3/17/26: From 600 Leads to 6 Deals: The AI Lead Gen Playbook
Hey friends as always, we are here for another GTM AI deep dive. I started this podcast 3 years ago with a simple goal: give founders, operators, and GTM pros an honest look at how AI actually works in practice. No hype, Just what’s real.
Since then my man Jonathan Moss showed up and helped me take it to the next level and now 4,000+ of you showed up. So we decided to go bigger.
We’ve restructured to give you more value, more consistently, across more formats.
Here’s what we’re now cooking up:
Tuesdays & Thursdays → GTM AI Pods 🎙️
Think of it like a cooking show, but for AI in business. Real practitioners. Real builds. We’re bringing in the people who are actually doing the work and having them show you exactly what they built, how they built it, and what they’d do differently. No theory.
Sundays → Under the Hood 🔧
This is where we go deep. Long-form articles covering how to actually use AI, how to build agents that work, and the underlying systems that make it all hold together. If Tuesday/Thursday is the cooking show, Sunday is the cookbook.
Regularly → Signal v. Noise 📡
AI is moving fast. Model drops. New capabilities. Things that matter, things that don’t. This is your curated filter... I’m sifting through the noise so you know what actually deserves your attention.
Always growing → AI Business Network + AI Education for Professionals 🏗️
This is the layer I’m most excited about. Step-by-step guides. Video learning content. A community of peers who are figuring this out alongside you. Time with Coach K and me. Virtual events, live learning sessions, and more on the way.
The mission underneath all of it hasn’t changed: help you stay connected to what’s changing in AI, learn how to actually use it, build your strategy, implement it and ultimately become a 100x version of yourself professionally.
If you want in, you can go to the main page and get hooked up or subscribe below
Now without further ado, a few weeks ago on Linkedin I posted about a team I was working with that was getting leads like crazy on Linkedin, so I had them join the podcast to deep dive into how they are doing that!
PS Under the Podcast is a bunch of prompts and skill.md descriptions you can use in your own AI account :)
You can go to Youtube, Apple, Spotify as well as a whole other host of locations to hear the podcast or see the video interview.
How to use AI to Find and Convert High-intent leads on Linkedin
Most outreach tools do one thing: blast messages. Gojiberry AI does something different. It uses Claude to find leads, score them by intent, write personalized messages, and rank them so the warmest contacts get hit first. One of the founders, Roman Czerny, walked me through the full system on the GTM AI Podcast. His numbers are hard to argue with: 50% of replies convert to blueprint requests, 70% of demos close, and 35% of free trials become paying customers. Here’s what stood out.
1) AI-scored intent signals beat job title targeting by a mile.
Traditional outreach filters by job title + location. That’s a starting point, not a strategy. Gojiberry layers in real-time behavioral signals: who just followed your company page, who interacted with a competitor’s post, who changed jobs this week, who’s been active on LinkedIn in the past 48 hours.
Claude analyzes hundreds of potential leads and filters down to the handful that actually match your ICP and show buying behavior right now. Roman’s team filtered 600 people to surface 6 qualified leads in one example. That’s a 99% noise reduction before a single message goes out.
The shift: stop targeting profiles. Start targeting behavior.
2) The “blueprint before demo” funnel flips traditional outreach on its head.
Roman doesn’t ask for demos in outreach messages. He offers a free blueprint. A detailed, valuable resource showing the prospect exactly how to solve the problem Gojiberry addresses. When someone says yes and reads the blueprint, most don’t need a demo. They sign up for a free trial on their own. The ones who do request a demo are so pre-educated that Roman closes 70% of them.
Why this works:
You’re asking to give value, not take time. Response rates skyrocket.
No-show rates drop because the prospect initiated the meeting.
The prospect pre-qualifies themselves by engaging with the content.
Your lead magnet does the selling. The demo just confirms.
The proof: Roman’s best lead magnet went viral. Prospects started sharing it with other prospects. That’s when outbound becomes inbound.
3) One person runs 7 LinkedIn accounts, 3 X accounts, YouTube, and Reddit. AI is the multiplier.
