LeadRescue: An Open-Source AI Assistant That Helps Small Businesses Rescue Forgotten Leads

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle.

What if a small business could lose fewer potential customers simply by following up at the right time?

Every small business receives enquiries. They come from website forms, referrals, social media, spreadsheets, and direct messages.

But receiving an enquiry is only the beginning.

Some leads get an immediate response. Others get forgotten in a spreadsheet. A promising prospect gets marked as "contacted" but never receives another message. A potential customer who was genuinely interested eventually moves on.

Large companies can invest in sophisticated customer relationship management (CRM) platforms and sales automation tools. For small businesses, those tools can be expensive, complicated, or unnecessarily feature-heavy.

I wanted to explore a simpler idea:

What if a small business could turn an ordinary spreadsheet of leads into an actionable follow-up workflow, powered by open-source AI?

That idea became LeadRescue.

🔗 Live Demo: https://lead-rescue-b4i8.vercel.app

💻 GitHub Repository: https://github.com/dhruv-dagar/LeadRescue

💡 The Problem: Leads Don't Always Need More Marketing

Sometimes, businesses don't need more leads. They need to handle the leads they already have more effectively.

Imagine a small agency with 50 enquiries. The owner is busy managing projects, answering calls, and handling daily operations. There is no dedicated sales team.

A few days later, several enquiries are still waiting for responses.

Which ones should receive attention first? Which prospects have already been contacted? What should the next message say?

Without a clear workflow, answering these questions becomes another manual task.

LeadRescue focuses on this gap: helping small businesses organize enquiries, identify follow-up priorities, track progress, and prepare messages without requiring a complicated CRM setup.

🚀 What Is LeadRescue?

LeadRescue is an open-source lead management and follow-up assistant built around a straightforward workflow.

Instead of treating a lead spreadsheet as a passive list, it turns the information into something a business owner can act on.

Core capabilities

• Lead import: Import leads from CSV files, with CSV and Excel (.xlsx) support in the full local application.
• Lead organization: Keep lead information and follow-up status in one dashboard.
• Follow-up prioritization: Identify leads that need attention based on their current state and available information.
• Status tracking: Track progress as leads move through the follow-up process.
• AI-assisted message drafting: Use a locally running open-weight language model to prepare concise follow-up messages.
• Human review: Review and edit suggested messages before sending them to a customer.

The goal isn't to replace human sales conversations. It's to reduce the friction between receiving an enquiry and taking the next useful action.

🤖 Why Open-Source AI?

This project is part of the Hacktoberfest Open-Source AI Challenge, so using AI was not enough. I wanted open-source AI to play a meaningful role in how the product works.

LeadRescue uses Ollama to run an open-weight language model locally. Its default model configuration is qwen3:4b-q4_K_M.

That gives the full local application a way to generate follow-up drafts without requiring every lead and customer conversation to be sent to a third-party hosted AI API.

Why this matters

1. More control over data

Lead information can be sensitive. Running inference locally provides more control over where data is processed, although users still need to secure their own machine, database, and backups.

2. No per-request hosted AI API bill

Once the model is available locally, generating drafts does not require paying a separate hosted AI API provider per request. Local inference still uses hardware resources, electricity, and storage.

3. A more accessible starting point

Small businesses and independent developers can experiment with AI-assisted workflows without immediately committing to a paid AI service.

4. Inspectable and adaptable technology

The application and its integration with the model can be inspected and modified. The model itself is distributed under its own terms, so users should check the applicable model licence before commercial use.

For me, the key lesson is that open-source AI is most interesting when it changes the economics, privacy trade-offs, and accessibility of a real workflow—not when it is added merely to make an application sound intelligent.

🛠️ How It Works

The full local application follows a simple architecture:

• React + Vite: The dashboard and user interface.
• Node.js + Express: The backend API for lead operations.
• SQLite: Local database storage.
• Ollama + Qwen3: Local language-model inference for follow-up drafts.

The workflow looks like this:

1. Import a CSV or Excel spreadsheet containing lead information.
2. Review the leads in the dashboard.
3. Identify leads that need follow-up.
4. Update their statuses as the conversation progresses.
5. Generate a draft follow-up message using the local AI model.
6. Review the draft and decide what to send.

The AI assists with preparation; it does not autonomously contact customers.

This distinction is intentional. An AI-generated message should be a starting point for a human decision, not an instruction to send something without review.

