Deal Pilot AI

Inspiration

Online shopping has become easier than ever, but the experience is still largely manual. Customers spend time browsing products, comparing prices, searching for discounts, and navigating checkout processes. Meanwhile, small businesses often struggle to provide personalized assistance, negotiate fairly with customers, and manage sales efficiently.

We asked ourselves a simple question: What if shopping could feel like having a conversation with a knowledgeable sales assistant?

That question inspired DealPilot AI, an intelligent shopping platform that combines conversational artificial intelligence with PayPal's payment infrastructure.

Our vision was to create a shopping experience where customers could simply describe what they wanted, receive personalized recommendations, negotiate within fair pricing limits, and complete their purchases seamlessly.

We wanted to demonstrate how AI could go beyond answering questions and take meaningful actions that connect conversations to real commerce.

What it does

DealPilot AI is an AI-powered shopping and negotiation platform that transforms natural-language requests into real shopping actions.

Instead of manually searching through a catalogue and adding products one by one, customers can interact with an intelligent shopping assistant.

For example, a customer can say:

"Add five T-shirts and two caps to my cart."

The AI understands the request, identifies the relevant products, and automatically updates the shopping cart.

Customers can also ask questions about available products, request recommendations based on their budgets, and negotiate discounts.

However, unlike an unrestricted chatbot, DealPilot AI operates within predefined merchant pricing rules. It cannot invent products, offer unauthorized discounts, or override business constraints.

Key features include:

Conversational shopping: Customers browse and interact with products using natural language.
AI-powered cart management: The assistant can add products and quantities directly to the shopping cart.
Intelligent negotiations: Customers can request discounts while the system enforces merchant-approved limits.
Dynamic quotations: The platform calculates item totals and applicable discounts transparently.
PayPal Sandbox checkout: Customers approve payments through PayPal's checkout experience.
Verified payment capture: The backend validates and captures authorized payments before confirming orders.
Order history: Customers can view successfully completed purchases.
Seller workspace: Merchants can access tools for managing their catalogue and pricing rules.
Guest access: Visitors can interact with the shopping assistant without needing a ChatGPT account.
Modern responsive interface: A polished storefront with realistic product imagery and clear, accessible shopping controls.

The result is a connected experience, from the customer's first message to a completed PayPal payment.

How we built it

We developed DealPilot AI as a full-stack web application, bringing together modern web technologies, OpenAI's API, and PayPal's developer platform.

Frontend and user experience

We built a responsive e-commerce interface featuring a product catalogue, conversational shopping assistant, interactive cart, quotation summary, order history, and seller workspace.

The interface was refined to use a professional blue-and-white design, realistic product photographs, clearer navigation, and stronger visual affordances.

Artificial intelligence

We integrated the OpenAI API through a secure server-side endpoint. The assistant interprets customers' natural-language shopping requests and uses the application's product catalogue and merchant rules to produce grounded responses.

Rather than allowing the model to control prices independently, the application enforces business rules and validates shopping actions through its backend.

PayPal integration

We integrated PayPal's Sandbox environment using its checkout capabilities and Orders API.

The application creates payment orders, requires the customer to approve the transaction, captures the authorized payment, verifies the result, and records successful purchases in the order history.

This allowed us to demonstrate a genuine end-to-end payment workflow using sandbox funds.

Security and reliability

Sensitive credentials are stored as server-side environment secrets rather than embedded in frontend code. We implemented pricing validation, guest-access controls, and error handling.

We also maintained automated tests covering key functionality, including shopping behaviour, payment rules, and application access.

The complete source code, documentation, setup instructions, and MIT License are available in our public GitHub repository.

Challenges we ran into

Building DealPilot AI presented several challenges, particularly because we wanted to deliver a functional product rather than a static demonstration.

1. Integrating real AI functionality

Our initial shopping assistant used simulated responses. Moving to an actual AI model required implementing secure API communication, connecting the model to product data, and ensuring that responses respected merchant rules.

We also encountered API usage-limit errors before configuring the appropriate API billing.

2. Connecting PayPal payments

Setting up PayPal's Sandbox environment was one of our biggest technical learning experiences.

