Sprouts Put Agentic Intelligence in 55 Stores. Here’s What Happened Next.

Author: Microsoft Cloud · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Sprouts Put Agentic Intelligence in 55 Stores. Here’s What Happened Next. · Channel: Microsoft Cloud · 19:47 · Watch on YouTube

Topics: AI agents, Enterprise AI applications, AI evaluation and benchmarks

Summary

Sprouts Farmers Market developed Store Assist Intelligence to help team members answer customer questions and spend less time on repetitive, non-value-added work. The company began with a two-store test, prioritized accuracy and employee feedback, and expanded the pilot to about 55 stores, with a national rollout planned for later in the year. Adoption has depended on useful tools, engaged store leadership, and business ownership, while AI-supported workflows are intended to redirect labor toward customer experience rather than simply reduce staffing. Sprouts plans to keep improving the assistant using candid feedback and explore ways to support product knowledge, inventory reliability, and personalized value.

Key points

Sprouts’ mission is to help people live and eat better, and the AI effort was shaped around that goal rather than technology for its own sake.
Store Assist Intelligence is designed to assist team members, not replace their judgment or expertise.
The assistant aims to reduce time spent searching for resources and handling repetitive tasks so employees can engage more with customers.
Sprouts involved team members from the start and used their feedback to identify friction points and guide development.
The pilot grew from an initial test in two stores to about 55 stores; a national rollout was planned for later that year.
Accuracy and trust were priorities, so Sprouts tested the assistant in real stores before expanding.
Store and district leadership, a clear vision for employees, and tools that address real needs were important to adoption.
AI-supported workflows can compress multistep processes; Sprouts intends to repurpose labor toward customer experience rather than focus on removing labor.
One example involved helping a customer choose a wine pairing for a steak, giving a team member confidence to assist across departments.
Sprouts encouraged team members to give negative feedback as well as positive feedback to help improve the assistant.

Chapters

Start	Chapter	Summary
00:00	Sprouts’ mission and the opportunity for AI assistance	Sprouts explains how its mission to help people eat and live better led to an AI assistant aimed at freeing team members to help customers.
03:11	Building with team members and starting small	The team describes gathering frontline feedback, prioritizing accuracy, and expanding from two stores to a 55-store pilot.
05:47	Driving adoption and coordinating stakeholders	Sprouts discusses the role of store leadership, change management, business ownership, and regular communication in making the tool useful.
10:26	Business outcomes and repurposing work	The company focuses on delivering expected outcomes, giving employees product and process information, and redirecting labor toward customer experience.
13:34	A customer-service breakthrough and future plans	A wine-pairing example shows how the assistant can help employees serve customers across departments, while future ideas include curated products and personalized value.
17:12	Feedback, change management, and a more human retail experience	Sprouts encourages critical feedback and emphasizes using AI to remove burdensome work so team members can focus on human connection with customers.

Quotes

we help people live and eat better. — 00:00

change doesn’t happen to people. It happens with people. — 03:42

You lose trust with the team members. You lose trust with the customer. — 05:16

Don’t be afraid of the thumbs down. — 17:43

this should make retail more human, not less human. — 18:46

Concepts and entities

Sprouts Farmers Market, Store Assist Intelligence, Evan Naylor, Microsoft, Agentic workflows, change management, real-time feedback

Full transcript (original language)

00:00 Hey, good afternoon. Brad Mark. I am leading our AI team for our retail and consumer goods industry. My name is Evan Naylor and I’m the VP of AI transformation at Sprouts Farmers Market. Fantastic. We’ll look forward to the discussion this afternoon. Maybe before we get started, we’d love for you to maybe just share a little bit about the Sprouts mission and your role. Yeah, so I mean, our mission pretty much runs through everything we do, including my role, which is we help people live and eat better.

