3 Starter Tips for Your "AI" Stack
A new framework for custom GPTs
Thanks to everyone who participated on the custom gpt poll on the last post. More to come on this, updates below.
Over the past couple of weeks I’ve been lucky enough to connect with both founders and AE’s discussing numerous topics ranging from business development strategy to outreach tactics. Many of those discussions touched on leveraging AI in this process, be it strategic or tactical.
Much of what I’m going to say today bodes well with a recent episode of the Pomp podcast featuring perplexity CEO, Aravind Srinivas. If you have a minute, I’d highly suggest listening.
A few things we’ll cover today:
Judgement over Creativity
Offloading 80% of Workload
Practical Approaches to AI Agents + Custom GPT
Judgement over Creativity
The most interesting takeaway I’ve found from both sales people on the field, along with non-sales founders, was the question of: “Can AI replace sales?”.
People fell into three camps: one of genuine curiosity, staunch confidence that it will replace or that it surely won’t replace. You know I fall in the latter.
It’s hard to back this up with evidence however, primarily because we don’t have data and the field is developing quickly. Quickly being a vast understatement. Aravind, CEO of Perplexity put it best:
“We have the opinion that we are the most creative people on the planet, but it’s sort of what Rick Rubin says. You need to have incredibly good judgement and the actual process of creation, like knowing which tools to use, which instruments to use, which software to use, all that isn’t goin to be super valuable anymore. It’s more like, can you look at the end output and say like what’s actually going wrong there, and then give that as feedback to an AI or computer that you’re interacting with and can make good use of”
This makes a lot of sense.
Writing out campaigns is laborious and time consuming. Even more interesting is that humans struggle with a blank canvas. This workload is now outsourced to AI, 80-90% is now done leaving you the room to critique, refine, and deliver based on the nuances of what you, the individual, know based on your experience.
We are much better at JUDGING something and refining.
Think about how a sales pitch is the same words for a team. Notice how each sales person on that team says it a different way? We take our own individual lens on how to deliver content, what to say, and to whom.
This shows acumen in judgement rather than creativity.
Offloading 80% of the Workload
This of course begs the question that if 80-90% the workload can be replaced by AI, wouldn’t that threaten our jobs?
The better question is: Where does the value lie?
Ironically, the final 10-20% of any point in a sales cycle is the hardest.
Think about it:
Booking a meeting: easier to blast out an email and get a response, but how much harder is it to actually secure the meeting?
Qualifying questions at first glance are easy but now place them in the dynamic of a person who is closed off or doesn’t want to share that much information - how challenging is that?
You’ve received a verbal “yes” on a deal commitment. How many resources and how much time is dedicated to get someone to *actually* sign the paper and ensure someone is emotionally invested?
Practical Approaches to AI Agents + Custom GPT
I’m providing a glimpse of my AI “stack”, if you will. How I’m setting up workflows for things like outreach and discovery. This will most certainly produce a number of posts in the coming weeks.
Why?
Intially, I had been trying to create a massive workflow to replicate what I do with custom gpt’s. Candidly, I’ve been banging my head up against a wall.
A big thanks to Daniel Lemire at AI Mistakes, who helped me rethink my approach. Which is to take pieces of the workflow and THEN chain them together.
Here’s a quick example of what I’m working on which I’ll walk through more next week:
Persona Matrix: with challenges, emotions, associated with ICPs. I’ve written about this before here.
Messaging Matrix: combined with a persona matrix to templatize key messages based off trigger events or relevance signals to help me develop more pointed outbound cadences based on the research I do.
All this is a combination of outputs from GPT amd perplexity.
Now that I’m thinking about replicating “pieces” of a chain rather than an ENTIRE workflow, it can help my day to day, and yours. Imagine this would make for a lot more efficient CUSTOM gpts.
If you have any questions or want to see something specifically that AI can help with, feel free to shoot me a note: andrew@hackingsales.xyz or @askobylarz on X.
As always, thanks for reading and see you all next week.
-Andrew K