Roman is the entire marketing department. He manages content across seven LinkedIn accounts posting daily, three X accounts, a YouTube channel, and Reddit. His system: Gemini with persistent memory handles copywriting. Whisper handles voice-to-text so he talks ideas instead of typing them. Nano Banana Pro on Gemini generates thumbnails.
For YouTube, Roman films one-take competitor review videos. Title format: “[Competitor Name] Review 2026.” Half the video covers the competitor, then he pivots to showing Gojiberry as an alternative. These rank fast for high-intent search terms. Five minutes per video. One per day.
The hack he dropped: create a dedicated YouTube channel named after the competitor review keyword. A channel called “ZoomInfo Review” with a video titled “ZoomInfo Review 2026” ranks above almost everything. Aggressive, effective, and costs nothing.
4) AI end-to-end still loses to human-AI handoff. For now.
Roman tested fully automated outreach where AI handles every step including responding to replies. The result: way fewer demos. When a human takes over at the reply stage, demo bookings jump. Prospects are testing for AI with trick questions (”give me a cake recipe”) and the human touch at the conversion point still matters.
The tactical split that works right now:
AI handles lead discovery, scoring, ranking, and initial message
Human handles replies and demo booking
AI assists the human with context and suggested responses
This changes fast. But today, the hybrid model wins.
The bottom line:
The single biggest ROI move from this episode: stop reaching out to people who haven’t been on LinkedIn in two months. Target people active in the last 24-48 hours. Roman says this alone doubles response rates across every campaign they’ve tested.
The AI Lead Gen Playbook: How One Founder Books 20+ Demos Per Week With a Team of One
Roman is one person running outbound across LinkedIn, cold email, YouTube, X, and Reddit. He books 20+ demos per week. His close rate on those demos is 70%. His trial-to-paid conversion is 35%.
Those numbers would be impressive for a 10-person sales team. He does it solo. AI is doing the heavy lifting at every step except one (more on that later).
I’m breaking down the full system so you can steal the playbook.
1. Stop Targeting Profiles. Start Targeting Behavior.
Every outreach tool on the market starts with the same filters: job title, company size, location. That gets you a list. It doesn’t get you responses.
Roman’s system adds a layer most teams skip entirely: real-time behavioral signals. His AI (Claude, running in Gojiberry’s backend) tracks who just followed his company page, who interacted with a competitor’s post, who changed jobs recently, and who’s been active on LinkedIn in the past 48 hours.
Then it scores them. Not a basic “fit score” from static data. A combined score using ICP match (40% weight) and intent signal strength (60% weight). Intent gets more weight because a good-fit lead who hasn’t logged in for two months is worth less than a decent-fit lead who was active yesterday.
Here’s the math from one of Roman’s campaigns: the AI analyzed 600 leads. It surfaced 6 qualified prospects. That’s 99% noise filtered out before a single message was sent.
The move you can steal today (no tools required): Before sending any outreach, check if the person has been active on LinkedIn in the past 24-48 hours. Look at their recent posts, comments, or shares. If there’s nothing recent, skip them. Roman says this alone doubles response rates across every campaign they’ve tested.
How to prompt AI for lead scoring:
I'm going to give you a list of leads with their LinkedIn profile data
(job title, company, industry, company size, recent activity).
My ideal customer profile:
- [Role/title]: VP Sales, Head of Revenue, CRO
- [Company size]: 50-500 employees
- [Industry]: B2B SaaS
- [Other criteria]: Series A-C funded
Score each lead 1-10 on ICP fit and 1-10 on intent signal strength.
Weight intent at 60% and ICP at 40%.
Sort by combined score descending.
Flag anyone scoring 8+ as "reach out today."
2. The Blueprint Funnel: Why “Want This Free Guide?” Beats “Book a Demo?” by 10x
Roman doesn’t ask for demos in his outreach messages. He tried it. Everyone tries it. The response rates are garbage because every sales rep on LinkedIn is doing the same thing. People are exhausted by Calendly links from strangers.
Instead, he offers a blueprint. A detailed, free resource that walks prospects through exactly how to solve the problem his product addresses. The outreach message looks something like this:
“Hey [name], noticed [specific signal]. We put together a blueprint on [topic that matches their signal]. Want me to send it over?”
That’s it. No pitch. No “15 minutes of your time.” Just an offer to send something valuable.
