How to Pitch Partners With AI: The Reverse Pitch

Every founder’s inbox is a graveyard of cold sales emails. “We’d love to help your team with X.” “Quick question about Y.” “Can I get 15 minutes?” I used to delete them all. Now I have an AI agent that reads every pitch, figures out which ones are real, and auto-drafts a personalized reverse pitch turning their sales email into a partnership opportunity for Theanna. It runs every morning at 9am. I wake up, check my drafts, and hit send on the good ones. Here’s how I built it, the exact tech stack, and how you can build your own in an afternoon.

What’s in This Post

• The Problem: Founders Are Drowning in Cold Pitches
• The Idea: Flip the Script With AI
• How the Reverse Pitch Agent Works
• Step 1: Scan and Classify With Claude
• Step 2: Rank by Partnership Potential
• Step 3: Auto-Generate Reverse Pitches
• Step 4: Draft Replies in Gmail
• Step 5: Summary Email and Cron Automation
• The Tech Stack
• What the Output Looks Like
• Results So Far
• How to Build Your Own Reverse Pitch Agent
• Why This Matters for Founders
• What Else You Can Build With This
• Copy This Prompt and Build It Yourself

The Problem: Founders Are Drowning in Cold Pitches

If you’re a founder with any kind of online presence, your inbox is a war zone. I have 50,000+ followers across LinkedIn, Instagram, TikTok, and Threads. I post daily about building Theanna. And that means every sales team on earth thinks I’m a lead.

On any given day I get 10–20 cold emails from companies pitching me their product. Dev shops wanting to build my features. Accounting tools. Scheduling apps. AI analytics platforms. LinkedIn growth bots. Some of them are genuinely great products. Most of them are spam. All of them take time to read and decide what to do with.

The old playbook is to ignore them. Archive, delete, move on. But here’s what I realized: some of these companies are exactly the kind of partners Theanna needs. They have products our founders would use. They have audiences that overlap with ours. They’re already reaching out to us. The intent is there — it’s just pointing in the wrong direction.

“Every cold pitch is a warm lead for a partnership. You just have to flip the script.”

The Idea: Flip the Script With AI

Instead of ignoring these emails or manually replying to each one, I built an AI agent that does the entire pipeline for me. Every morning at 9am it:

• Scans my inbox and spam folder for the last 24 hours
• Classifies every email using Claude — is this a real company pitching a real product?
• Filters out the noise — mass marketing, scams, generic lead-gen, dev shops, acquisition spam
• Ranks the real pitches by how relevant they are for a Theanna partnership
• Generates a personalized reverse pitch for each one — acknowledging what they do and proposing a partnership swap
• Creates threaded Gmail draft replies so the response shows up in the original conversation
• Emails me a summary with a ranked list of everything it drafted

All I do is open Gmail, review the drafts, and hit send on the ones I like. Delete the rest. The whole review process takes me 5 minutes.

How the Reverse Pitch Agent Works

Let me walk you through the exact system. It’s a standalone Node.js/TypeScript project that connects to Gmail via IMAP and uses the Anthropic API to classify and generate emails. No Google Cloud project needed. No OAuth dance. Just a Gmail App Password and an Anthropic API key.

Step 1: Scan and Classify With Claude

The agent connects to Gmail via IMAP using a simple App Password — takes 30 seconds to set up at google.com/apppasswords. It pulls every email from the last 24 hours across both Inbox and Spam. Yes, spam. A lot of cold pitches end up in spam, and some of the best partnership opportunities are buried there.

Each email gets sent to Claude with a classification prompt. Claude determines: is this a real company pitching a real product that could become a partnership? It’s not just looking for “sales email” — it’s specifically filtering for companies with products that our founder community would actually use.

The classifier is trained to filter out:

• Mass marketing from big companies (Zoom upgrades, Google Workspace promos)
• Private equity acquisition spam
• Dev shops pitching to build your product for you
• SEO audit and LinkedIn growth hack cold outreach
• Vague “let’s collaborate” emails with no specific product
• Any email with sketchy domain mismatches or mass template vibes

In my first real run, the agent scanned 47 emails and correctly identified 7 legitimate pitches while filtering out all the noise. That’s a 15% hit rate — which honestly tracks with my manual experience.

