Capstone Project: Create an AI Product Business Proposal

Congratulations! You are very nearly done with this ND. You have learned so much in this program, and this project is an opportunity to put your new knowledge to work.

In this project, you'll develop a business proposal for an AI product. There are several important aspects of product development you'll need to think about and describe, and they are exactly the topics we've covered in this course. Here, we'll walk through what should be included in the proposal. You'll also find instructions in the Capsone Project Starter File, linked at the bottom of the page. You need to answer the questions in this file to complete the project.

This project is open-ended in that you can propose any product in any industry that you want! If you're stuck, feel free to think about and research one of the use cases we've already discussed in this course. As a tip, you may find that your strongest ideas come from business arenas you are interested in and may already know a lot about.

The Business Goal

You'll need to describe what the product is, and how it will provide value to the business. It's important in this section to describe exactly what the product will do and why/how this helps the business. Focus on linking the AI/ML task to business goals such as increasing revenue or customer happiness.

Success Metrics

You'll also need to describe how you'll know whether the product is successful. Think about measurable, predictive, comparable, and benchmarked metrics for business success.

Data

You should carefully consider how you will acquire the data to train your model, and issues that may arise during data collection. Important considerations include: buying data vs. collecting it, personally identifying information (PII) and data sensitivity, cost, and whether data will be continuously available or acquired in one large batch (and need to be refreshed).

Model

When thinking about how you will build the model, will you use an in-house data science team because none of the out of the box platforms have your use case, or because you want the ability to control a particular aspect of your model? Frequently, when you use an external platform to build and host the model, their terms of service will require you to give them access to your data. Will this be an issue?

Minimum Viable Product (MVP)

You'll want to think about what your product is. What does it look like? Who uses it, and how do they use it? How will you actually build it?

Post-MVP-Deployment

Finally, what happens after launch? How do you ensure your product continues to perform well? Who will monitor it, how will they monitor it, and how often?