The results from Roman’s funnel:
50% of people who reply say yes to receiving the blueprint
90% of conversions happen without a demo. People read the blueprint, visit the website, and start a free trial on their own.
When a demo does happen, 70% close. The prospect already understands the product. The demo just confirms.
No-show rates are near zero because the prospect initiated the meeting.
Trial-to-paid: 35%. The self-serve path filters for people who actually want to use the product.
Compare that to traditional outbound: 3-8% reply rate, 20-30% close rate on demos, 30-40% no-show rate.
The breakthrough moment: Roman’s best lead magnet went viral. Prospects started sharing it with other prospects. When that happens, your outbound becomes inbound and you didn’t spend a dollar on ads.
He told me a story about a company called Instantly. They sent him a blueprint so good he forwarded it to 10-15 people. He’s been a customer for 3-4 years now and has spent $30,000+ with them. That’s the ROI of one great lead magnet.
What makes a blueprint worth sharing:
Most lead magnets are glorified product brochures. Nobody shares those. The ones that go viral solve a complete problem end-to-end without requiring you to buy anything. They’re specific enough that someone can execute the entire workflow after reading.
The test: would the reader forward this to a colleague before they even finish reading? If not, add more tactical depth until the answer is yes.
3. AI Writes Every Message. But a Human Closes Every Deal.
Roman tested full automation. AI finding leads, AI scoring them, AI writing messages, AI responding to replies. End-to-end, no human touch.
It booked way fewer demos.
The reason is simple: prospects are testing for AI. People respond to outreach with “tell me a cake recipe” or “are you an AI?” to see if they’re talking to a bot. When AI handles the reply, it either fails the test or gives itself away. When a human takes over at that point, demo bookings jump.
The winning split right now:
AI does: lead discovery, scoring, ranking, initial message writing, content generation
Human does: reply handling, demo booking, relationship building
The initial message can be AI-written because it’s one-directional. Nobody interrogates a first message. But the moment a prospect engages, they want to know there’s a person on the other end.
Here’s the nuance Roman shared: his AI-generated messages perform better than templates because they reference specific behavioral signals. “Hey Sarah, noticed your team just started following [competitor]. We put together a blueprint on [relevant topic].” That’s not a merge field. That’s Claude analyzing the lead’s actual behavior and writing a unique message. The prospect reads it and thinks a human wrote it because it’s contextually relevant.
The AI personalization prompt:
Write a personalized LinkedIn message for this lead:
- Name: [name]
- Title: [title] at [company]
- Signal: [what they did, e.g. "commented on a post about AI outreach tools"]
- My offer: [one sentence about what your blueprint covers]
Rules:
- Reference their specific signal naturally
- Offer the blueprint, don't ask for a demo
- Under 300 characters for connection request
- Conversational, not salesy
- End with: "Want me to send it over?"
4. One Person, Seven LinkedIn Accounts, Five Channels. AI is the Multiplier.
Roman is the entire marketing department at Gojiberry. Here’s what he runs daily:
7 LinkedIn accounts posting content daily (his own plus team members)
3 X accounts doing content + lead magnets
1 YouTube channel focused on competitor review videos for SEO
Reddit with consistent community engagement
Cold email via Instantly repurposing leads from Gojiberry
His single metric: “How many blueprints did I distribute today?”
Every channel feeds the same funnel. LinkedIn posts drive blueprint downloads. X posts drive blueprint downloads. YouTube videos drive blueprint downloads. Cold email offers blueprints. The blueprint does the selling.
For content creation, Roman uses Gemini with persistent memory. He’s loaded it with his past content and brand voice so every new idea he feeds in comes back as a draft that only needs minor editing. He also uses Whisper for voice-to-text. Click a button, talk through the idea, let Gemini clean it up into a post. The whole content creation loop takes minutes.
The YouTube SEO hack that costs nothing:
Roman films one-take competitor review videos. No editing. Title format: “[Competitor Name] Review 2026.” The first half honestly reviews the competitor. The second half pivots: “If [weakness] is a dealbreaker, here’s an alternative” and demos Gojiberry.
