← Projects
  • Case study
  • AI strategy

Beauty Formula Decoder

An AI concept for Ulta Beauty

I led a 6-person team, created both of our AI concepts, chose Ulta, interviewed an Ulta executive and helped build the working prototype. Our class voted us #1 of 6 to present at Innovation Day.

Title slide: Beauty Formula Decoder, by Aahil Penwalla, Chandler Besse, Daniela Jimenez, Sara Pec, Shree Meher and Zachary Yujnovich, above Ulta Beauty Collection products on orange.
Our Innovation Day title slide. Most of the photos and illustrations in our presentation and poster were AI-generated, created to fit the theme and the concepts we wanted to convey. They are not photos of real Ulta stores, products in use, or customers.

01 The Problem

Skincare and makeup shoppers are buried in options. Searching for a fix turns up endless products and generic AI tips that never answer the real question: are these products safe to use together? Ulta has the products and the loyalty data to answer it, but nothing that guides a shopper through their own routine.

  • Presentation slide: The modern beauty pain point, with close-up photos of acne, forehead lines, eye wrinkles and smudged makeup.

    The pain point, from our Innovation Day presentation.

  • Presentation slide: Current solutions are overwhelming, with shoppers surrounded by products and tutorials, and a chatbot giving a generic list of products.

    Today's options: more products, more tutorials, generic AI lists.

02 My Process

6 steps, from picking the client to the prototype. Pick one.

Research deck slide: Ulta's Playing to Win strategy, with its winning aspiration, where to play, and how to win.Research deck slide: the same Playing to Win breakdown for competitor Sally Beauty.

Step 01 · My idea

Choosing the client

Our IPRO course asked for a generative AI strategy for a real company, and I proposed Ulta: a US beauty retailer with 1,300+ stores, a huge loyalty program and personalization as a stated goal. We mapped Ulta's strategy with the Playing to Win framework, then did the same for Sally Beauty, CVS/Walgreens and Bluemercury to see where AI could set Ulta apart. The interview later showed how big that loyalty program is: about 43–44 million members, driving about 95% of sales.

  • Playing to Win
  • Competitive analysis

03 The Solution

The Beauty Formula Decoder: a shopper uploads their routine or scans any product, from Ulta or anywhere else. The AI checks the ingredients for conflicts, such as retinol with vitamin C, suggests the order to apply them, and builds a bundle of products that work together, with a plain reason for every pick and a beauty advisor to confirm it.

  • Presentation slide: What's the future, showing a lighting booth in an Ulta store, a purple consultation room with lit mirrors, and two shoppers browsing makeup.

    Our in-store vision: a lighting booth and consultation space for color and skin checks.

  • Presentation slide: Why it matters, with customer impact, business impact, validation and governance.

    Why it matters.

  • Presentation slide: AI transparency builds trust, with our sample AI disclosure.

    The AI disclosure shoppers would see.

04 The Outcome

In the Innovation Day auditions, our class voted our team #1 of 6, and only two teams were picked. We then presented the Decoder at Illinois Tech Innovation Day with our research deck, poster and working prototype. I wrote the content for our final presentation slides. It taught me that the strongest AI idea is the one people can trust, and that one interview can change a whole pitch.

From Innovation Day

  • Our full Innovation Day poster.

    Our poster

What I'd change: Test with real shoppers and Ulta beauty advisors instead of relying on secondary research and AI stand-in feedback. Next, I'm designing a mobile version of the prototype on my own.

← Back to all projectsWork with me →