This is a 2 part series that will teach you 95% of what you need to be successful with AI
(the other 5% is you, dear reader–but you’re good for it, right?)
PART 1: 3 terrible myths that every GTM leader falls for (easy to fix)
PART 2: 4 simple steps to use just ChatGPT to find evidence of why specific customers bought and turn that into an engine to find new customers.
Talking to our new AI overlords is basically like talking to aliens: They are immensely powerful, and also profoundly stupid. The way we use these tools today mostly just amplifies that stupidity.
Read the $46M dollar message below.
Does this feel like the same moment you experienced the true power of ChatGPT?
Or does it feel like eating a 34-day-old banana? (Ignore the fact that a robot just asked me to meet in person.)
What $46.1M in funding for AI promises (left) v.s. What it delivers in 2025 (right)
Salesforce may have invested billions in Agentforce, but their emails aren’t any better.
Anyone can hate—I mean, look at every single post on X—but not many people can drink that Haterade and turn it into a winning marathon run. That’s me, baby! Advisor to Clay since 2020 ($3B) and Tenner ($605M), OpenAI just sent me a trophy for running 10B tokens, and I’m the co-founder of a top 50 Business Substack (Cannonball, #46 as of today).
By the time you finish this immensely long article, you should be able to create a campaign you’d be proud to send to thousands of prospects with no other tools than ChatGPT and Claude Code. Or, screw it, copy and paste this mofo directly into ChatGPT and say, “I refuse to think; run this process for my company.”
Your message won’t sound anything like that crap above. Buckle up.
And when you see this emoji 🟢, it means actionable tip. Prompts are highlighted in orange.
The Top 3 Reasons Why Your AI Strategy is Failing – And What to Fix First
To help you think right, it’s important to dispel the most common myths, and also realize that we need a totally new mental model for AI. Because there are some places where AI is a god-like entity, and some places where it’s a garbage factory.
1. TRAP: You’re Horizontal SaaS & You Let Your SDRs Write Their Own Emails
Which Location Do You Believe Has Better Sushi?
Problem:
Horizontal SaaS companies fall into the thinking trap that adding verticals + features + personas + products = more revenue.
But the more surface area you cover, the greater the understanding gap between your solution and the customer’s reality, making it near impossible to scale your go-to-market with messages you’re proud to send.
That gap leads you to just blindly claim you have all of the world’s best food and you serve it at your buffet. Meanwhile your competitor is like Jiro, who trains people to make only rice for two straight years.
In order to go to market, you need to focus on one dish, and talk about why you’re the best at that dish. And it’s hard (this is why I only work with vertical SaaS companies).
Even the Buffet at the Bellagio only gets 3.3 stars on Yelp (and it’s DELICIOUS).
Solution:
You need to think about buyer segments, the insights you know about them, and how to quickly run campaigns against those insights—not about TAMs (more on this trap later).
The value of AI is that you can launch campaigns fast that you will be proud of, but only by focusing on a specific segment and extracting the insight of that segment.
The best campaigns are a combination of the segments where your customer gets 10x the value from your product AND has a problem that is extremely demonstrative with data (from the web or inside of your company).
A demonstrative problem is one that can easily be found with public or internal data, and also easily described.
Demonstrative
Good: Your 15 engineers qualify you for $5,000 in tax rebates per engineer
Good: Your 5 BR 710 airplanes are required to do X; we do it for $1k, 90% cheaper than most
Bad: Save 5 hours per engineer per month
Bad: Get off Excel, so you never have to worry about formulas breaking
A segment is not just industry + persona + product (although those things are often correlated). It’s really fenced around what data + message would everyone on the list recognize themselves in while they’re reading.
Here’s how to do it…
🟢 Segment By Pain, Not Product
To solve this problem gracefully, you need to narrow your focus on a segment.
Do:
1. Find the customers that get 10x value
2. Get each of their recent sales transcripts
3. Ask ChatGPT “can you find any public evidence that this customer was in this situation based on what they have said”
^---- Do this 1 by 1 with each customer with ChatGPT Thinking + Heavy Thinking on and then build your data strategy by combining.
2. Your Leadership Doesn’t Prospect or Vibe Code
You Can’t Lead What You Don’t Know
Problem:
“With direct LLM access, about 15% of all U.S. worker tasks can be done significantly faster at the same quality; with LLM-powered tools, that rises to 47–56% of tasks (2023).” -OpenAI Study (2023)
AI replaces tasks, not people. If you’re a GTM leader and don’t use ChatGPT for like 30%+ of your day, or you don’t vibe code, you have no idea what AI does well, what it does poorly, and how to help your team offload their tasks to AI.