00:32 And that’s what I think it’s great about coming from the business side of the house and not necessarily a technical background, is that I really keep our team focused on that vision of how do we help our customers live and eat better? How do we make sure we understand their preferences, what’s important to them, and not just start with tech for tech’s sake? And I think that’s really been important as we kind of go through the story today and talk about how that transformed some of the things we’re doing with our AI shopping assistant at Sprouts. I love that. And on that theme of coming from the business, you saw an

01:03 opportunity at the store level within your team members. Talk to us a little bit about that opportunity that you saw. Yeah, I think at the heart of our team members is they have a real passion for helping our customers live and eat better. And when they come in the store, a lot of customers are starting on a journey and they’re not really sure. How do I start this healthy living and eating journey? What’s the best right products for me? What’s my best first next move that I should take in order to start my journey? And so our team members really love

01:34 to interact with our customers, give them expertise, knowledge, give them their experience to help them start their journey. And so when we saw that, it was a problem for our team members. It was really about what’s getting in their way of being able to accomplish that. And we saw a lot of non-value added work, a lot of trying to gather resources from different places to be able to help them to address those customers’ needs. We’re just playing work getting in the way of their ability to interact and engage with our customers. And so when we talked about this idea of an AI team member assistant, it

02:07 was really about how do we leverage that in order to take all this non value added work off their plate and all these scenes that were interacting between the information that they were trying to get to and just eliminate all that so that they can spend more time engaging the customer and helping them in the first step or second step or even on a lifelong journey of healthy eating and living. So the solution you built is Store Assist Intelligence. It seems you were very purposeful with the word assist.

02:38 Yeah, it was a little bit about that. Yeah, well, our artificial sounds so menacing. So I think a lot of people are trying to get away from that word. But for us, it truly is about assistance. This is about how do we enable our team members to, again, get all the non value added work off their plate, focus on what they want to do, which is engage our customers and do things that really matter to them most. And so for us, it’s really about assistant. It’s not about removing their judgment from the equation or their expertise from the equation. It’s really about helping them connect in a more human way because

03:11 now they’re not bogged down by tasks or doing things that distract them from that human connection and that human interaction. So, AI has been around for a while, but I doesn’t solve everything. And you’ve been we’ve met a couple of times and you’ve shared with me that it’s really important to get real time feedback and talk to us a little bit about how you got real time feedback. Maybe how you adjusted the execution plan along the way. Yeah. So when we first started with our

03:42 team members, we started with our team members. And so I always have this saying that change doesn’t happen to people. It happens with people. And so we involve them right from the start. What are your biggest customer pain points and what’s not what’s getting in your way of helping you be able to do your job and the things that are most meaningful and impactful to your customers in the business. And by sitting and listening with our team members and starting to remove those friction points, it all started with what’s right for the customer and what’s right for the business and how do our team members drive that intersection.

04:13 It did not start with the technology and hey, everybody else is doing X and everybody else is doing Y and did you just read this article and this tech is really cool and it’d be really cool if we could do this. But what does that matter to our team and what does it matter to our customers and so saying true to that and building this with them was a key principle and priority of what we did from the start. Maybe you share a little bit about how fast how slow you went. How many stores how many team members a little bit around maybe the

04:44 operating models we can get a good understanding of the scale of Store Assist Intelligence. Yes, I think it’s and it was important for us to start actually start a little bit slower right and make sure that we had some good foundations in place. You know, I think I echo some of the comments from the main stage earlier where they were talking about sometimes you have to kind of slow down to speed up and you have to make sure that you have the right data in place. You have the right infrastructure in place to be able to help you get it right and get it accurate because the worst thing you can do is start to drive an AI assistant and it doesn’t

05:16 get it’s not accurate. It doesn’t pick up on the right data. It doesn’t give the right inference and then all of a sudden you lose trust. You lose trust with the team members. You lose trust with the customer. So for us it was really important that we drive accuracy and what we were building and so having those team members along that journey and starting with a small group of just 2 stores very small group get it right understand how it played out in the real world. Not just in a lab and then expand that to a few more stores and then now we’re in about a 55 store pilot