These videos rank for high-intent search terms. Someone typing “ZoomInfo Review 2026” is actively evaluating tools. They’re the warmest lead you could ask for.
The advanced move: create a dedicated YouTube channel named after the competitor keyword. A channel called “ZoomInfo Review” with a video titled “ZoomInfo Review 2026” ranks first almost every time. You’d need a separate channel per competitor, but each one takes 10 minutes to set up and the video takes 5 minutes to record.
5. The System in One Page
Here’s the full playbook distilled:
Find leads (AI):
Configure intent signals: page follows, post interactions, competitor engagement, job changes, keyword activity
AI scores each lead: ICP fit (40%) + intent signal strength (60%)
Filter to top 1% of qualified, active leads
Write messages (AI):
AI generates personalized message per lead using their specific behavioral signal
Message offers a blueprint/resource, not a demo
Warmest leads get contacted first
Distribute (AI-assisted):
Send via LinkedIn and cold email simultaneously
Post content across LinkedIn, X, YouTube, Reddit pointing to the blueprint
YouTube competitor reviews capture high-intent search traffic
Convert (Human):
Human handles all replies
Blueprint readers self-qualify: most convert without a demo
When demos happen, close rate is 60-70% because prospects are pre-educated
Measure (One metric):
Blueprints distributed per day. That’s the leading indicator. Everything downstream follows.
Skill.md for you to use:
---
name: ai-lead-scoring
description: >
AI-powered lead scoring, qualification, and ranking system. Use this skill whenever Coach asks
to score leads, qualify prospects, rank a lead list, build a scoring model, prioritize outreach
targets, or analyze intent signals. Also trigger when someone mentions ICP scoring, lead
prioritization, warm leads, intent signals, behavioral scoring, prospect ranking, or any
workflow that involves deciding which leads to contact first. This skill applies a two-layer
model (ICP fit + intent signal strength) with configurable weights and outputs ranked,
actionable lead lists. Works for manual processes and AI-automated pipelines.
---
# AI Lead Scoring
## Purpose
Replace gut-feel lead qualification with a repeatable, AI-powered scoring system that ranks prospects by both who they are (ICP fit) and what they're doing right now (intent signals). The core insight from practitioner data: intent signal strength predicts response rates better than demographic fit alone. Weight accordingly.
## The Two-Layer Scoring Model
### Layer 1: ICP Fit (Default Weight: 40%)
Static profile match. Answers: "Would this person be a good customer if they bought?"
Scoring dimensions (each contributes to a 1-10 composite):
- Job title/seniority alignment to buyer persona
- Company size within target range
- Industry match
- Geography relevance
- Funding stage or revenue range
- Tech stack indicators (when visible)
When building ICP criteria for Coach, always ask:
1. What does your current best customer look like? (reverse-engineer from closed-won)
2. What disqualifies someone immediately? (negative signals save time)
3. Are there "nice to have" signals that push a 6 to an 8?
### Layer 2: Intent Signal Strength (Default Weight: 60%)
Behavioral signals showing buying mode right now. Answers: "Is this person likely to respond today?"
Signal hierarchy (ordered by predictive strength):
- **+3 points:** Active on LinkedIn/platform in past 48 hours (baseline qualifier)
- **+2 points:** Engaged with content about your product category
- **+2 points:** Interacted with a competitor's page, post, or content
- **+2 points:** Posted about a problem your product solves
- **+1 point:** Changed jobs in past 90 days
- **+1 point:** Company raised funding recently
- **+1 point:** Visited your profile or company page
Intent gets 60% weight because timing beats targeting. A perfect ICP match who's been inactive for 2 months is worth less than a decent ICP match who commented on a competitor's post this morning.
### Combined Score
```
COMBINED = (ICP Fit × 0.4) + (Intent Signal × 0.6)
```
Weights are defaults. Adjust based on Coach's context:
- High-ticket / long sales cycle → increase ICP weight to 50%
- High-volume / fast close → increase Intent weight to 70%
- New market entry → increase ICP weight (you're still learning who buys)
## The 48-Hour Activity Filter
Before running the full scoring model, apply this pre-filter: remove anyone who hasn't been active on the target platform in the past 48 hours. Move them to a "check again next week" list, don't delete them.
