Overview
I enrolled in a 9-week online course, "Artificial Intelligence: Business Strategies and Applications" with UC Berkeley Executive Education, which required the completion of a final capstone project. The capstone project required students to "develop and refine a Business Case and Plan for a project or initiative at your own organization aimed toward the utilization, advancement, or leveraging of AI to transform one or more aspects of the business." Since I was not employed during the time I was enrolled in this course, I decided to develop a business case for my next biggest and most time-consuming challenge at the time - weekly meal planning.
Want to jump straight into the presentation?
Challenge
The capstone project's prompt outlined specific parameters to work within. The prompt asked for the following slides (get ready, the prompt is quite lengthy!):
- Slide 1: Summarize the AI application in 2-3 sentences. Focus on the problem. Briefly describe the industry context.
- Slide 2: Strategy: How will your application generate business value? Describe if it's an improvement over an existing technology or if it solves an entirely new problem. Does this application create/help you maintain a competitive advantage? How? Are you a first-mover or a second-mover? Is your application financially viable? If it requires investment, provide a rough estimate of the expected ROI (focus on the different components of investment and return rather than the exact numbers).
- Slide 3: What metrics will you use to evaluate the success of your application? Describe how this metric is connected to the underlying problem need. Describe how you would interpret particular values of this metric concerning actionability (e.g., What action would you take if your metric rose above some threshold value or below some floor?).
- Slide 4: Technology: What AI/ML technique will you use for your application? Briefly describe the technology to the extent we covered it in Modules 2-5 (e.g., Supervised/Unsupervised/Reinforcement learning, does it involve neural networks, vision, speech). Will you adopt a readily available algorithm, or do you have to develop your own? How will you train the algorithm, and what are the training objectives?
- Slide 5: Data: Most of today's AI applications are data-hungry. If yours isn't, describe why not. If yes, how will you get the data? Do you already have access to it? Who owns the data? Where does the data reside in your information systems? If you don't have the data yet, how will you obtain/generate it? To the degree that integration is required, how will you link data across systems?
- Slide 6: (Modules 7-8) Humans and AI: What organizational and human factors are crucial for the success of your project? Describe any organizational change or acceptance that your project necessitates. How will you make the above happen? Does your AI application interact with humans within or outside the organization? What steps are required to make the interaction smooth? Discuss any potential moral or ethical implications that your project may have.
- Slide 7: Measurement: Design an experiment to test the success of your AI application. The experiment should include the following: Experimental design (e.g., A/B test with randomization, pre-post test, or some other design), which success metrics you're measuring and what you're comparing to, and your test period/population.
- Slide 8 (Optional): Experimental results and analysis. If feasible, we would love to hear from you about the deployment of this experiment. This is not required but would offer some opportunities for reflection and feedback.
I kindly opted out of slide 8. 🙂
Approach
As a self-proclaimed foodie who holds myself responsible for weekly food planning and prefers home-cooked dinners for the majority of the week, I started by listing my own requirements and processes:
- Review my recipe bookmarks to see if any of them re-pique my interest
- See what recipes are trending on social media
- Review my own pictures of food to remind myself of what worked well
- Make sure I don't repeat the same recipe too frequently to avoid dish-fatigue, yet remember to bring back the familiar tried and true
- Find dishes that incorporate seasonal ingredients
- Make sure the recipes I select are also kid-friendly
I then researched already existing apps to see what's out there, such as Mealime and Platejoy Health. I also looked up pricing from popular paid recipe websites who also have their subscriber rates documented online, like NY Times Cooking and Bon Appetit.
I went down rabbit holes on Hugging Face to search for existing datasets and data models. I came across:
- Some work has already been done by the New York Times for extracting ingredient information as described in this article. This team created datasets and trained models: https://huggingface.co/datasets/napsternxg/nyt_ingredients, which would be great for generating shopping lists.