05:47 with a national road out coming later this year. I really like that approach and as we know customer service is so important and now you’re giving your team members more time to provide customer service. Talk to us a little bit about maybe the change management the adoption from your frontline team members and how how that’s progressed along the way. Yeah, I mean the first thing I’ll say about adoption, especially in a retail environment and multi unit is the leadership of the store and the

06:18 district leadership matters a lot. It’s really important and we have seen some divergence in that. So we would you know in some stores that we tested where the leadership maybe not have been involved. We saw a certain level of adoption and certain level of engagement and then we saw a couple of our stores with the leadership made it a priority to enable their team to help drive this and it just took off in the store. It impacted sales. It impacted shrink waste and they were really taking an active role in making sure that their team members knew how to use the technology understood

06:50 what it meant to them. So there’s also this factor of with adoption and change of if you don’t give people a vision of what they’re going to, they’re scared to move away from. Right and so I’m scared to move away from what I normally do because I’m really not sure what you’re moving me toward. And so if you paint a vision of here’s what this is going to enable you to do. Here’s how we see your role progressing. Here’s how you can engage more with your customers. You really give them that’s what what’s in it for me or the with them to be able to move to something and

07:21 so that’s a really critical part of adoption. The leadership is a really critical part of adoption and also building useful tools things that matter. So it’s really cool if the if the engineers want to go build all these bells and whistles and like hey we could do this and we could do that. But again how is it going to impact the team members? Does it make their life easier? Does it remove friction for them? Does it remove friction for their customer? And so I think there’s there’s several adoption points along the way but it’s got to be functional and useful. It has to be able to be leadership driven and leadership

07:51 has to get behind it. And when you comment to a combination of those things, that’s when you really start to see adoption tank off. When it comes to stakeholders and how you’re managing the operating model and how you over communicate throughout this whole process, maybe just a little bit deeper on stakeholder management and how often are you communicating on progress on this journey. Yeah, and there’s a there’s a lot of stakeholders in this journey, including Microsoft. We’ve worked

08:22 really closely with the teams at Microsoft and they’ve been great partners at helping us to move things along where maybe our third party build partner was a little bit technically stuck. You guys were able to bring in partners and help us kind of move forward. Our technical build partners really helping them understand what the business requirements what we were trying to accomplish. And you do run into those points where people try to say, well, you know, but if we could do these things instead, that would be really great. It’s like, that’s really great, but that’s not what our teams are asking for. And that’s not really going to be a

08:52 business viable or business impact solution. And so you have to constantly calibrate a little bit, you know, on the on the tech side. We actually went through and we started building out some more engineering capability after we started building with our third party build partner. So bringing some of those folks in and they became key stakeholders with our it and our engineering teams as well. Then you’ve got your HR partners to make sure that we understand the impact on people, the whole change management side. We’ve got our chief of stores who we meet with regularly in our stores

09:23 team members and getting them involved and really owning it. So as an AI transformation group, we kind of sit in the middle of the business across the enterprise. And so it’s really critically important that if you’re delivering AI or a new technology into a piece of the business, that they really take ownership of it. So if they don’t take ownership of it, you’re trying to kind of push it from outside that functional area. It becomes a little bit more difficult for adoption and for them to really, you know, take pride in what they’re what they’re doing. So a lot of stakeholders meet regularly with most of these people on a weekly

09:54 or biweekly basis to make sure we stay connected and also make sure we understand the problems that are encountering. What’s working, what’s not working, what did we think was going to work in a lab, but didn’t work in the real world? And how do we go back and solve for that problem? What do we think is going to be a quick adoption, but ends up not being a quick adoption and the team members didn’t like it at all. And so you constantly have to go back and calibrate with all these stakeholders to make sure that you’re building something that is useful and impactful. I’m sure while you’re managing this project, some of the stakeholders are also asking you to do even more

