In my IT for Networked Organizations class at the University of Illinois, the final project was to use a variety of tools to create a proof of concept for an app. Throughout the course, we learned how different Information Systems can be used within businesses and the advantage they provide. Later on, we started shifting more toward AI, and how people can use these tools to enhance their workflow. Throughout the course, our professor gave us access to a discord bot that allowed us to call OpenAI’s ChatGPT 3.5 model to help assist us throughout the course. This bot, Erin, was the foundation of my creating LocalePal, a Recommendation System that learns more about the user to provide personalized results through generative artificial intelligence.
Wireframe & Pitch Deck
After speaking with Erin bot to solidify our idea, we then had to create a wireframe and a pitch deck for our product. For my wireframe, I utilized Erin to generate the app’s logo and also to help come up with how I wanted my app to feel. To create the wireframe for LocalePal, I utilized the tool Figma. Click here if you would like to check out the prototype of the wireframe.
Next, we move on to the pitch deck. This is where the Erin bot really showed its strengths, as the assistance it provided for the project was astronomical. Our professor provided us with this example slide deck to use as a template for our pitch deck, and Erin was able to help generate all the information that was required. Speaking to the bot, I was able to refine my business plan, different tools I would use to actually create LocalePal, different market strategies, created fake managers of the company, and even provided data that I could use when creating the deck. Erin was also able to generate fake people that I could use as the faces of said fake managers, however, I hit an image creation limit. To work around this, I utilized This Person Does Not Exist, which is another website that allows users to use AI image creation to generate faces of people who are not real. After finishing the Pitch Deck, I then used Murf.AI to generate an AI voiceover for my pitch deck. Combining all of these tools, I was able to create this Pitch Deck.
Throughout the process of creating the wireframe and pitch deck for Locale Pal, I found different strengths of AI that I never considered before. Before this project, I held an opinion that AI would do all of the work for you, and it would limit the creativity of people in the future. After this project, however, I hold the opposite opinion. Using these tools helped me flush out my ideas, and allowed me to see the product that I had envisioned. Using Erin bot, I was able to bounce ideas back and forth, much like with a coworker, to help wrinkle out any problems I was currently facing in the design phase. I could ask for different options when facing an issue with funding opportunities, ask to help refine my idea on the freemium payment style I was aiming for, and much more. I feel that these tools can help enhance your business creation phase, allowing for different insights that you may never think of otherwise.
AI functionality
The next, and final, step of our project was to create a working prototype of our app through the use of AI technology. After trying out a multitude of chatbot generators, like Langflow and Botpress, I finally decided to work with Flowise. Flowise is an open-source tool that allows users to drag and drop different commands to utilize different Large Language Models (LLMs). Using this, I provided my LLM a JSON file with 60 different businesses, which were generated by Erin, that the bot would recommend to users. These businesses were either classified as a “Restaurant” or an “Activity”, and the bot would use these differentiations to help decide what to recommend. For example, if a user asked for a place to go eat dinner, it would not recommend the user a bike rental shop. Within the restaurant classification, I had Erin generate what type of food they serve (think American, Greek, Indian, French, etc.), how expensive they were, and the average user rating. This allows the user to tell the bot what kind of food they like, and give them recommendations on these. This same type of breakdown was in the Activities as well, with what type of activity they provided, price point, and average user score.
With the use of LLMs, rather than just coding prompts, I was also able to make the chatbot itself create fake distances for the users. When calling the ChatGPT API, there is a variable named Temperature. To put it simply, temperature tells ChatGPT how much freedom it has to be creative with its response. You chose a number between 0–2, and the larger the number is, the more free range ChatGPT has. For example, when I had the temperature set to 0.5 and asked for restaurants that were near me, my chatbot would respond that it did not have any location data available. When I increased it to 1.5, however, my chatbot's answer completely changed. It told me which restaurants that fit my criteria were within walking distance (something it created on its own), and even different activities that were near the restaurant. This allowed me to further demonstrate what I want LocalePal to do in the future, without actually having to code this process in.
Here is a short demonstration of my chatbot in action.
Next Steps
Even though this is all that is required for my project, I plan on continuing to work on this as a personal project. My first goal is to implement Yelp’s API, which provides up-to-date information on businesses across the world. Instead of the fake businesses I had the Erin bot create, I plan on having my chatbot recommend actual places to my users. I feel this will be a great next step in scalability, as I will then have more businesses to work with when I start developing an actual recommendation system. Currently, I only have users tell the bot what kind of food and activities they enjoy, and the bot randomly suggests a business that fits this criteria. In the future, I want to incorporate collaborative filtering to help hone in on what each user will enjoy, based on other users' similar preferences.
Conclusion
Throughout the process of creating LocalePal, I learned a multitude of ways to incorporate AI when creating a business. I now strongly feel that AI can especially help in the early developmental stage of businesses, where you can quickly make a working app, and build off it from there. These tools are easy to use, which allows anyone to create a proof of concept of their idea in a matter of hours. I am quite excited to see where this technology goes in the future, and how people will incorporate it into their everyday lives.