Yonathan Cohen's Recent LinkedIn Posts

Most salespeople don't know why they lose deals. This Granola recipe tells you. It runs across all your sales calls and shows: → The objections that keep coming up. → The exact moment prospects disengage. → The patterns you repeat in lost deals. → What's different in the calls you close. Use Granola for your next few sales calls. Then type / and run the recipe. One click, one monthly review, and your real blind spots are exposed. Access the recipe here: https://lnkd.in/dBv5R7Ny Repost this to help someone close more deals.

Most AI avatars die when you talk to them. I tested 6 this month for a client project. 5 loop the same idle animation while you speak. Then I tried Tavus Phoenix-4. When I talk to it: → It nods when I make a point → It shifts expression when I change tone   → It reacts before I finish my sentence Every pixel is generated in real-time. No pre-set triggers, no canned reactions. It reads your tone, your face, your intent. I'm now testing it as part of a lead gen workflow: → AI avatar handles first-touch outreach via video → Personalized to the prospect's company and role → If they engage → auto-booked meeting in the CRM We've been adapting to machines for years. This is a machine adapting to us. --- Create your own avatar: https://www.tavus.io/

Most AI agents break after 5 minutes. This one builds itself and runs 24/7: 1 → Go to Base44 and open Superagent. 2 → Describe what you want. 3 → Connect your tools (Gmail, Calendar, Slack,...) 4 → The agent builds itself. No code or setup. 5 → It starts running. 24/7. Without you. What mine does every day: → Checks my calendar and sends me a morning brief. → Researches prospects before every call. → Sends follow-up emails on my behalf. → Monitors competitors and alerts me on WhatsApp. Most people spend days setting up AI agents. Docker, SSH keys, API configs, tutorials. This took me 5 minutes and one sentence. Try it → https://lnkd.in/dY_wn6Up

Get fresh leads inside Google Sheets every day. Here's how to build it in 5 minutes: ☑ Set your ICP (industry, seniority, location,...) ☑ Every morning, new leads appear in your sheet. ☑ Review them. Mark the ones you want as "Enrich". ☑ One click Emails and phones filled automatically. You just need Google Sheets + FullEnrich. Step 1 → Open the template and paste your API key. Step 2 → Fill in your ICP filters (takes 2 minutes). Step 3 → Set the schedule. Close the tab. Every morning: fresh leads in your sheet. Review, select, enrich. Most people spend hours building prospect lists manually. This does it while you sleep. Comment "LEADS" to get the template.

Just saw this Reddit post about AI agents. Finally, someone telling the truth. WHAT THEY SELL YOU: "AI agents work 24/7 automatically!" WHAT REALLY HAPPENS: Your AI texts customers who left in 2012. Why? Your customer data is in 3 different broken Excel files. THE REAL PROBLEM: The AI works fine. Your 20-year-old computer system doesn't. Company still uses Windows XP. Took 3 months just to connect the AI to it. WHAT ACTUALLY WORKS: Company wanted to automate everything. Instead, built tool that only checks if forms are complete. That's it, saves 3 hours weekly and they're happy. Start with one tiny task, not your entire business. AI agents aren't smart employees. They're toddlers with database access. Everyone's pretending this is easy. It's not. But nobody admits it.

Send a personalized video call to every prospect. Here's how to build it in 5 minutes: 1 → Add a prospect's website in Google Sheets.  2 → AI reads their site to understand their business. 3 → Tavus creates a live conversation link for them. 4 → Send. Prospect clicks and talks to your AI. Teams using video outreach report 3x higher response rates. This is just the foundation. You can plug in: ☑ FullEnrich → Find & enrich the right contact. ☑ Lemlist → Send the link via email sequence. ☑ HubSpot → Log conversations & track pipeline. Most people send the same cold email to 100 prospects. This sends a live video conversation to each one. Workflow in the comment or ask for it.