10:26 projects and more things. How are you managing kind of the supply and demand with AI being such a strong topic in today’s world right now? Yeah, I think it’s you really do have to be disciplined in a couple of things. You have to be disciplined in your focus. So let’s get out the door what we already agreed we were going to get out the door and not just get it out the door. But make sure it has the impact that it’s expected to have. There’s a lot of people want to deliver a technical technological solution and say, hey, we’re done. Technology’s there. Okay,

10:56 but it is actually achieving the result that we expected. And is it getting us the outcome that we intended to get? And so I think it’s important to continue to think about, you know, how you continue to drive towards that outcome versus just the technical delivery. And so that that became something that was really important to us. Obviously, don’t share any of the numbers, but just high level a little bit on the TCO, the ROI, the business case side of things. Just how do you and Sprouts kind of make sure you’re aligned as you’re

11:27 constantly evolving the solution? Yeah, I think when I look at some of the things that we developed, we kind of developed a suite of things that would be both things that drive sales. So helping with product knowledge and information. Sprouts opens a lot of stores every year. You know, we open about 10% of our network every year. And so when we do that, we have to go back and make sure that we are able to give our new team members in new areas, tools that help them ramp up fast. And so that was one of the impetuses behind building a tool that helps them have product knowledges

11:59 product knowledge, standard operating procedures, you know, mill planning and production guides all at their fingertips. In our business in the past, you had to go and sit down at a desktop to retrieve that information. And so any of this manual repeatable work or places that took me off the sales floor instead of keeping me on the sales floor to interact with customers. We really went after and attacked that work first because it keeps people engaged in a point where they can be engaged with customers. And then we also looked at certain workflows.

12:30 If I can take a workflow and say, look, this workflow typically takes 14 processes or 14 steps in the process. Well, then you ask yourself the question, why? Why is it taking 14 steps to do this? And oftentimes we find that actually, AI can come in and probably do about 10 of those steps and we can compress it down to four. And so we picked a couple of those high value use cases where we could really take a whole process, get that non value added work again off the team members plate and have them focus on the things where they matter most, which is their judgment, their

13:01 decision making when it comes to that and then their customer experience. And so by doing that, we were able to come up with some agentic workflow cases to get labor out of the stores and take that labor and readjust it towards customer experience. Because for us, customer experience is really a differentiator. So unlike other retailers or some other businesses, we’re not really interested in taking tons of labor out of the business, but we do want to repurpose that labor to make sure that it is meaningful and impactful for the customer experience. That’s great. Any successful project

13:34 has a breakthrough moment. I know you’re still a little early days and you’re not scaled out across every store today. But if you had to maybe mention one or two breakthrough moments so far, what comes top of mind? I would have to say I get excited, like super passionate about when you hear how a team member has finally connected the dots in the store and helped the customer and actually built a basket. So I love that there’s a couple of stories that we have where, you know, a non team member who is not working

14:07 in perishables, they weren’t in the meat department. But somebody had just purchased like a really beautiful, nice, marbled steak and the team member was passing by and they said, hey, would you mind recommending? What would be a good wine to go with this? And, you know, if you’re not somebody who really understands wine, that’s probably a pretty scary question. But this person was like, you know, I’m not sure, but let me check. And so they pulled up the application that we created and they punched in, you know, what type of steak it was. And they asked the customer a couple more questions about the event they were having and said, help me recommend a wine pairing that would go with this.

