Stop Using GenAI to Send More Spam
GenAI is turning sales upside-down—but not by cranking out thousands of templated emails that clog inboxes. The real unlock is using large-language models as a research engine to spot buyers who are already in pain and then earning the right to start a conversation by showing them you understand their problems.
To dig into the GenAI sales playbook, we sat down with Jordan Crawford, CEO & founder of Blueprint GTM. Jordan’s data science firm helps high-growth startups convert public data into revenue, and his lessons have become gospel for teams at Clay (valued at $3 billion), Tennr (valued at $600 million) and dozens more. He spends every Friday publicly building a complete go-to-market system in an hour live on Cannonball GTM, and he has a wealth of insights and success stories to share.
Below are Jordan’s four commandments for putting GenAI to work in outbound sales.
1. The List Is The Message
“Most startups are writing their way out of a targeting problem. You can’t write a good message to the wrong person.” - Jordan Crawford
Every struggling outbound program has the same blind spot: it starts with copy rather than context. SDRs pull a generic ICPs from ZoomInfo, fire up a cadence tool, and hope witty copy can overcome the fact that nobody on the list actually has the problem they solve.
Jordan flips the order:
- Interrogate your best customers. Who got value fastest? Who expanded? Who would be devastated if you disappeared?
- Translate that into observable data. What can you see from the outside—there’s a ton of bleeding data. Website updates, public permits, board meeting minutes notes for schools, Google review velocity and patterns, who installed Generac generators—there’s 10x more public data than you think and a lot of it correlates with the pain you solve.
- Pull a list that exactly matches those traits. Only then, Jordan says, should you bother writing.
When you do send an email, the content is the research. To illustrate this with an example, Jordan talks about a message he built for a company that sells rental-management software to construction firms whose cranes sit idle. He identified those firms via public permit (turns out the DOT requires you to file a permit every time you move your large equipment) and then build the data set before writing a single line of copy.
This was the email:
Subject: Your Liebherr LTM 11200-9.1 idle?
Dug through some permits recently, and I found your Liebherr LTM 11200-9.1 doesn't seem to be currently in use (couldn't find a Texas DOT permit)?
Maxwell Construction (owner, max, max@maxwellconstruction.com, 405-145-1461) just filed a building permit to install some large wind turbines (Enercon E-101) that likely need your crane for 3-6 days. That's about a $45k→$180k rental.
Want an intro to Maxwell?
Still Workings 9 to 5,
-Dolly Parton
2. Use GenAI to Surface Pain, Not to Spray More Emails
“The AI SDR is like bolting a robot to the top of a horse. The horse gets nowhere and it still shits along the way.” - Jordan Crawford
Point-and-shoot “AI sales assistants” claim they’ll write the perfect email for you. Jordan calls this nonsense: “Automation layered on a broken process just speeds up the garbage,” he says.
Instead, he wields GenAI as an investigative analyst:
- Scrape public footprints. “ChatGPT has the best deep‑research capability on the market,” says Jordan. “Feed it permits, SEC filings, GitHub commits—anything noisy humans hate sorting.”
- Ask the model to summarize patterns. LLMs excel at clustering noisy data into usable signals (e.g., “show companies adding ≥4 properties per quarter but scoring <90 on mobile page-speed”).
- Quantify impact. Insert revenue, utilization, or churn numbers so the prospect sees dollars, not buzzwords. “If you can’t attach a dollar sign, keep digging,” says Jordan.
- Package the finding as a consulting report. Jordan’s emails often read like mini-consulting reports the prospect would gladly pay to receive. Jordan calls them “Permissionless Value Props,” or PVPs. “The message should be so useful the prospect would pay just to read it,” says Jordan. “If that bar feels high, good. It means you’re on the right track.”
Because each email is anchored in verifiable pain, Jordan’s reply rates skyrocket, and the replies open with ‘How did you know this?’—the best segue a seller could want.
3. Validate Manually Before You Automate
“Don’t scale what you haven’t proven by hand.” - Jordan Crawford
GenAI can map a market in minutes, but Jordan warns founders not to delegate to GenAI too soon. He insists that you should close your first 10-30 deals manually first. Why?
- Language-market fit hones future messaging. When you talk to prospects one-on-one, you steal their exact wording for pain points; those phrases anchor future prompts and keep LLM copy from sounding robotic.
- Objection intel shapes the filter. Each manual call reveals a new criterion to bake into the list.
- Automation scales mistakes. “Spraying bad logic at 50,000 inboxes just gets you blocked 50,000 times faster,” says Jordan. You need to understand who you are selling to, why they care, and you need to spot patterns. You can only see these if you do it by hand.
Only after you can reliably win deals at low volume should you spin up Clay, Instantly or Smartlead and you can use tools like Serper or Claude Code to save cash. At that point, you’re industrializing a pattern you already understand.
This slower-is-faster mindset prevents you from building an ineffective GenAI sales workflow before any human has closed a deal.
4. Your First GTM Hire Probably Isn’t an SDR
“You don’t need someone to knock on more doors. You need someone who knows what doors to knock on.” - Jordan Crawford
Founders typically default to hiring a junior SDR once pipeline stalls. Jordan counters with a different archetype: the Go-To-Market Engineer (sometimes dubbed a “chief of systems”).
Traits to recruit for:
- Obsession with experimentation speed. How quickly can they run an idea from hunch to live test?
- Fluency in data tooling. Think SQL, Clay, Apify, Claude Code / APIs.
- LLM craftsmanship. Can they explain the difference between Claude, ChatGPT, and Gemini? Show favorite prompts? Articulate what the models can’t do?
- Discernment. Your hire must be able to separate out AI hallucinations at a glance. “The single skill that matters most in 2025 is discernment,” says Jordan.
Once this person proves a repeatable loop—data in, PVP out—then you should be able to scale the play out with a low number of people.
Conclusion: The Hidden Magic is Discipline
GenAI can now surface signals humans never had bandwidth to notice. But Jordan’s playbook makes one truth brutal: software amplifies genius and garabage alike. Skip the research, automate too early, or hire for volume instead of discernment, and AI will crash‑land your quota faster than any human could.
Start with the unsexy question no spammer bothers to ask—“Which 50 companies feel this pain today?”—and let GenAI prove the answer. Then email the evidence 1 by 1. When the prospect replies, you won’t sound like a stranger begging for 30 minutes; you’ll sound like the only vendor who actually did the homework.
And that, Jordan grins, is how you swap a robot‑on‑a‑horse for a rocket engine. As Jordan puts it, “Do the research. Eat your vegetables. Then step on the gas. The rest takes care of itself.”
Want more insights from Jordan? Attend his weekly Cannonball sessions, or take his course to work with him directly.