Most AI agents fail before they finish. Here's the math nobody talks about: 1 step → 99% success rate. 5 steps → 77%. 10 steps → 60%. 20 steps → 36%. Every step you add is a new failure point. 4 agents coordinating to do what one prompt could. For a task that's basically "read this, extract that." The stack that actually works: ☑ Claude → one tight prompt with examples ☑ n8n → webhook or cron trigger ☑ Supabase → if persistence is needed That's it. Simple agents make money. Complex agents make problems.

Nobody Googles anymore, they ask AI. Here's how to check if AI recommends you: I built an Airtop agent that audits your brand across every major AI. ☑ It visits your website and gets your business. ☑ Asks 3 AIs about your industry. ☑ It checks if they mention you. ☑ It checks what sources they cite. ☑ It generates a full visibility report in a Google Doc. What you get: → Are AIs recommending you or your competitors? → What do they say about your brand? → Where are the gaps in your visibility? → A concrete action plan to fix it. Most businesses have no idea if AI mentions them. Here's how to try it: Step 1 → Open the agent https://lnkd.in/dhxWxY85 Step 2 → Paste your website URL. Step 3 → Add your keywords. Step 4 → Get your full audit report.

Turn viral TikTok content into infographics in 2 min. Here's how: 1 → Set your keywords. 2 → Apify finds the most viral TikToks for it. 3 → Pick the best one. 4 → Paste the ready-made prompt into Gamma. 5 → Gamma creates a professional infographic. Viral content already did the hard work. You just repackage it. ☑ No design skills needed. ☑ No hours scrolling TikTok. ☑ No starting from scratch. The best ideas are already going viral somewhere. This finds them and turns them into your next post. n8n template here:  https://tally.so/r/NpL15O

I got 29% reply rate on LinkedIn with this. Here's how to copy it in 2 min: 1 → Connect Gamma to Claude 2 → Create a Project in Claude "sales deck generator" 3 → Upload the file I shared to the project. 4 → Type /sales-deck and tell Claude what you sell. 5 → Paste a prospect's website URL. 6 → Gamma creates a personalized deck for them. That's it. Each deck is unique. About THEIR business. Selling YOUR services. Ready in 30 seconds. For each prospect just paste a new URL. Claude remembers your product and messaging. Most people send the same cold message to 100 prospects. This sends a custom deck to each one. Comment "GAMMA" to get the file.

Stop creating carousels from scratch. Turn Reddit's best posts into slides in 2 min. Most people stare at a blank page trying to find carousel ideas. The best ideas are already on Reddit.  Thousands of upvotes an proven engagement. This workflow turns them into ready-to-post carousels: Step 1 → Enter your topic (ex: "Claude Code"). Step 2 → It finds the most engaging Reddit posts. Step 3 → AI reframes them into carousel scripts. Step 4 → Gamma designs the slides. What you get: → Professional slides with perfect layouts. → Relevant visuals added automatically. → Export-ready for LinkedIn, Instagram,... → No design skills needed. Reddit validates what resonates. Gamma makes it beautiful. You just post it. Comment "SEND" to get the templat & step-by-step.

Find any prospect without leaving Slack. Here's how to build a lead research bot in 10 minutes: I built a bot that lives inside my Slack with FullEnrich ☑ I type: "@FullEnrich Bot find 10 sales at Shopify" ☑ It returns names, titles, locations, and profiles. ☑ Directly in the Slack thread. If you found someone interesting, get their details: "@FullEnrich Bot [paste their profile URL]" ☑ It pulls their verified contact info in seconds. ☑ All inside Slack. Most people jump between 4 tabs to find 1 prospect. I type one message in Slack. Your team now finds leads without leaving the conversation. Comment "SEND" to get the template. Here's how to build it:

Everyone creates content on LinkedIn. Nobody turns it into pipeline. I built a system that changes that. → Tracks who engages with your posts → Filters out AI comments and low-intent signals → Detects real patterns (multiple touches) → Enriches top engagers with verified contact data → Scores them against your ICP → Delivers warm leads to your CRM every morning Not every like or comment is a buying signal. (especially with AI) 3 real comments in 2 weeks → Intent. I tested it on my own content: 23 real engagers → 7 conversations → 2 clients. This is one workflow from my lead gen library. What makes it work: → Calibrating it to YOUR audience. --- DM "SPRINT" if you want this built for your business.

