Born in an earlier gold rush, San Francisco-based Levi Strauss & Co.’s headquarters are just a stone’s throw from many of the companies at the center of today’s AI gold rush, like OpenAI and Anthropic.
And the $6 billion apparel company is working fast to take advantage of what the latest wave of AI tech can do, whether it’s recommending outfits to shoppers using the Levi’s mobile app, or helping streamline the process of processing incoming wholesale orders.
At the helm of the 173-year old company’s digital strategy is Jason Gowans, Chief Digital & Technology Officer, who joined the company three years ago, not long after the launch of ChatGPT. “What was happening here was something exponential, and we needed to move quickly,” he says.
We recently caught up with Gowans to talk about his strategy, and some early AI case studies. LS&Co. operates in roughly 120 countries, with approximately 3,300 retail stores and 19,000 employees worldwide.
• • •
You began working at LS&Co. a few months after OpenAI introduced ChatGPT. How did you tackle your new assignment in the midst of a new technology that offered you the opportunity to fundamentally rethink how LS&Co operates its business?
When I joined LS&Co, I really had two mandates. The first was to reignite growth in our e-commerce channels; the company had and still has a stated ambition to 3X our e-commerce revenue, and that’s all part of our direct-to-consumer (DTC) first ambition.
The second mandate was what we internally call digital transformation. Most concretely, if you think about Levi’s and you think that from the moment you imagine a product to the moment it shows up on the shelves — that’s our go-to-market process. As we make that shift to being largely a DTC-first company, one of the things that has to happen is…we’ve got to get a whole lot faster. When you think about how we design products, how we forecast demand and supply, how we allocate our products, and how we price our products, you can imagine that we are using obviously technology and AI to drive a lot of the decision-making in those processes. So those are the two mandates just to set the stage coming into Levi’s.
Given those two mandates, what are you focused on?
Job one on e-commerce was really a diagnostic of, let’s figure out where are we with e-commerce, what’s happening with the consumer experience? …We really went through a deep diagnostic dive of what was happening with that particular channel, and thus where we had opportunity. From that emerged three pillars of our strategy. Number one: fix the fundamentals. Number two: evolve new assortment. Number three: build towards creating what we call a digital flagship experience.
…Our results show that we have been very successful with that strategy. On the digital transformation side, I’ve been doing machine learning for a really long time. Prior to Levi’s, I was with Nordstrom for 10 years.
I was the company’s first-ever director of data science, and was responsible for building out the company’s product recommendation engine. So…coming into Levi’s, I was pleasantly surprised to find there was actually a lot of good work in areas like price optimization, good use of machine learning, and in demand forecasting. So a lot of it was really just refining the strategy, and finding opportunities to improve the core solution, but then also drive deeper adoption within the teams that were using these at LS&Co., like the merchants, the planners, etc.
…We understood early on with GenAI, and with OpenAI in particular…that what was happening here was something exponential, and we needed to move quickly.
…For us at Levi’s, we understood early on with GenAI, and with OpenAI in particular, and subsequently all the other models, that what was happening here was something exponential, and we needed to move quickly. So as early as 2024, we were already using Microsoft Copilot, which of course is using ChatGPT under the hood.
Last year… we went from individuals using tools like Copilot, ChatGPT and Gemini to ask questions, do research, and improve their individual tasks, [and] transitioned towards scaling our agentic work and rolling out agents. In the past year alone, we’ve since deployed more than 800 agents across the enterprise in pretty much every single function in the company, and where we see this going is really this notion of a hybrid workforce.
One of the things we did last year was with Harvard Business School and Microsoft. We became part of this cohort called the Frontier Firm AI Initiative that is tasked with figuring out how do enterprises… responsibly adopt these capabilities and embed them in the workforce successfully.
…What are the opportunities to reimagine workflows and automate the execution of large bodies of work? What are the human skills that you need to operate this new hybrid workforce, where your workforce is a combination of humans and AI agents? That really is the journey that we are on.
Can you give our readers two or three examples of what you consider to be a high-impact generative AI use case?
A good example for us would be an agent that we call Stitch. Stitch is an agent that we roll out to the stores. The core premise here is as you think about [the] stylists — the folks who work in the stores — selling denim and representing denim is a complex business. When you think about all the different fits, sizes, and colors, you can imagine that there are lots of questions that customers have about buying a pair of jeans in our stores. When you also think about the…back-of-house tasks that a stylist has to perform in the store, [those can involve] processing goods that are arriving at the store, filling online orders, or even thinking about how [they] submit a time off request. There are lots of questions that a stylist might have in the course of doing their shift. And so Stitch was really built to eliminate that moment where you have to say to a customer, “I don’t know.”
… Stitch was born out of a hackathon. One of the folks who actually conceived of this and subsequently built it — his name is Michael Buchanan — he actually worked in a store for a number of years, he had gone through our own internal machine learning boot camp, he’d accumulated all of these skills, and now he was part of the data science and analytics team. So when we had a hackathon, his idea was Stitch.
