Personalized Salads, Predictive Trends: Inside Just Salad’s AI Strategy
HT spoke with Just Salad CTO Matt Silverman about what the fast-casual brand has learned from its AI-powered order personalization tool, launched in January.
How It Works
In the Just Salad mobile app, users can navigate to the Salad AI section, complete a short preference survey, and instantly receive personalized order recommendations based on their flavor profile.
Silverman, who had worked at the salad brand in 2009, rejoined the company in 2023 after working at an AI blockchain startup. “I’ve been following AI closely, and we recognized that it is really important to us and for me to keep us at the cutting edge. Where can AI help our corporate employees? How can it help our customers? We’ve been tackling it in those two realms,” he said.
Early last year, Silverman added an AI director to focus on those two buckets: guest- and employee-facing possibilities.
To start, Silverman and his team wanted to create a guest-facing AI that was "insulated from a lot of the problems" that exist today with large language models (LLMs). “They can make mistakes, and particularly when you're allowing guests to put in pure text into a chatbot,” he explained. Plus, they wanted to avoid bad actors "jailbreaking" (bypassing the security guardrails) of the AI model.
The team intentionally avoided using a traditional text-input chatbot. “If you output raw text from AI, you don't have tight control over what it says. We were looking at doing it in a more controlled way but still using AI under the hood to power decisions based off the menu. We control what values (guests) could select. We looked at what flavors people are most interested in. It’s more a templated approach; we're taking in what they're saying but using AI under the hood to match their flavor preferences with our entire menu offering and every ingredient we offer.”
Model Agility and Prompt Engineering
The world of major LLMs and AI is quickly evolving. “What is fascinating … is as you're developing it, a new model could come out that's much better,” he said.
The team invested heavily in prompt engineering—crafting the inputs and context fed to the AI, which included the entire Just Salad menu. “We experimented with all the top models before we launched, and we found one that we needed the balance of what's the quickest and the most accurate.”
“Ultimately, it was ChatGPT-4o that worked best for us at that time. But if we're developing it today, we may have selected a different one,” Silverman said.
Lessons Learned
In the six months since launch, Just Salad has logged tens of thousands of interactions with the tool. “When we went into it, we didn’t really know what to expect. We didn’t know how many people were going to use it. We just really went into knowing this was something cool. We want to be on the cutting edge. (Our thoughts were) let’s see what happens.”
The Data Says...
“We know based what customers are ordering, but we now see people explicitly select over and over their preference. It helps shape tons of decisions we are making, including our culinary, menu development, customer service—all that stuff,” Silverman explained.
The AI director generates regular reports—both 30-day and cumulative views—highlighting customer preferences and trends. These reports are shared with the full management team, including culinary.
Protein-powered meals have been popular, along with customized toppings that combine savory and crunchy, he said. “A lot of what guests are selecting track with kind of how our menu has naturally developed over time, because we listened to our guest feedback,” Silverman said.
The data is being used to inform menu decisions.
For example, if Just Salad was choosing between a couple of new seasonal salads, the data helps guide decisions. “If one of them checks more boxes with what we see the trends are, with flavors, that (data) would help steer us in the optimal direction,” he said.
While the data hasn’t yet influenced inventory planning, Silverman sees it as a promising area for future AI applications. “It's hard to get any accurate prediction for something like inventory (from the Salad AI data.) But Inventory is a great example of an area that's kind of ripe for innovation with AI. "
AI for the Employee Experience
Silverman and his team are also exploring how AI can improve the employee experience, especially at the corporate level. “With a lot of people, there's some fear about AI replacing jobs; we see it very differently as we stay at the cutting edge of what AI can do. What we see it doing is allowing people to focus on what they love most about their job and helping take care of some of that kind of more repetitive parts,” he explained.
At the corporate level, all employees using Google Workspace now have access to Gemini, Google’s AI assistant. Training sessions are personalized to each individual or group within the company to help them find their best use of AI. Some may use it to organize documents, generate checklists, or streamline routine tasks. “Everyone at our corporate office is using Gemini, and we're kind of at the exploratory stages right now on how we can use it. How can AI improve or add an extra check to some process we're already doing,” Silverman said.
Looking Ahead
Looking to the future, Silverman said to expect Just Salad to stay on the forefront of using AI to improve the guest experience. “Anytime we're looking at what can we do to innovate the guest experience, it's safe to say that we're going to continue to stay at the cutting edge of what's possible with AI,” he said.