning AI have emerged. I’ve grown increasingly weary of the constant pummeling of Twitter with these absurd claims.
Create a GPT Chatbot for Free in Just Five Lines of Code!!!
Setting aside the other 5495 lines of code you’ll need to make it all work with your data, tantalizing offers like this should provide you with something valued at roughly precisely what you paid -
Another ChatGPT app incapable of doing anything.
These claims produce the AI equivalent of a street vendor’s Rolex knockoff. Maybe this guy knows some Python too.
Please don’t fall for it. The pathway to commercial-grade AI software is not trivial.
As I see it, there are two paths to consider when standing up business-level conversational AI applications.
Make it.
Rent it.
Make It
What’s involved? Experience with Embeddings, OpenAI APIs, LLMs, prompt engineering, a tech stack like a chat UI framework, servers, security, etc. Many companies have followed this path and succeeded. Most haven’t achieved their hopes, and you can read about many of them in the OpenAI forum.
Rent It
Pay some really well-qualified people to rent you their platform. There are many like those listed here. My company uses CustomGPT. We’ve tried ChatGPT (no API, hard pass) and others. We wanted something that fully embraces automated processes. Pay close attention to the API of all these vendors because it signals good architecture if you want your document assets to flow magically into your AI systems automatically.
Keep reading with a 7-day free trial
Subscribe to Impertinent to keep reading this post and get 7 days of free access to the full post archives.