I love this and the fact you are taking the proactive approach in designing your own. It takes a special type of mind to jump into this topic.
So right off the bat, very impressive.
The last year and a half we have been traveling the US meeting contractors and hosting podcasts with them to learn what motivated them and how they became the contractors they are today.
We been learning and documenting all the pain points industry wide and then created our own "blueprint" or challenge list we see everyone typically run into.
Thats the most recent, plus im a very calculated person, when I started contracting, I probably made every mistake in the book and then wrote it down on how I overcame it. What I should have done before the problem / after etc etc.
so all this data, I write up and created my own SOP for running our business. I created graphics and lessons for myself so I wouldn't repeat the same ones over.
The hardest part of the GPT is educating it, giving it, its own foundation.
if you dont you get the generic replies everyone sees today.
For example lots of GPT responses all have the same language, they also give the same grammar indicators.
We are rolling into a new phase of Ai where now you can see if someone else is using GPT already.
So if you dont train your own, you may be using Ai but you will not grow.
So the foundation, learning, and input...aka DATA is the most important, and hardest.
Plus if you have a bad foundation - then your responses will be wrong.
What do you think?
Absolutely agree, foundation is everything! You're right about the generic response problem. I've definitely noticed those AI grammar patterns (the em dashes, overly structured responses) showing up even when I think I've trained it properly.
What's been working for me is feeding it back the actual responses I ended up using after making changes. I'll tell it "FYI, I ended up sending this response instead" and paste the real version so it learns from what actually worked versus what it originally suggested.
Your approach of documenting mistakes and building SOPs around the solutions makes total sense. That real-world experience is way more valuable for training than just general business knowledge.
How are you handling the foundation-building? Are you feeding it your documented lessons gradually or building the entire knowledge base first?
The podcast research sounds like a solid way to gather industry-specific data. Having those real contractor pain points has to make the responses much more authentic than generic business advice.
When a topic comes up, ill talk about it with our team, make some content on it, then graphic, then ill test it online with the public see how they respond.
its like putting it through the fire and forging it to see any points where we have to tweak.
But whenever we have a solid graphic or update then I'll give it to the GPT.
I can't just rush it because then I may input some bad values and **bleep** it all up. On our end, with the PRO version, you can only use / upload 30 pieces of "core knowledge / foundation" so you need to be able to get your principles and standing within that limited response, like your guardrails,
and then make sure to give supplemental knowledge on other topics so it can weave it all together.