Step 2: Rank by Partnership Potential

Not all pitches are equal. Claude assigns a relevance score from 1–5 based on how well the product fits Theanna’s audience of early-stage founders. A 5 means “every founder needs this” — think accounting tools, productivity apps, AI platforms. A 1 means “technically a real pitch but not our world.”

The summary email ranks them highest to lowest, so I see the best opportunities first. On a typical day I’ll have 2–3 high-relevance pitches worth responding to and a few more that are interesting but not urgent.

Step 3: Auto-Generate Reverse Pitches

This is where it gets fun. For each identified pitch, Claude generates a personalized reverse pitch email. The agent has a pitch.md file that contains everything about Theanna — who we serve, our reach, our current partners, and what partnership models we offer. Claude reads this context and writes a reply that:

• Acknowledges their product — shows we actually read their email
• Bridges to the partnership — connects what they do to our founder audience
• Hits them with the numbers — 50k+ followers, 3k newsletter subscribers, 300k quarterly impressions
• Proposes the swap — premium placement in our content, founder discount from them, newsletter swap, cross-visibility
• Drops social proof — mentions current partners like Customer.io, Wispr Flow, and Granola
• Closes warm — “Down to chat about what this could look like?”

Here’s an example of a reverse pitch the agent generated for a fintech startup that pitched us their accounting tool:

“Hey! Love what you’re building — AI-powered accounting for VC-backed startups is exactly what founders need when they’re drowning in spreadsheets. Quick flip: instead of you helping us, what if we helped each other? We’ve got 50k+ followers, 3k newsletter subscribers, and 300k quarterly impressions. The swap: premium feature in our content, founder discount from you, newsletter swap. We’ve done this with our current partners — founders love the perks, partners love the signups. Down to chat?”

That email was generated in under 10 seconds. It references their specific product. It’s in my voice. It includes real numbers. And it’s sitting in my Gmail Drafts ready to send.

Step 4: Draft Replies in Gmail

The agent doesn’t just generate the text — it creates actual Gmail drafts that are threaded on the original email conversation. So when the recipient gets my reply, it shows up in the same thread as their original pitch. Not a random standalone email. Not a new conversation. A reply.

This is critical for response rates. When someone sees a reply in their own thread, it feels like a real conversation, not an automated blast. The draft includes the right `In-Reply-To` and `References` headers so Gmail threads them correctly.

The drafts are also formatted as proper HTML emails — flowing paragraphs, clean bullet points, bold text where it matters. Not the broken line-wrapped plain text that makes you look like you copied and pasted from a terminal.

Step 5: Summary Email and Cron Automation

After all drafts are created, the agent sends me a summary email with a ranked list:

Reverse Pitch Report — 2026-05-06
========================================

Found 7 pitches. Drafts are ready in your Gmail Drafts folder.

1. [Company A] ★★★★★
   Pitching: AI-powered accounting for VC-backed startups
   Draft: Ready in Drafts

2. [Company B] ★★★★☆
   Pitching: Video recording/transcription with 50% startup discount
   Draft: Ready in Drafts

3. [Company C] ★★★★☆
   Pitching: AI content analytics for creators
   Draft: Ready in Drafts

The whole thing runs on a cron job. Weekdays at 9am. I don’t think about it. I just wake up, check my inbox, open Gmail Drafts, and decide which partnerships to pursue today. Five minutes and I’m done.

The Tech Stack

The entire agent is a lightweight Node.js/TypeScript project. No frameworks. No cloud infrastructure. No database. It runs on my laptop.

No OAuth flow. No Google Cloud Console. No refresh tokens. You generate an App Password at google.com/apppasswords, paste it into a `.env` file, and you’re done. The whole setup takes about 10 minutes.

What the Output Looks Like

Let’s look at a real day. On May 6, 2026, the agent scanned 47 emails across my inbox and spam folder. It filtered out 40 non-pitches (newsletters, notifications, personal emails, spam) and identified 7 real company pitches:

I sent 3 of those drafts. Two already replied within 24 hours wanting to set up a call. That’s a 43% response rate on cold partnership outreach that took me zero effort to generate.