🟢Solution:
Since AI is replacing tasks, you can’t very well help your team eliminate tasks if you don’t know what they are doing each day.
With your team:
Shadow an SDR and watch what they do all day
Write 10 cold emails to just closed prospects after listening to the transcripts
Cold call prospects yourself
Talk to customers
Build an MCP server to connect to your internal data (no code needed, promise)
With yourself:
Install Claude Code now (you don’t need to know anything about code)
Ask ChatGPT “don’t execute this just improve the quality of the prompt”
Ask ChatGPT “don’t execute what I asked for JUST ask me more questions.”
Ask Gemini “Here’s a screen recording of 20 minutes of my work, what could I do better with AI?” ← Gemini is the only model that can “watch” videos.
Get Wispr Flow and only ever talk to AI, not type, speak in 2-4 paragraphs at a time.
Ask ChatGPT “given all of our history, tell me all the best features of ChatGPT I have never used and how I should use them. Like Tasks. Or Pulse.”
3. You Use ZoomInfo to Build ICPs and Personas, and You Don’t Know Your Customer’s Secrets
Find Your Customer’s Pain, Then Build Your Own Data
Henry David Thoreau wrote in Walden (1854) that "men have become the tools of their tools."
Problem:
In the past, we only had one type of tool to target our customers (Apollo, ZoomInfo, etc). Then, nearly overnight, we were granted a tool that can find data, connect data, translate data, and crunch data all in minutes—and yet still we use it to write “personalized messages.”
Our biggest challenge is that our capabilities changed so drastically, and we didn’t know how to use them.
Now, we need to answer this one terribly difficult question:
With unlimited time and unlimited resources and perfect knowledge of my customer’s internal state, what message would I send them?
We can’t answer this question, because for so long, all we had was headcount and revenue filters, so none of our messages at scale were ever that different.
For Horizontal SaaS, what you’d say differs WILDLY by customer. So you need to segment by who the best message applies to, and who it does not.
🟢 Solution:
Stop building your list with ZoomInfo filters, and stop trying to sort or enrich an entire TAM. It’s just too complex and unnecessary.
Your customers, not ZoomInfo, know why they bought, what they were doing before they bought, and where they are now.
Inhabit Your Buyers’ Situations to Learn Their Secrets
Go to your customers and ask:
Why did you buy us v.s. continue to do what you were before?
What hard earned insight did you learn (that relates to our product)?
What do you believe we do much better than our competitors?
What was the one reason you decided to buy?
If you had to price your value, how would you do it?
What was happening in your role and in the company that led to this purchase
Who else was involved in the purchase and why did they buy?
You can’t know what data to find, or proxies for that data, without deep customer insight. Drop these transcripts into ChatGPT and say “find anything for blahblah.com on the public web that would have indicated this was going on inside of the company around”
SIDE NOTE: Even With Unlimited Time on the Web, Some Problems Can’t Be Detected
Some Problems Aren’t Discoverable Publicly (Like if Sally in Accounting is Using Excel)
Problem:
Some companies solve problems that happen mostly with 0 public data footprint (like accounting). My methodology is not designed for those companies.
I will say that you’d be surprised how much data is out there (e.g. certain companies have to publish a LOT of their financial information in some markets), so don’t give up too early, as there are often proxies for this information.
If there truly is no data from LinkedIn profile content, regulations, filings, posts online…there are other GTM motions to solve problems where great precise targeting is damn near impossible.
To be continued…
Next week: The Exact Steps to Launch a Campaign in 2 Hours.
Fun fact: Illustrations are hand drawn from creatives at the non profit Big Wonder. Want to help counteract the teen mental health crisis by helping young people find inspiration? Donate here
This Week Across Topline
Sam’s Corner
Something I am thinking about a lot these days…
In the modern world, content is a commodity. Even the brilliance of this newsletter. Anyone can make content. AI can make content. Ideas are everywhere—but who knows if you can trust them?
Two people can share the same piece of information, but if I know and trust one of them, it’s valuable. If I don’t, it’s noise. The difference is context.
For example: There are ten million “experts” on LinkedIn who haven’t done the job in years—and who may or may not be full of beans. Without context, there’s no way to know.
Interestingly, the people who’ve actually done it don’t need to shout—they just show up and share what’s real. As Lil Wayne said, real Gs move in silence like lasagna.
At Pavilion I think this means we need to nail putting the right people in the right rooms. Real GTM operators who know what they are talking about. Not trying to get the maximum number of members.
In your business, you should also be wondering what changes if content is a commodity and context is the resource that matters.
Become a Topline insider by joining our Slack channel.
We want to hear your feedback! Let us know your thoughts.
Sam’s Corner written by Sam Jacobs.