Inspiration
Our inspiration was to create a personalized product explorer chatbot that feels like the in-store experience and conversing with a Verizon retail sales associate. Currently, the chatbot application on the company’s website connects customers with a “Live Agent.” Therefore, we began by chatting with Agents through the current chat feature on Verizon’s website to gauge the current capabilities of the service and the general sentiment of the employees who interact with customers. According to Lauren, an agent with Verizon who replies to customers in the chat interface, “Our AI bot will only initiate in the start of the conversation if there's no available agents yet at the moment during a high volume of queue.” “But the bot may just provide general information, like for example, if the customer is looking for an upgrade, the bot may [gather] FAQs related with upgrade options or may provide the latest deals and promotions we may have,” Lauren said. Based on the current, limited capabilities of the AI bot feature, we were inspired to develop a GenAI chatbot that assists customers in exploring and selecting Verizon products based on his or her specific needs and preferences, which would eliminate the need for a “Live Agent” and expedite the process for the consumer. We implemented a recommendation system to suggest smartphones by evaluating the user’s purchasing behaviors, preferences and interaction history with the chatbot.
How we collaborated
Our team includes five people representing the University of Florida’s Open-Source Club. The members are as follows: Robert Conde, a computer science student with a background in mathematics, Nikhil Iyer, a computer engineering student, Grayson Kornberg, who studies computer science with an expertise in statistics, Jonathan Mesa, another computer science student and the president of our student organization, and Rebecca Weinstein, a public relations student responsible for communicating our project. We divided the work by playing on our personal strengths. Conde is a front-end developer wizard, so he led the efforts with coding the interfaces. Iyer debated with ChatGPT and constructed Vivian Verizon’s functionality. Kornberg, given his statistical skills, compiled the data for identifying the products that the AI can suggest. Mesa became a UI/UX designer over the 48 hours we worked on this project and coded a responsive website that is visually appealing and aligned with Verizon’s brand voice. Weinstein used her experience in communications to give Vivian Verizon a personality, tone of voice and created the presentation deck. Having a person on the team with a background in public relations helped us with establishing a creative platform, researching demographics for strategic purposes and providing a SWOT analysis for Verizon.
What it does (the functionality)
Vivian Verizon is a friendly, digital sales representative created with artificial intelligence. She offers insights into Verizon’s plans, promotions and deals with a positive attitude. Her job is to help Verizon customers purchase the perfect phone in the most efficient way. Vivian Verizon is dedicated to assisting customers with purchasing a phone and tries her best to support purchasing products without the need to go to a physical store.
How we built it
We collected data on Verizon and its products to prompt and train our model, which uses GPT 4.0, the newest and best publicly available large language model. Then, we created a React.js application as the front-end for the user to interact with. We made an interface between GPT and our website to allow Vivian Verizon to interact with users more dynamically than a typical chatbot. To facilitate communication with our data, the user and the ChatGPT API, we created a Flask backend. Finally, we connected the front-end to the backend to present Vivian Verizon to consumers.
Challenges we ran into
Vivian Verizon understands her job description, but in some iterations, she developed a sense of humor. About 12 hours before submission, around 5 a.m., we asked Viv for help purchasing a phone. She provided us with a URL to Verizon’s website that did not actually exist, so we asked her to try an embedded video. That’s when Vivian Verizon Rickrolled us. She gave us a link called “Click here to watch!” and it redirected us to a video of Rick Astley singing. It was a needed reminder for our team as it helped us recognize the developer’s responsibilities to have an ethical AI thinking model and cleared Vivian Verizon’s context.
Accomplishments that we are proud of
Verizon prides itself on providing a fast and reliable network; therefore, we are proud that our software performs on par with the company’s mission and values. This software is another simple solution from the company to connect customers effortlessly to Verizon’s products. Taking into consideration the intense timeframe of this challenge, we were still able to simulate a proper product development life cycle. We started at the initial idea stage with conducting user interviews to provide insight into consumer behaviors, applied multiple forms of prototyping and ensured we performed quality assurance testing.
What we learned
The power of artificial intelligence grows daily, which in developing this project, we were fascinated by the actions the AI took. We learned how to use technologies and softwares that we had no prior experience using, like Flask, React-Bootstrap, GPT API and Figma. Also, we began understanding large model theories to grasp concepts we would have never otherwise considered. While we are all members of the Open-Source Club, the five of us did not work together in a team setting before starting this hackathon, so we learned how to work quickly to create a solution and leverage our individual skills.
The future of Vivian Verizon
Vivian Verizon, or Viv for short, can become the digital spokesperson for the company like Flo from Progressive or Lily from the AT&T commercials. Considering she is the first of her kind, Verizon has an opportunity to disrupt the industry. By launching Vivian Verizon as an AI support system for customers, it would redirect share of voice to the company, distinguish Verizon from its competitors and provide more opportunities for consumers to purchase Verizon products.
Additional Links
Built with
Love by Team Open-Source Club ❤️
Languages: Jupyter Notebook, TypeScript, Python, HTML, CSS, JavaScript, JSON, React-Bootstrap
Built With
- embedding
- flask
- json
- llm
- openai
- python
- react
- typescript
