Outreach Cupid

Where deals meet destiny.

Comment

Inspiration

Problem definition

Discovery Difficulties: We lack alignment on sales motions at a company scale, which leads to uncertainty about what to pursue. This uncertainty leaves Product and Engineering blocked and frustrated by a lack of knowledge about what prospective deals are coming through the pipeline or being outreached to.

Learning Gaps: Across the company, it’s challenging to gather and share clear information on target companies, including their details, communication styles, and potential regulatory challenges, leaving people feeling uninformed about the types of people we’re working with.

Engagement Inefficiency: We’re losing a lot of time manually reaching out to companies; the tools we’ve wanted to use still require too much manual setup and customization to be effective. And much of the time, the research is owned by a single person, leading to a “bus problem” if they’re OOO or leaving.

Value Add

Time Saved:

  • Reduces prospect qualification time from hours to minutes - Traditional manual research of a single company (checking LinkedIn, news, tech stack, decision makers) takes 2-3 hours. Your swipe interface + automated research can process 20+ companies in the same timeframe.
  • Eliminates repetitive deck customization - Creating a personalized pitch deck currently takes a long time and feels blocked by Marketing. Automated Gamma generation reduces this to minutes, enabling the team to pursue 10x more opportunities.
  • Cuts meeting prep time by 75% - Auto-generated executive summaries mean no more scrambling to research a company 30 minutes before a call.

Revenue Impact:

  • Increases pipeline velocity - Faster qualification and outreach result in more opportunities. If you can reach 50 qualified prospects instead of 5 in a quarter, even with the same conversion rate, you're looking at 10x revenue potential.
  • Improves win rates through personalization - Generic outreach has ~2% response rates. Highly personalized, research-backed outreach can achieve response rates of 15-30%, directly impacting closed deals.
  • Captures opportunities before they're lost - Companies evaluating blockchain solutions often move to competitors within 30-60 days. Your system enables same-day response instead of week-long research cycles.

Risk Reduction:

  • Prevents compliance violations - A built-in review workflow ensures that all outreach is reviewed before sending, protecting against regulatory missteps in the financial services sector.
  • Reduces reputational risk - AI-generated content review checkpoint prevents embarrassing errors or inappropriate messaging that could damage Cosmos Labs' brand.
  • Flags high-risk prospects early - Automated research identifies regulatory red flags, problematic jurisdictions, or competitive conflicts before investing significant BD resources.

What it does

Discover: Easily find companies we care about

Step 1: Outreach Cupid

  • Tinder-style card stack that displays prospect companies one at a time with key info (name, industry, size, location)
  • Swipe gestures (or button clicks) to qualify leads: right = interested, left = pass, up = high priority
  • Real-time sync with Google Sheets - every swipe decision instantly updates your prospect database
  • Session tracking shows how many companies you've reviewed, qualified, and passed in real-time

Step 2: Google Sheets Trigger

  • Monitors "Qualified Leads" worksheet for new company entries
  • Captures company name, industry, and website URL

Step 3: Clay - Create New Company Record

  • Adds the new company to Clay workspace for research
  • Populates initial company data from the spreadsheet

Learn: Understanding our company needs

Step 4: AI - Generate Solutions Targeting

  • Uses Claude AI to analyze company and create targeted keywords
  • References sales templates and strategy documents
  • Produces customized solution targeting recommendations

Step 5: Clay - Update Company Record

  • Adds the AI-generated solutions targeting info back to Clay
  • Enriches the company profile with targeting insights

Step 6: Clay - Retrieve Updated Company Data

  • Fetches the complete company information including new targeting data
  • Ensures all data is available for next steps

Engage: Combining and automating research & outreach

Step 7: Google Docs - Create Executive Summary

  • Generates a formatted executive summary document
  • Saves to "Hackathon Project Files" folder with company name

Step 8: Slack - Send Notification

  • Posts summary to #test-dealflow-notification channel
  • Includes Google Drive link to the executive summary

Step 9: Approval Workflow

  • Sends approval request to David Zhou via Slack
  • Includes all company details and executive summary for review
  • Waits for approval before proceeding (7-day timeout)

Step 10: Gamma - Generate Presentation

  • Creates a 10-slide customized enterprise pitch deck
  • Uses executive summary and company data to personalize content
  • Focuses on Cosmos blockchain infrastructure solutions

Step 11: AI - Draft Email Content

  • Uses Claude AI to write personalized outreach email
  • Incorporates company-specific targeting and context
  • Creates both subject line and email body

Step 12: Gmail - Create Draft Email

  • Generates email draft with AI-written content
  • CCs key team members
  • Ready for manual review and sending

End Result: Complete automated pipeline from lead identification to personalized outreach materials ready for human review and sending.

How we built it

Data Sources:

  • Google Sheets (trigger)
  • Clay (data enrichment)
  • Clay enriches company data, identifies key decision-makers, and runs a multi-step email waterfall validated by multiple email verification products to deliver reliable contacts.

AI Processing:

  • Claude AI (targeting analysis & email writing)
  • ChatGPT (general research and image gen)
  • Gemini (tag line production)
  • Midjourney/Nijijourney
  • Suno (background music generation)
  • ClayAgent (company research)
  • Gemini / Nano banana (image generation)
  • Cursor AI (FE/BE code generation and bug resolution)

Document Creation:

  • Google Docs (executive summary)
  • Gamma (presentation)

Communication:

  • Slack (notifications)
  • Gmail (outreach drafts)

Workflow Management:

  • Zapier (workflow automation)
  • Approval tool (human review checkpoint)

Challenges we ran into

  • Engineering the prompts - It was hard to figure out a cohesive and very easy prompt to input in order to get AI to do what we want and that does not break anything.
  • Clay was very slow for some prompts, especially if they relied on other columns
  • Trying to build the swipe/tinder components from scratch using AI wasn’t very good. Eventually gave up and found some code online to re-use as a base to bootstrap the application
  • AI wrote some buggy code, had to fine-tune a lot of code to fix issues, and AI introduced a lot of regressions while adding new features, or adding features I didn’t prompt for that weren’t desired.

Accomplishments that we're proud of

  • Awesome landing page with perky punch lines
  • Bridging anime aesthetics with corporate use cases
  • Creating an extensive Zapier workflow that automates the discovery phase until outreach
  • UI looks good. Very kawaii.

What we learned

  • AI is still not perfect; you still need to do a lot of manual tweaks when using it.
  • Prompting is difficult; being broad gives more variety but less specificity
  • AI can really uplevel anyone on the team especially if we have basic foundations of company knowledge
  • Setting exact roles early on sets us up for success especially when we’re on a time crunch

What's next for Outreach Cupid

  • Make it more reliable by AI generation delays
  • Improving the formatting of email, presentation and Executive Summary
  • Make the database dynamic ( constantly adapting information)

Built With

Share this project:

Updates