This is cross posted from LinkedIn.
TLDR – I created a LLM to read and highlight keywords in RFQ’s so they were easier to evaluate and route to the proper people inside my company. Stick with me and I’ll tell you how to make your own.
Reading is a Drag
Request for Quotations, Qualifications, & Proposals come at me like a firehose pumping water. Way too fast to drink. It bothers me as a business development guy to think that inside any particular bid (that’s 100 pages long), there might be an environmental scope that I don’t know about.
But reading takes time, and time is money. I’m continually conflicted between feeling anxiety for not reading a bid, and feeling anxiety for spending too much time reading bids and not doing more important work.
We’re Generally a Subcontractor
Environmental Consulting scopes are less frequently the main thrust of a bid request. Most times it’s buried inside an engineering project of some sort. Our scopes feel like we are searching for a needle in a haystack. I wanted to have a better way to evaluate opportunities that might be marginal based on the title of the RFQ but could be good projects for my team.
How I Built My RFQ Reader (Robot)
In my opinion there are only 2 good LLM models, Anthropic’s Claude and OpenAI’s ChatGPT. My company has a subscription to ChatGPT so that’s the one I used. There are pro’s and con’s to using ChatGPT for this, one of which is ChatGPT is hard to say. We call our LLM, “Sparky“. That’s how I’ll refer to him going forward.
The paid, enterprise, version of ChatGPT is $25 a month. It won’t break the bank, you get less functionality when you use the free version. You’ll need the paid version to see the tools I’m using.
Keyword Text File
Sparky is really good at summarizing longer text but I didn’t want a summary. I wanted the text unchanged and my eyes drawn to specific keywords that indicated environmental scopes.
I created a text file, as opposed to any other format (PDF, DOC, MD) because text files are easy for machines to read and there’s not a lot that can go wrong with them from a corruption standpoint. The text file contains all the keywords I wanted to find.
Markdown File Instructions
LLMs are good at writing what are called “Markdown” which some millennial decided was better than a “Markup” file. If you’re Gen-X like me, you’ll remember the M in HTML is “Markup”. Because the LLM is good at writing the .md files, it’s also good at reading them and they are useful for instructions, back to the LLM. (It appears they won’t let me link a .md here, so I included as a TXT)
Putting It All Together
- On the left side navigation bar, click on “GPTs”
- Across the top, on the right, click “My GPTs”
- Click + Create a GPT
- See the image below, it’s pretty explanatory.
How to set up your GPT =>
See big Picture =>
What The Result Looks Like
This is a sample bid I ran through the machine.
Publishing & Sharing
Once you’ve saved your GPT, you can share it with people in your company so they can use the tool too without having to remake it. That Icon is in the upper right hand side.
When you use the tool, the button in the center of the screen starts the conversation. The LLM will ask you one question about your file at a time, beginning with, “Please upload the bid”. After that, the rest is auto-magic.