We had to understand the difference between merchant and buyer sandbox accounts, configure application credentials, and implement the payment approval and capture process.

We eventually completed a successful sandbox transaction and verified that the purchase appeared in the customer's order history.

3. Balancing AI flexibility with business rules

A conversational agent needs flexibility to understand customer requests, but an e-commerce platform cannot allow arbitrary prices or unlimited discounts.

We addressed this by grounding the assistant in the catalogue and enforcing merchant-approved pricing through application logic.

4. Making the application accessible to judges

Although our website was publicly accessible, we discovered that the shopping assistant initially required visitors to sign in with ChatGPT.

We corrected this limitation and successfully tested guest AI-powered cart management in a signed-out browser session.

5. Time constraints

We developed the project while balancing university examinations and academic responsibilities.

Working under a tight deadline taught us to prioritize essential functionality, preserve stable versions, and focus on completing a reliable end-to-end shopping experience.

Accomplishments that we're proud of

Our biggest accomplishment is transforming an initial idea into a working AI-powered commerce application.

We are particularly proud that:

We successfully integrated a real OpenAI-powered shopping assistant.
Our AI understands natural-language product requests and updates the shopping cart.
The assistant enforces merchant-approved discount rules rather than blindly accepting customer demands.
We completed a real PayPal Sandbox checkout, including payment approval, capture, and order confirmation.
Successfully captured transactions appear in the customer's order history.
Visitors can access and use the AI shopping assistant without signing in to ChatGPT.
We redesigned the interface into a polished and intuitive shopping experience.
The final source export passed 26 automated tests, TypeScript checks, and a production build.
We published the complete application source code and documentation on GitHub.

Most importantly, we built a prototype that demonstrates the full journey from AI conversation to an approved PayPal payment.

For us, this was more than completing a hackathon project. It was an opportunity to apply our technical knowledge to a practical business problem.

What we learned

Building DealPilot AI helped us understand how several important technologies work together in real-world software.

AI can do more than generate text.

We learned how to turn natural-language instructions into meaningful application actions, allowing an AI assistant to participate in the purchasing process.

AI-powered systems need clear boundaries.

We discovered the importance of grounding AI responses in reliable business data and enforcing critical operations through deterministic backend validation.

Payment integration requires careful engineering.

Working with PayPal taught us about developer credentials, sandbox environments, customer approval, server-side payment capture, and transaction verification.

User experience is part of functionality.

We gained a deeper appreciation for interface design principles, including visual hierarchy, affordances, signifiers, accessibility, and clear feedback.

Testing matters as much as implementation.

Automated tests helped us protect existing functionality as we added new capabilities, while manual tests confirmed that real AI requests and sandbox payments worked.

Finally, we learned the importance of teamwork, perseverance, and building incrementally under pressure.

What's next for DealPilot AI

Our vision extends beyond the hackathon.

We want DealPilot AI to evolve into an intelligent commerce platform that helps small and medium-sized businesses deliver personalized shopping experiences without requiring large sales or customer-support teams.

Our planned improvements include:

1. More advanced AI shopping agents

Enable the assistant to compare multiple products, create personalized shopping plans, and guide customers through more complex purchasing decisions.

2. Merchant intelligence dashboard

Provide deeper insights into sales performance, popular products, customer demand, discount effectiveness, and inventory trends.

3. Multi-merchant support

Expand the platform to allow multiple independent merchants to manage their own catalogues, business rules, and storefronts.

4. Personalized recommendations

Improve product suggestions based on customer preferences, budgets, and shopping behaviour, with appropriate privacy protections.

5. More advanced payments

Explore additional PayPal capabilities, including refunds, payment lifecycle management, and merchant onboarding.

6. Production readiness

Strengthen authentication, fraud prevention, infrastructure monitoring, security controls, and scalability before introducing real-money transactions.

Ultimately, our goal is to make online commerce more conversational, intelligent, and accessible.

DealPilot AI — Good things. Better deals.

From conversation to commerce, powered by AI and PayPal.

Live Demo: https://dealpilot-ai-bongani.sugaryhero129.chatgpt.site

GitHub Repository: https://github.com/Bongani017/dealpilot-ai