Results So Far

I’ve been running the agent for a week. Here’s what’s happened:

• 47 pitches identified across 5 business days
• 32 reverse pitch drafts sent
• 14 replies received (44% response rate)
• 6 partnership calls scheduled
• 2 deals in progress (one newsletter swap, one founder perk partnership)
• Time spent per day: 5 minutes reviewing drafts

Before the agent, I was responding to maybe one pitch a week — if I had time. Now I’m having 1–2 partnership conversations a day with companies that approached me first. The pipeline is full and I didn’t have to cold outreach a single person.

“The best partnerships come from companies that already want to talk to you. You just need a system to talk back.”

How to Build Your Own Reverse Pitch Agent

You don’t need to be technical to build this. If you have Claude Code, you can literally say “build me an email agent that reads my inbox and drafts partnership replies” and it’ll scaffold the whole thing. Here’s the high-level process:

1. Set up a Gmail App Password. Go to google.com/apppasswords. Generate a password. This is how your agent authenticates with Gmail without needing a full Google Cloud project.

2. Get an Anthropic API key. Go to console.anthropic.com. Create an account. Generate an API key. This powers the Claude classification and pitch generation.

3. Create your pitch.md file. This is the most important part. Write down everything about your company: who you serve, your reach, your current partners, what partnership models you offer, and how you want the tone to sound. This is the single source of truth the agent uses to write your pitches.

4. Build the agent with Claude Code. Open Claude Code in your terminal and describe what you want. It’ll create the IMAP connection, the classifier, the pitch generator, and the Gmail draft creator. The entire codebase is about 400 lines of TypeScript.

5. Set up a cron job. One line in your crontab schedules it to run every weekday morning. You never think about it again.

The whole build took me about 2 hours with Claude Code. Most of that was refining the pitch voice and tightening the spam classifier. The actual infrastructure — IMAP, Claude API, Gmail drafts, cron — Claude built in minutes.

Why This Matters for Founders

Partnerships are one of the highest-leverage activities for early-stage founders. A single good partnership can get you distribution, credibility, and revenue you’d never get from paid ads. But most founders don’t have time to do business development on top of everything else.

This agent solves that by flipping the funnel. Instead of cold outreach, you’re responding to companies that already want to talk to you. Instead of spending hours crafting partnership emails, Claude writes them in your voice in seconds. Instead of a BD team, you have a script that runs at 9am.

At Theanna, partnerships with companies like Customer.io, Wispr Flow, and Granola have been some of our highest-value relationships. They didn’t come from a sales team. They came from conversations. Now those conversations start automatically.

If you’re a founder sitting on a mountain of ignored cold pitches, you’re sitting on a mountain of potential partnerships. Build the agent. Flip the script. Let AI do the heavy lifting while you focus on the relationships that matter.

What Else You Can Build With This

The reverse pitch agent is just the starting point. Once you have an AI agent that reads your email and understands your business, you can bolt on a full outbound partnership pipeline. Here’s what I’m building next — and what you can too:

A full outreach agent that does the research for you. Right now the reverse pitch agent responds to inbound pitches. The next version proactively researches companies before drafting. Imagine: the agent reads a pitch from a fintech startup, pulls their LinkedIn page, checks their blog for recent funding announcements, looks at their team size, and uses all of that context to write a hyper-personalized partnership pitch. Not “Hey, love what you’re doing” — but “Congrats on your $4M seed round last month. Your 12-person team is exactly the kind of lean startup our 3,000 founders want to learn from.”

Auto-log every contact in your CRM. We use Folk as our CRM for partnership tracking. The next version of the agent will connect to the Folk API and automatically create a contact for every company it pitches — with the company name, sender info, what they pitched, what we pitched back, and the date. No more manually logging partnerships in a spreadsheet. Every conversation is tracked from the first email.

Track the full pipeline. Once contacts are in Folk, you can build pipeline stages: Pitched, Replied, Call Scheduled, Deal in Progress, Active Partner. The agent handles the first two stages automatically. You pick up from “Call Scheduled” onward. That means your partnership pipeline is always warm and always organized — without you touching a CRM until there’s a real conversation happening.

Draft follow-ups automatically. Someone didn’t reply to your reverse pitch? The agent can check Folk for contacts in the “Pitched” stage with no reply after 5 days and auto-draft a follow-up. Warm, casual, one line: “Hey Sam — wanted to bump this in case it got buried. Still think there’s something fun here.” Same voice. Same thread. Zero effort.

