Do 0-1 B2B GTM like Rob Snyder
An agent skill for Claude or ChatGPT. Written by me, Rob — here's the whole story.
Tell me if this sounds like you:
You're trying to figure out B2B sales/GTM for your early-stage startup. You've stumbled across my book, blog, or podcast, and the path to fast growth finally seems straightforward.
But you're trying to implement my advice, and you're finding it difficult to apply to your specific situation. Claude gave you terrible feedback on your outbound messaging. Your homemade “AI Rob” skill is somewhat better, but you're pretty sure that it's nowhere near “Real Rob.”
That's why I created the official “AI Rob” skill. It helps you do early-stage GTM really well. With it, you'll do outbound, run sales calls, and analyze where the pull is like I do.
But AI Rob does way more than that:
- It can connect to your company brain and autonomously analyze your metrics weekly to tell you where your bottleneck is.
- It can help you problem-solve a weird deal you're stuck on.
- It can craft your landing page.
- It can structure a pilot.
- It can give you a way to think about which possible ICP actually has pull.
- …and it can do a lot more.
I am constantly refining AI Rob so that it is as close to “Real Rob” as possible. This includes writing most of the files by hand, and testing it against hundreds of founders' situations, calls, and metrics.
AI Rob is free if you work with real Rob 1:1 or join my GTM accelerator. Otherwise, it's $99/year (for now). Start with one month free.
How to actually use AI Rob
Invoke it in any chat or session by tagging /ai-rob.
For example: /ai-rob here's my outbound message and stats; roast it please!
You can also add it to regular workflows: Every time there's a new sales call in the company brain, have /ai-rob review it.
Founder stories
“I spent the last six months trying to do outbound unsuccessfully. We'd get maybe five meetings per month via outbound. Everything else about our funnel works, customers love our product, but we just couldn't figure out top-of-funnel. I installed AI Rob and gave it my outbound message for feedback. AI Rob tore it apart and gave me a new message. I wasn't sure if AI Rob's message would work, but I gave it a try… and in two days, I scheduled six meetings via outbound, which was more than I'd scheduled in my best outbound month before AI Rob!”
“We got to $5M in revenue but it was all me pushing. We couldn't hire sales reps, we struggled with retention, and it wasn't clear why. After reading Rob's book, it makes sense that I've been pushing and need to find PULL. So I started using AI Rob's call reviews to help me think differently about my sales calls and ICP. After every call, I get an AI Rob review and chat with AI Rob to help me improve how I sell, but more importantly help me figure out what PULL really is.”
“I'm working with real Rob, but I also use AI Rob all the time. I can review every sales call and deal with AI Rob, and also ask it my 2am existential questions I don't want to bother real Rob with. I use real Rob more strategically: Rather than having him review every call, I can have him deep dive into a few specific deals and then zoom out and help me with the bigger questions.”
Why isn't a homemade AI Rob skill good enough?
This is something I constantly hear when people upgrade from a homemade AI Rob skill to the official AI Rob: Official AI Rob is MUCH better, why? I'm not sure I have a good answer, but my current hunch is that experts develop extremely strict reasoning hierarchies: We know that, in Z domain, X is what is most important, and it is upstream of Y.
This is the exact opposite of how AI seems to work, which is to select any one plausible reasoning pattern out of infinite plausible reasoning patterns.
To make this concrete, let's think through outbound messaging. If you ask AI to help you write an outbound message, it can follow many reasonable reasoning patterns:
- This is an outbound message, so we need to pitch the pain point
- This is an outbound message from a founder who isn't selling anything, so we need to ask for feedback and say we're not selling anything
These — and the infinite other plausible reasoning patterns — will seem right when you analyze them. They will produce an outbound message that looks like an outbound message, and seems like it could work.
And the resulting outbound message won't work by default, because there are very few “right” reasoning patterns among a set of infinite plausible-seeming reasoning patterns. AI is looking for a needle in an infinite haystack.
So: Unless you enforce insanely strict reasoning rules, AI will, by default, “escape” into the infinite haystack and produce something that won't work.
Which is why just feeding in all of my blogs into your own AI Rob (or why feeding everything I've ever written into Delphi) will generate things I wouldn't say — it has a lot of content, but lacks the strict reasoning rules.
And so: I spend a lot of time thinking about these strict reasoning hierarchies and hard-coding logic. Yay.
Try AI Rob for a month, free
$99/year after that (for now). Free if you work with real Rob 1:1 or join the GTM accelerator.