I build workflows everyday. My go-to tool for turning web data into leads: Bright Data Here's a workflow I run for myself: → I give it a website URL → It analyzes the site to understand their offer → AI builds the ideal customer profile automatically → Searches Crunchbase for matching companies → Scores each one on 5 signals: · Company size fit ✓ · Uses competitor tech ✓ · Web traffic growing ✓ · Recently funded ✓ · Right industry ✓ Result: 50+ scored leads from your website URL in minutes. I built a free tool so you can test the power yourself. Try it → https://lnkd.in/dmEXyDcB

Stop asking ChatGPT about your prospects. Here's how to close more deals with real data: ☑ Tavily is a search engine built for AI. ☑ It searches the web in real time, not stuck in 2024. ☑ Connect it to Claude. It searches the web for you. ☑ You get live data Most people paste a company name into ChatGPT and trust the output. That output is months old, sometimes wrong. I let Claude pull fresh data before every outreach. → Copy this prompt to try it: "Search for the latest news about [company name] from the last 7 days. Include funding, new hires, and product updates. Give me 3 talking points I can use in a cold email." That one prompt replaces 20 minutes of Google. Here's how to set it up (2 minutes):

I create automation for a living. But my own communication was fully manual. I type 100+ messages a day. Slack, email, WhatsApp, Notion. ~2 hours a day, all typed, all slow. So I automated it. I switched to Wispr Flow. I talk → it types. Any app in 500ms. Here's what makes it different from basic dictation: ☑ It reads your screen.  It knows the context before you reply. ☑ It adapts tone per app.  Slack stays casual. Email goes professional. ☑ It learns your words.  Client names, technical terms, abbreviations. But here's what sold me → Whisper mode. You can dictate so quietly that nobody hears a thing. Waiting for coffee ➤ whispered all my Slack replies. Walking to lunch ➤ answered 10 emails hands-free. Sitting in a meeting ➤ wrote tomorrow's to-do list. 85% of my messages need zero corrections. I automate workflows for a living. This is the one that saved me 2 hours a day. --- Now on Android too. Link in comments.

Stop writing follow-ups after sales calls. My prospect gets a personalized deck in 5 min. Step 1 → My call ends. Granola transcribes my notes. Step 2 → Zapier triggers automatically. Step 3 → Claude writes the deck content from my notes. Step 4 → Gamma builds a personalized deck. Step 5 → Email sends to the prospect with the deck. What the prospect gets: → What you discussed. → Key takeaways from the call. → Next steps. → A CTA. Granola captures the full conversation. Their words not a generic template. Most people finish a call and say "I'll send you more info." Then forge or send a generic PDF 3 days later. This workflow sends a personalized deck in 5 min. Guess who stays top of mind. Comment "GRANOLA" to get the template. Here's how to build it:

80% of AI projects fail. Because of this one thing: We implement before we understand. Company hears about AI → Rushes to implement → No clear problem → Project dies → "AI doesn't work" The mistakes: ⃠ Solution without a problem "We need AI" instead of "We waste 20 hours on X" ⃠ No success metrics Can't even define what "working" looks like. ⃠ Too big, too fast Revolutionize everything = achieve nothing. It's not the AI that failed. It's the approach. The ones that work all start the same way: One small problem, one clear metric, one workflow. That's it.