🌐 Try the Demo

You can try LeadRescue without installing anything:

https://lead-rescue-b4i8.vercel.app

The public demo is designed to let people explore the core experience. It includes fictional sample leads, CSV import, browser-based persistence, status updates, and illustrative follow-up templates.

One important distinction: the public demo does not connect to the live AI backend. Its example messages are templates, not responses generated by an active language model. The full local application is where SQLite-backed storage and Ollama-powered generation are available.

I have kept this distinction explicit rather than presenting a browser-only demonstration as a fully hosted AI service.

💻 Run LeadRescue Locally

Want to explore the actual application and local AI integration?

Prerequisites

• Node.js and npm
• Ollama
• Git

1. Clone the repository

git clone https://github.com/dhruv-dagar/LeadRescue.git
cd LeadRescue

2. Download the model

ollama pull qwen3:4b-q4_K_M

Make sure Ollama is running before trying AI-assisted message generation.

3. Start the backend

In your first terminal:

cd backend
npm install
npm run dev

The backend runs on port 3001 by default.

4. Start the frontend

In a second terminal, from the repository root:

cd frontend
npm install
npm run dev

Open the local URL printed by Vite in your terminal.

For additional configuration, API details, and troubleshooting, refer to the repository's README:

https://github.com/dhruv-dagar/LeadRescue#readme

💰 Could Someone Actually Pay for This?

That is the question behind the Side Hustle challenge.

I don't want to assume that building a useful tool automatically creates a business. LeadRescue is an early MVP, and finding out whether small businesses will pay for it is part of the next phase.

The initial target users are small agencies, freelancers, consultants, and other small businesses that receive enquiries but don't have a dedicated sales operations team.

A possible business model is a freemium approach:

• Free: Basic lead organization and manual follow-up tracking.
• Paid: More advanced prioritization, configurable workflows, team collaboration, and reporting.

These are potential future offerings, not features currently being sold.

Before building a larger product, I want to validate a few assumptions:

• Do small businesses actually lose meaningful opportunities because of inconsistent follow-ups?
• Would they use a lightweight tool instead of their existing spreadsheets?
• Is AI-generated message preparation valuable enough to justify paying?
• Would they prefer local AI for data control, or would they prioritize convenience over self-hosting?

My next step is to speak with potential users, demonstrate the workflow, collect feedback, and see whether anyone is willing to use it consistently. If the problem is real but the solution is not compelling enough, that feedback should shape the next iteration.

🧠 What I Learned

Building LeadRescue reinforced a few lessons for me:

Start with a workflow, not a model. The business problem is forgotten follow-ups. The model is only one component of the solution.

A demo needs honest boundaries. A public interface and a locally running AI backend have different capabilities. Making that distinction clear matters.

AI suggestions should remain reviewable. Customer-facing communication deserves human oversight, particularly when generated text could misrepresent an offer or relationship.

Deployment is part of the product. Local inference is useful, but it introduces installation, hardware, and configuration requirements. Making the first experience simple is just as important as choosing a model.

There is still work to do around usability, validation with real businesses, deployment, and deciding which features genuinely deserve to exist.

🔭 What's Next?

My priorities for LeadRescue are:

• Gather feedback from small-business owners and independent professionals.
• Improve follow-up prioritization based on real user needs.
• Make local setup and model configuration easier.
• Add more useful workflow controls without turning the product into an oversized CRM.
• Validate whether there is a credible paid use case before investing in more features.

I'd rather build a smaller product that solves a real problem than a larger application full of features nobody uses.

🙌 Try It, Explore It, and Share Feedback

LeadRescue is an experiment in combining an everyday business problem with open-source software and local AI.

If you run a small business, manage enquiries, or have experience building sales tools, I'd love to hear your thoughts.

• Would this workflow be useful to you?
• What would prevent you from adopting a tool like this?
• Which feature would make the biggest difference?
• Would you prefer local AI, hosted AI, or the option to choose?

Try the demo: https://lead-rescue-b4i8.vercel.app

Explore the code: https://github.com/dhruv-dagar/LeadRescue

If you have suggestions, issues, or ideas for improvement, please share them.

The larger goal is simple: make useful AI-powered tools more accessible to people who cannot afford complex software or expensive AI infrastructure.

Let's see whether LeadRescue can turn that idea into something small businesses genuinely want to use.