The endgame: a partnership OS. Inbound scanning, outbound research, CRM logging, pipeline tracking, automated follow-ups — all running in the background while you focus on the actual relationships. This is what business development looks like when you’re a solo founder with AI. You don’t need a BD team. You need agents.

“The future of founder BD isn’t hiring a team. It’s building agents that do the busywork so you can focus on the handshake.”

Copy This Prompt and Build It Yourself

You don’t need to reverse-engineer anything. Here’s the exact prompt you can paste into Claude Code (or any Claude instance) to build your own reverse pitch agent from scratch. Just replace the placeholder info with your own business details.

Build me a reverse pitch email agent as a standalone Node.js/TypeScript project.

What it does:
Scans my Gmail inbox and spam folder via IMAP for cold sales/pitch emails from the last 24 hours. Uses Claude to classify each email as a pitch or not. For real pitches, it ranks them by relevance (1-5 stars) based on how well the company aligns with my business. Then it auto-generates a personalized reverse pitch — flipping their inbound sales email into an outbound partnership opportunity from me. Creates threaded Gmail draft replies (not standalone drafts — replies on the original email thread) and sends me a summary email with the ranked list.

Tech stack:
- Node.js + TypeScript (run with tsx)
- imapflow for IMAP email access
- Gmail App Password for auth (no Google Cloud project or OAuth needed)
- @anthropic-ai/sdk for Claude API (classification + pitch generation)
- nodemailer for sending the summary email via SMTP
- dotenv for environment variables
- node-cron for scheduling (optional)

Project structure:
- src/config.ts — env vars (GMAIL_ADDRESS, GMAIL_APP_PASSWORD, ANTHROPIC_API_KEY), constants
- src/gmail.ts — IMAP fetch from inbox + spam, create threaded HTML drafts, send summary email
- src/classifier.ts — Claude prompt to classify pitch vs not-pitch, extract company name + pitch summary + relevance score (1-5)
- src/pitcher.ts — reads pitch.md for my business context, generates personalized reverse pitches with Claude
- src/run.ts — main entry point: fetch → classify → rank → generate pitches → create drafts → email summary
- pitch.md — single source of truth for my business info, partnership value props, and tone

Classifier rules:
- YES: companies actively pitching a product or service that could be a real partnership
- NO: newsletters, receipts, personal emails, mass spam, dev shop cold emails, vague "let's collaborate" messages, LinkedIn growth hacks, SEO/audit spam, PE acquisition outreach
- Return JSON: { isPitch, companyName, pitchSummary, relevance }

Reverse pitch tone:
- Write as me (the founder), not a marketing team
- Reference their specific product (show I read their email)
- Use bullet points for the value props
- Include real numbers (followers, subscribers, impressions)
- Warm, direct, collaborative — not corporate or salesy
- Keep it short — 150 words max

Draft threading:
- Use In-Reply-To and References headers pointing to the original message ID
- Set the subject to "Re: [original subject]"
- Append the draft to [Gmail]/Drafts via IMAP so it shows as a reply in the original thread
- Use text/html content type so the email renders like a normal email (not monospace/code)

Summary email:
- Send to my own email address
- Subject: "Reverse Pitch Report — YYYY-MM-DD"
- List all pitches ranked by relevance with star ratings
- Note that drafts are ready in Gmail Drafts folder

My business info (replace this with yours):
- Company: [YOUR COMPANY NAME]
- What we do: [ONE LINE DESCRIPTION]
- Who we serve: [YOUR TARGET AUDIENCE]
- Audience: [FOLLOWER COUNT] followers, [SUBSCRIBER COUNT] newsletter subscribers, [IMPRESSIONS] quarterly impressions
- Current partners: [LIST YOUR CURRENT PARTNERS]
- Partnership offer: premium placement in content, founder discounts, newsletter swaps, qualified lead flow, co-marketing
- My name: [YOUR NAME]
- My email: [YOUR EMAIL]

Setup: Gmail App Password (Google Account → Security → 2-Step Verification → App Passwords), Anthropic API key. Three env vars in .env file. Run with: npx tsx src/run.ts

Optional: add a cron job to run it every weekday morning automatically.

Paste that into Claude Code, fill in the business info section at the bottom, and let it build the whole thing. You’ll have a working reverse pitch agent in under an hour.