Practitioner data: targeting people active within 48 hours doubles response rates compared to outreach based on profile data alone. The dropoff after 48 hours is steep. This is the single highest-ROI change to any outreach campaign.
## Signal Stacking
Multiple signals compound. Use these tiers to set urgency:
| Tier | Signals Present | Action |
|------|----------------|--------|
| Warm | Active 48h + ICP match | Reach out this week |
| Hot | Active 48h + competitor engagement + ICP match | Reach out today |
| On Fire | Active 48h + competitor + job change + ICP match | Drop everything, reach out now |
## Action Tiers From Scores
| Score Range | Category | Action |
|-------------|----------|--------|
| 8-10 | Priority | Reach out today. Personalized message referencing strongest signal. Human attention first. |
| 5-7 | Nurture | Add to outreach sequence. Re-check signals weekly. Upgrade if signals strengthen. |
| 1-4 | Skip | Don't waste a message. Revisit in 30 days if ICP fit exists but intent is missing. |
Goal: never send a message to someone scoring below 5.
## Ready-to-Use Scoring Prompt
When Coach needs to score a lead list, use this prompt (adapt the ICP section to their context):
```
I'm going to give you a list of leads. For each lead, I'll provide
their LinkedIn data and recent activity signals.
My ideal customer profile:
- Role: [fill in]
- Company size: [fill in]
- Industry: [fill in]
- Other criteria: [fill in]
Score each lead on two dimensions:
1. ICP FIT (1-10): How closely do they match the profile above?
2. INTENT SIGNAL (1-10): How active and relevant is their recent
behavior? Weight these signals:
- Active on platform in past 48 hours = +3
- Engaged with content about [product category] = +2
- Interacted with [competitor names] = +2
- Posted about [problem you solve] = +2
- Changed jobs in past 90 days = +1
- Company raised funding recently = +1
COMBINED SCORE = (ICP Fit × 0.4) + (Intent Signal × 0.6)
Output format for each lead:
Name | Company | Title | ICP: X/10 | Intent: X/10 | Combined: X/10
Primary signal: [the strongest intent signal detected]
Recommended action: [reach out today / add to nurture / skip]
Best outreach angle: [one sentence referencing their signal]
Sort by combined score, highest first. Flag anyone 8+ as priority.
```
## Benchmarks
| Metric | Without AI Scoring | With AI Scoring |
|--------|-------------------|-----------------|
| Lead qualification accuracy | 30-40% | 70-85% |
| Time to qualify 100 leads | 4-6 hours manual | 12 minutes |
| Reply rate on outreach | 5-8% | 15-25% |
| Noise-to-signal ratio | ~70% wasted messages | ~1% (600 leads → 6 qualified) |
Source: Gojiberry AI practitioner data, GTM AI Podcast Feb 2026.
## Integration With Other Skills
- After scoring → use `lead-magnet-ai` to create blueprints for the outreach offer
- After scoring → use `blueprint-funnel` to structure the value-first outreach sequence
- For Momentum deals → apply scoring to deal-specific prospect lists in `Momentum/deals/`
- For content about scoring → use `gtm-linkedin-post` or `gtm-ai-newsletter`
## Reference Files
- `references/scoring-prompt-variants.md` — Alternative prompts for different sales motions (enterprise, PLG, channel)
- `references/signal-definitions.md` — Detailed definitions and data sources for each intent signal typeThe Fastest Win From This Playbook
You don’t need any new tools to start. Today, before your next outreach batch, do this:
Check each prospect’s LinkedIn. Have they posted, commented, or liked something in the last 48 hours? If not, remove them from the batch.
For the ones who are active, look at what they engaged with. Reference it in your message.
Offer something useful instead of asking for time. A relevant article, a short guide, a template. Anything that shows you can help before you ask for anything.
That’s the core of Roman’s system stripped down to zero tools and zero cost. The AI and automation amplify it. But the principle works manually: target active people, lead with value, let them come to you.
This playbook came from my conversation with Roman Czerny on the GTM AI Podcast. If you want the full breakdown every week (AI tools, GTM strategies, live demos, and the stuff I can’t fit into a podcast), subscribe to this newsletter. It’s free, it’s tactical, and I don’t waste your time.