Hackathon · Soapbox AI & Data Science Hackathon3 min read

AI & Data Science Hackathon: Lead Discovery for UPS

Our team proposed ConnectifyAI for UPS at the Soapbox hackathon. Afterwards I built an n8n proof of concept that screens company lists with AI.

From Hackathon Idea to Working PoC: Automating Lead Discovery for UPS with AI

Last week, I had the opportunity to participate in the AI & Data Science Hackathon hosted by Soapbox, where our team worked on a real business challenge from United Parcel Service (UPS). The experience didn’t stop at the hackathon itself, it sparked a deeper exploration that led me to build a working Proof of Concept (PoC) demonstrating how AI-driven automation could significantly improve UPS’s customer acquisition process.

Open the post on LinkedIn

The Challenge: Manual Lead Discovery Doesn’t Scale

Within the UPS sales organization, sales representatives spend a considerable amount of time manually searching the internet to identify potential customers. This process has several drawbacks:

  • No standardized “good-fit” checklist
  • Inconsistent decision-making across sales reps
  • Duplicate efforts and wasted time
  • Difficulty prioritizing which companies to contact

Despite the fact that valuable signals exist across public datasets and internal systems, they aren’t consolidated into a single workflow. The result is lower productivity and reduced conversion rates — especially problematic at scale .

The Idea: AI-Driven Lead Identification

Our hackathon proposal, ConnectifyAI, focused on a simple but powerful idea: replace manual company research with an automated AI-driven workflow.

The goal was to enable UPS to:

  • Automatically filter raw company lists
  • Identify high-value prospects using consistent criteria
  • Prioritize leads based on AI-generated insights

By shifting sales teams away from data screening and toward relationship-building, UPS could dramatically improve efficiency and decision quality .

From Concept to Proof of Concept

After the hackathon, I wanted to validate whether this idea could work in practice. To do that, I built a Proof of Concept using n8n, an open-source workflow automation platform.

The PoC demonstrates how an AI-assisted workflow could be deployed with minimal overhead while remaining flexible and scalable.

The n8n workflow of the lead discovery proof of concept

How the Workflow Works

  1. Automation Trigger The workflow can run on a schedule (daily, weekly, or monthly) or be triggered manually.

  2. Data Collection Company data is fetched from a public dataset (for the PoC, a sample of 50 companies was used).

  3. Data Storage The collected data is stored in a lightweight database (Google Sheets for simplicity).

  4. Existing Customer Filtering Companies that are already UPS customers are filtered out via a (theoretical) Salesforce API integration.

  5. AI-Driven Analysis Each remaining company is analyzed by an AI reasoning model, generating:

    • Company overview
    • Potential UPS services
    • Business scale and geographic reach
    • A priority score from 1–10
  6. Result Aggregation Companies are ranked by priority score, with the top prospects selected.

  7. Automated Reporting A summarized report of the top leads is automatically emailed to sales teams .

Results: Scalable, Effective, and Surprisingly Cheap

The PoC successfully demonstrated that this approach is both technically feasible and operationally practical.

One especially interesting result was cost: even when using one of the most expensive AI APIs available, processing roughly 300 company records across multiple runs cost less than $0.01 USD.

This highlights a key insight: AI-powered automation doesn’t require heavy infrastructure or in-house model development to deliver real business value .

Why This Matters

This project reinforced a few important lessons for me:

  • AI delivers the most value when embedded directly into workflows
  • Automation is as much about consistency as it is about speed
  • Lightweight tools can still support enterprise-scale use cases
  • Hackathon ideas can — and should — evolve into real prototypes

The PoC shows strong potential for integration into UPS’s customer acquisition process, improving productivity, decision-making, and scalability without major operational disruption.

Final Thoughts

What started as a hackathon challenge turned into a concrete demonstration of how AI + automation can solve real-world business problems. I’m excited about the possibilities this approach unlocks — not just for UPS, but for sales-driven organizations facing similar challenges.

If you’re interested in AI-driven workflows, automation, or turning ideas into working systems, I’d love to connect and exchange thoughts.

Published
6 Oct 2025
Topics
n8n, AI