Book meetings from your website 24/7. Here's how to set up an AI sales rep in 10 minutes: Most websites have a chatbot. This is not a chatbot. It's Tavus a real-time AI video conversation. ☑ Visitor lands on your website. ☑ An AI replica of you greets them face-to-face. ☑ It talks, listens, and responds in real time. ☑ It qualifies the visitor. ☑ It books the meeting directly. No missed visitors or "we'll get back to you." Most companies lose 95% of website visitors. This AI talks to every single one. Here's how to set it up:

You can now create your own AI board of directors. Here's how (inspired by Andrej Karpathy): ☑ You ask one question. ☑ Multiple AIs search the web and answer. ☑ They rank and criticize each other's answers. ☑ A fifth AI picks the best answer with sources. Four AIs debating with real-time data. Best answer wins. So I built my own version with Bright Data to unlocks the full web for each AI. ☑ ChatGPT, Perplexity, Gemini, and Grok in mine. Here's how to try it: Step 1 → Open the link. Step 2 → Ask any question. Step 3 → Get the best answer from four AIs. Try it: https://lnkd.in/dzh7MZpK

Claude runs my entire sales process. Here are the 6 tools plugged into it: ☑ "Find this person's contact info" FullEnrich delivers. ☑ "Research their company" Bright Data searches the web. ☑ You hop on a call.  Granola transcribes everything. ☑ "Write a follow-up from my notes" Claude drafts it. ☑ "Create a personalized deck for them" Gamma builds it. ☑ "Send the email with the deck" Zapier sends it. 1 chat, 6 prompts → Deal moved forward. No tab switching or copy-pasting between tools. Just Claude connected to your entire stack. How to set it up: Step 1 → Open Claude → Settings → Connectors. Step 2 → Connect each tool (2 minutes each). Step 3 → Start prompting. Most people use Claude to summarize text. This is Claude running your entire sales process.

Claude Code is not just for coding anymore. Here are 5 MCPs that make it dangerous: ☑ Figma MCP  → Inspect designs and generate React code from frames. ☑ Zapier MCP  → Trigger any Zap from your terminal. Thousands of apps, one prompt. ☑ Notion MCP  → Read and update docs without leaving your editor. ☑ GitHub MCP  → Read diffs, suggest fixes, open PRs in natural language. ☑ Supabase MCP  → Design schemas, write migrations, query your database. With these MCPs, you describe the outcome. Claude handles the glue work: → "Convert this Figma frame into React components in my design system." → "Propose a bug fix, open a PR, and update the Notion doc." → "Query all users who signed up this week and push a report to Notion."

Claude Code just shipped Auto Dream. It fixes the memory problem nobody talks about. Auto Memory launched a month ago. By session 20, it becomes a mess of contradictions and stale context. Auto Dream was built to fix this! Think of it like REM sleep. Your brain doesn't store everything from the day. It replays, cleans, and consolidates overnight. Auto Dream does the same for Claude Code: → Scans memory files & session transcripts → Merges duplicates + resolves contradictions → Converts vague dates to real calendar dates → Keeps MEMORY.md clean (<200 lines) It runs in the background with zero interruptions. Check if you have it: /memory → look for "Auto-dream: on" Auto Memory writes everything down. Auto Dream knows what to forget.

This is how I book meetings on LinkedIn. (and it only takes 2 minutes per prospect) Step 1 → Connect Gamma inside Claude Step 2 → Drop a prospect + website URL into Claude Step 3 → Ask to build a Gamma deck about them Step 4 → Send the deck as a DM "Hey Jenny, made this for you → [lien Gamma]" Why this works: → They check your profile = curious → You send a personalized deck = surprised → The deck looks professional = impressed → Result = real conversation Most people send the same cold message to 100 people. You send a custom deck to 10. Guess who books more meetings. Gamma is now a native connector inside Claude.

You don't need n8n for most automations anymore. Now I do it in one prompt inside Claude. Most people still think Claude is a chatbot. But it’s becoming an action layer over your tools. The secret → MCPs. They connect your tools directly to LLMs. ☑ Your email ☑ Your calendar ☑ Your CRM ☑ Your browser ☑ Your files Once connected, you don't "chat" with Claude. You run workflows by talking to it. Most people build a 20-node workflow in n8n to send one personalized email. I type: "Research this company, check my calendar, and send them an email with my availability." One prompt but same outcome. n8n, Zapier, and Make still win for automations that run 24/7 without you. But for everything else, you're overcomplicating it.