Sales GPT

Sample Excel(.xls) sheet file of leads that the user uploads

Inspiration

Last year, I ran my own MLH Hackathon called WWPHacks and had to create over 300 personalized messages for business owners to sponsor my hackathon. I remember thinking to myself how laborious this task must be for an actual sales team contacting over 1000+ customers every day. Soon later, all my friends started using this online chatbot called ChatGPT to finish their essays. Hence, our team came up with the idea of using GPT to automate the task of creating personalized messages to customers where we convert leads into sales.

What it does

Sales GPT automates the grueling and laborious process sales teams have to endure when sending messages to customers/leads. Our app analyses the customer’s behavior, interests, and preferences to craft messaging tailored to their individual needs by using OpenAI’s GPT API. All the user has to do is connect our app to their LinkedIn sales navigator, Salesforce, Hubspot, or upload an Excel sheet file with their lead's information, and our app will return all the messages for each lead to the user!

How we built it

Our web application’s front end runs on the React.JS framework and was designed using Tailwind CSS. Our backend runs on Node.js and Express.js. We used OpenAIs GPT API in order to generate messaging that is not only personalized but also highly effective at converting leads into sales. We used MongoDB as our database where we stored user information, tokens, and the leads of that user.

Challenges we ran into

In order to analyze a customer's behavior and interests, we need to be able to have real-time data, but GPT-3 runs on a model that uses data from 2021. GPT-4 and its plugins allow us to access the internet for real-time information that would help us craft more personalized messages. Since we didn’t have access to the GPT-4 API, we decided to stick with GPT-3. Moreover, we had multiple errors in deploying our backend to Vercel due to some issues related to Express.js so we had to make a last-minute decision in deploying our backend to Heroku. We also didn't have time to integrate LinkedIn sales navigator, Salesforce, and HubSpot connection to our app.

Accomplishments that we're proud of

We are very proud to have a working prototype in just 24 hours.

What we learned

We learned how to use the gpt-3 API and gained much knowledge in gpt-4 and its plugins through research.

What's next for Sales GPT

We are going to add more features to our current prototype such as an account sidebar which was included in our UI/UX design but not our actual app. We will also include a custom prompt option where the user can create their own custom prompt for GPT to generate a response for the user leads. Once we have access to GPT-4s API and its plugins, we will include it in our backend so that it can create personalized responses using real-time data from the internet and the customer's social media accounts. We also hope to elevate our prototype to the MVP stage for our potential beta customers.