- An existing dataset and model for computer vision recognizing food and dishes was also available: https://huggingface.co/nateraw/food
- A paper that acknowledges that computer vision for food is still underdeveloped, and a team started to establish some foundations: Bossard, Lukas, Matthieu Guillaumin, and Luc Van Gool. "Food-101–mining discriminative components with random forests." Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part VI 13. Springer International Publishing, 2014.
Gaps I've noticed during my research:
- There's a massive gap when it comes to computer vision recognizing food and dishes.
- There's currently not much compilation of data across multiple data sources to facilitate meal planning.
- Social media isn't widely leveraged as a data source for meal planning among existing apps.
I made sure these areas were my focal points to make the SaaS AI-driven meal planning experience unique and worthwhile investing!
Next, I created a spreadsheet to list headcount/salary, overhead costs, and projected paid subscribers over 5 years to pull together some imaginary financials. I also made the assumption that the app was not receiving any advertising and sponsorship revenue to keep the numbers more realistic and conservative.
Since the prompt asked quite a few questions, my 7 core slides were text-heavy. However, considering I wouldn't be speaking to the slides, text was the only way to explain myself. I also decided to use Gamma, a presentation app powered by AI, to help design my slides in a more visual way to help offset my walls of text.
Solution
After extensive research, planning, and editing, my business case delivers a SaaS application designed to offer Meal Planning for Busy Foodies. It's an AI-powered app that understands the palate of each individual foodie through their food photography, favorite social media accounts, links to loved recipes, and food preferences and aversions. The app understands the foodie's cooking style and dietary needs to provide personalized recommendations, allowing them to explore new culinary experiences while maintaining a balance with their regular weekly rotation. The application generates weekly shopping lists and provides the option to purchase groceries through a delivery service, making it easy for foodies to stock up on ingredients to prepare their personalized meal plans.
Please see below for the slide presentation. If it's not rendering very well, you may also access it using this link.
Outcome
I put a lot of effort into my presentation, but I still felt uncertain upon submitting my capstone project. Was I too wordy in some places? Did I miss the mark on some questions? Did I provide insufficient data? Self-doubt was certainly settling in, but thankfully, I had a busy schedule to deter myself from dwelling on this.
Fortunately, in the end, I received a glowing review from the course facilitator and was able to successfully pass and complete this AI course!
Hello Grace,
Excellent presentation! It was one of the best deliveries of the course, and I found it very well-structured. I loved seeing that you also utilized other AI tools, such as Gamma.app, to enhance your delivery. I really liked your proposal, even though the reality it addresses is somewhat broad. You did an excellent job identifying the use cases and the technology used to accompany them. You presented a clear business case, provided useful metrics on the potential impact, and offered a comparison with other applications, as well as metrics to measure the success of your application. I really don't have much to add. Great work!
Lessons Learned
The Capstone project was both challenging and rewarding. It was genuinely enjoyable to pull together a business case on a topic I'm passionate about and believe in. Thankfully, I had the time to focus intently on this project. If I were juggling it with a full-time job like most other students, I likely would have struggled to give it my full effort and attention to detail.
Working on this capstone project was eye-opening, and I've gained invaluable experience that I will undoubtedly carry forward into future professional opportunities. Throughout the process, I gained deeper insights into the intricacies of developing an AI solution, from problem framing and data needs to implementation considerations and success metrics.
Perhaps most significantly, this course underscored the inextricable link between AI and digital transformation. As businesses strive to stay competitive and innovative, leveraging AI capabilities will be vital for driving strategic transformation initiatives. The skills and knowledge I've acquired in this course have equipped me with a solid foundation to effectively navigate the complexities of AI-driven digital transformation.
I now feel better prepared to identify opportunities where AI can create value, streamline processes, and unlock new growth avenues. Additionally, I have a deeper appreciation for the organizational shifts and change management aspects that must accompany the adoption of disruptive technologies like AI. This holistic understanding will be invaluable as I contribute to shaping and executing digital transformation strategies that harmonize people, processes, and technology for long-term success.