14:37 And it recommended three choices for the customer. So the team member was much more confident in their ability to approach that situation, guide the customer over to the wine and find a wine pairing that went with that. And we hear examples of that all the time where our team members are actually able to start building baskets because they understand the different categories and they can help the customer actually shop across the store. Whereas in the past, maybe they stayed a little bit closer to their area where they were a little bit more familiar with the product mix. Now they have that confidence to go and help the customer shop across the

15:08 store. And that’s exactly what we want. Right now, all of a sudden, you know, AI is helping us become more engaging and have better human connection, create that repeat business and that repeat traffic. I’m getting thirsty and hungry after that example. That’s great. If we think about the future, we think about what’s next. Assuming that this is going to be super successful, that you’re going to roll it out to your entire stores

15:39 nationwide. What do you think is next? Like how can you build upon this success and maybe some things that you feel are going to be some strong additives to Store Assist Intelligence? Yeah, I think there’s always the going back to what the customer prefers and the customer wants. I mean, for Sprouts, we turn a lot of our inventory, we turn a lot of our assortment, I should say, every year. And so we do that because we have to keep a very curated shelf. People come to us because they want to know what’s new and next in healthy eating and living.

16:10 And they expect us to be the authority in that. And so, again, that product knowledge from the Stores Assisted Intelligence helps arm our team members with knowledge even on emerging brands, emerging products. But then it’s also how do we make sure that we keep a curated shelf? How can it help power that? How can it help power reliability in our inventory? I think that’s another really important for one. How can we make sure that we’re giving the right value and affordability through right pricing and promo? It’s important to have the product in stock. It’s important to have a promo. And also, is it personalized

16:40 enough, like we heard earlier this morning, that that value is actually landing on something that somebody wants, right? And so it’s good to put promos out there. It’s good to have great pricing and affordability. But if you can actually show that what you’re doing is personalized to me and you know how I shop, then you’re going to show up for me in a very unique way. So those are things that we want to continue to build out. That’s great. Maybe last question before closing comments. Getting your end customer, your end consumers, direct feedback sometimes

17:12 is hard in this environment, but it’s super critical to get that real time feedback. How are you looking to continue to evolve that digital loop back so you can continue to be the sprouts that you are in the culture that you’re delivering to your team members? Yeah, I think it’s really interesting because when we started measuring all the different things that you can measure with AI, how’s the model performing, what’s the positive and negative feedback, and everybody always wants the positive feedback.

17:43 Hey, we gave the answer to the team member. The team member gave it a big old thumbs up and we all feel great about ourselves. We actually did something a little bit different. We actually encouraged and solicited team members to give the big old thumbs down because we wanted people to tell us when things weren’t working. We wanted them to be more adamant about, “Hey, this gave me a good enough answer, but it wasn’t exactly like I wanted, so I wish it would be better.” We gave them the carte blanche and we gave them the freedom to go in and give us some critical feedback there so we could improve the model and make it better.

18:14 Of course, now all that’s built into the way we grasp that feedback and are able to turn it around into useful features for our team members. I’d say the same thing with customers. Don’t be afraid of the thumbs down. It’s okay. You’ll get more feedback that way if you encourage that feedback and you’ll get more useful feedback to help improve your products. That’s great advice. I know we have a minute left. Any closing thoughts or best practices to share with the group here today? One thing that we talk a lot about is

18:46 how do you…this is the change management aspect. In a world of AI, sometimes it’s scary to people. They worry about what it’s going to do to my job and how is it going to impact me. Our heart is really that this should make retail more human, not less human. We want to continue to figure out ways that we can use the AI that we’re giving to take this…I want to reiterate this because it’s such a key important part… Is to take non-value added work away from team members so they can really focus in on taking care of our customers.

19:16 I think that you’re going to see people who are inclined towards that engagement with their customers are going to leverage AI that way to make sure that AI feels a lot more human than you would expect it to be. It’s about connection. It’s about engagement. It’s about giving people that trust and expertise. You can’t do that when you’re not on the sales floor because you’re off doing other things or you’re in the back room trying to do things that’s just bogging you down. We want to lift those weights off of our team members and make sure that they’re fully engaged with their customers. Thank you so much for sharing this

19:47 story. Just really exciting and look forward to hearing all the future success. Please join me in thanking Evan for his time today and a good story. Thank you.