How AI is Changing Grocery Shopping Forever
AI is in the news a lot today — here’s how it’s being used to wow customers and profitably grow businesses.
With the emergence and viral growth of ChatGPT, generative artificial intelligence (Gen-AI) and machine learning (ML) are making headlines. We’re already seeing their impact in software engineering (where they’re automating code production and editing), customer operations (where chatbots are quickly and effectively responding to consumers) and content creation (where AI is generating new, original creative). These are just a few examples.
In online grocery, many AI applications in search ranking, product recommendations, inventory management and targeted advertising have been employed for years. But with recent advances, and since food is one of the last sectors to go online, the industry is just getting started. With an AI-first approach, grocers now have the opportunity to deliver a more personal, affordable and sustainable experience to their customers.
How AI Has Been Built into Hungryroot’s DNA
Hungryroot started as a packaged food company selling a limited number of products but when our customers asked us to expand our offering, we realized that we’d need to offer something really unique to compete. In 2019, we pivoted into an AI-driven, truly personal shopping experience. Today, our AI selects 70% of what our customers buy, saving them time and worry, introducing them to new foods and helping them accomplish their health objectives better than they could on their own.
Grocery shopping is hard — the average store carries more than 35,000 items, many of which have long ingredient labels. The products are spread out across expansive spaces of 50,000+ square feet, and unlike online shopping, the store’s layout cannot be personalized to each customer’s specific needs and preferences. It’s also hard because we use the majority of our groceries for cooking meals, and yet today’s grocery stores aren’t designed around recipes to make it easy to do so. The result is that the average American makes two trips to the store each week and spends 3+ hours a week planning and shopping for their food. At Hungryroot, our customers trust us (and specifically, our AI) to do this work for them.
Here’s how we make that possible: First, our entire experience is designed around building trust and getting to know our customers. Before their first purchase, our customers fill out a 5 minute quiz, telling us about who they are and what they’re looking for in their food. How do they eat throughout their day? Are they sticking to a particular diet? What’s most important to them in selecting their groceries?
With each customer, we collect more than 100 explicit data points, far more than any other grocer. We also leverage countless implicit data signals: their ordering behavior, budget and how they edit our suggestions, for example. Using these data points, our large-scale constraint optimization algorithm fills customers’ weekly grocery cart leveraging our in-house inventory of modern food brands, our own private label products and over 6,000 proprietary recipes. Using data from over 7 million orders and over a hundred million groceries sold, we train our models to make predictions about what products and recipes will not only satisfy their dietary needs and preferences (and those of their families), but most importantly, foods that they’ll love, and our predictions are always getting better.
It’s like our AI is your own personal grocer, considering your preferences, health objectives, lifestyle and budget as it builds and then delivers the perfect grocery cart for you. It even considers how and when you eat, crafting simple recipes for your busy weeknight and on-the-go snacks for your kids. We’ve spent years designing an engaging way to accomplish this so that we delight our customers, save them time, and never ask for information that is either irrelevant or we can infer ourselves.
Knowing our customers so well also allows our algorithm to help customers shop more cost-effectively, which is especially valuable in today’s inflationary environment. For example, when protein prices were impacted by inflation earlier this year, our AI was able to swap chicken it knows the customer likes into their weekly delivery instead of beef, allowing us to fulfill the customer’s needs based on their preferences at no additional cost. We don’t have to burden our customers with this decision as it’s already built into our model and understanding of their preferences. And by arming our customers with a plan that includes recipes every week to guide them in how to use everything we send, we significantly reduce their food waste, resulting in even more cost savings.
AI is not only core to the personalized experience we deliver to our customers, it’s also integrated into our business operations. Because our algorithms get to know our customers well, we can run a streamlined fulfillment operation that doesn’t store (and ultimately throw away) unwanted products. We also use AI to continuously optimize orders throughout the week, so that we can effectively manage inventory and produce less waste, resulting in facility food waste of 2%, 80% less than the industry norm.
By flipping the traditional shopping experience on its head in this way — starting with a full cart, instead of an empty one — we leverage AI to be truly on our customers’ side. From day one, we emphasize that the data our customers share is explicitly used to build their cart and we reinforce this message throughout their interactions with us. This builds trust with them, leading to a willingness to share even more information with us, in stark contrast to the 40% of Americans who don’t trust companies to ethically use their data. The secret sauce isn’t the AI itself — it’s what it enables, which is a deeply personalized, relevant, helpful experience for customers.
The Future of AI at Hungryroot
AI will continue to positively impact our customer experience and our business over time. As AI evolves, we anticipate integrating large language models and our food-specific AI more directly into the customer experience. For example, after a customer receives their weekly Hungryroot box recommendations in a text message, they could reply that they’re not in the mood for soup this week since it’s unseasonably hot, and the Hungryroot chatbot could offer to swap in a grain bowl in a revised grocery cart. We’re also exploring how AI can optimize when and how we ask customers questions to gather impactful explicit data and enhance our recommendations. And as our operation grows to offer a larger assortment of groceries, AI will continue to be vital in sorting through many more data points.
Beyond our customers and the business, we also see great potential for AI to positively impact the lives of our team members. For our marketing and culinary teams we’re seeing AI impact content and recipe idea generation. In Customer Care, we’re excited about how AI can support our care agents with macro generation and issue analysis. Using AI to draft job listings, develop policies, and take meeting notes are examples of day-to-day impacts that add up to meaningful improvements for our team members.
AI has enormous potential to revolutionize the grocery industry. When it comes to leveraging AI to deepen and improve our customers’ relationships with food and optimize our business, we’re just getting started.