We rolled this out to a small pilot set of stores, [and] we scaled this in Q4 of 2025 to more stores in the U.S. One of the interesting things that we found is that the stores that make use of Stitch [have] seen an eight point rise in consumer satisfaction, [compared with] stores that don’t yet have Stitch…
Did Michael Buchanan get a raise?
You’ve done your research. He was well celebrated.
Outfitting is a feature in the mobile app that was built by your data science and engineering teams. Can you tell me about that one?
…On Levi.com, and really on any e-commerce site, there’s lots of personalization that is happening in the course of the consumer experience. If we think about product recommendations, if we think about search, if we think about navigation and, of course, if we think about in Levi’s case in particular, Outfitting. And so Levi’s [has] built our own personalization platform. We call it “tailored AI,” and it performs a variety of functions — product recommendations being an example, size guidance being another example…
Outfitting is a core way that we try to introduce our consumers to the fact that we sell more than just jeans.
Our stated goal of being the best denim lifestyle retailer in the world. What you didn’t hear me say is we want to be the best at selling jeans; what you heard me say is we want to be the best denim lifestyle retailer. For us, that definitely means always being an amazing place to buy jeans, but also being a great place to buy tops — jackets, shirts, sweaters… Outfitting is a core way that we try to introduce our consumers to the fact that we sell more than just jeans. Outfitting is a perfect example where we are using AI to help the consumer in their journey, and to look great in not just our jeans but also our tops.
Can I get one more example of a high-impact generative AI use case?
Like I said, we have more than 800 deployed at this point, so they are really everywhere.
…Our business is roughly 50 percent direct-to-consumer and 50 percent wholesale. …When you think about how we sell to those wholesale partners, you can imagine most of those orders are automatic, meaning we get an automatic transmission, and we assist them to automatically process those orders, and we send the jeans to the retail partner.
But there is also a large amount of small retailers that we do business with. Levi’s is in more than 100 countries, [with] more than 50,000 points of distribution. These smaller retailers…don’t always have these automated capabilities, and so because of that, we received a lot of manual orders. These might literally be a PDF or an email that arrives in our inbox that says, “Hey, I want to buy X amount of jeans that I can sell in my store.”
Last year, the team built an agent that can automatically process these manual orders. …Now, our wholesale partners are receiving their orders more quickly, because these orders are being processed more quickly and our folks are not having to do manual entry anymore. You can imagine that there are hundreds of examples like that within the company, where we’ve taken processes that were largely data entry [and] manual processing, and we’ve since automated those.
…One of the primary considerations we think about is, does this potential use of AI elevate the human in the loop?
What is the criteria you use to prioritize your projects? What does an AI project have to have to get the green light to go ahead with building it?
Aside from whether or not there is a financial benefit, one of the primary considerations we think about is, does this potential use of AI elevate the human in the loop? If you think about what humans are really good at it’s judgment, it’s creating — especially if you think about Levi’s, it’s discerning taste and trends and what we want to put out into the world. …You can imagine in design… we can really think about how we can use AI to alleviate things like data entry, so they can spend a whole lot more time designing products.
I have heard that generative AI, because it contributes to the output and performance of individuals as well as groups, is having an impact on performance reviews. Are you experiencing that?
If you mean specifically, is AI generating performance reviews, the answer is no. If you are asking, in the course of a performance review, are we discussing with a given individual the extent to which they are using AI, yes. I have conversations with my own boss about my own performance, and one of the questions is, tell me about your progress with AI. So those sort of conversations are happening, but there is no scenario here where a machine is spitting out performance reviews. These are people to people conversations.
…In the course of setting our objectives for this year, we rolled out an agent called Objective AI, and it helps me shape my objectives. So I’m still the one that’s thinking about what I need to accomplish. I’m still having the conversation with my boss about what is important, and what we want to get done this year. But if you think about the clarity of expressing that objective, and you think about the act of editing and refining that, that was the role of Objective AI — [helping] our employees shape their priorities.
With regard to educating your workforce, how are you training your non-AI skilled workforce?
…We are skilling up everybody around the world. This isn’t just a San Francisco thing or a US thing. One of the things that really brought that to life for me the other week was we actually had an internal hackathon, and we had 200 people across forty teams that volunteered and signed up to participate. I’ve been part of hackathons for 15 years now… and what struck me [was] they were using AI to generate videos and presentations that communicated their working demo, their core idea, the value proposition, and did it in a way that was really crisp and succinct. …[AI] is this kind of equalizer around communication, where now we are really focused on the idea and we aren’t focused on whether or not someone’s English is perfect or not.
In 2023, you entered a partnership with LaLaLand.ai, and your company received backlash over concerns that you were using a company that builds digital customized AI-generated models to avoid hiring models, particularly people of color. What did you learn about AI from this particular incident?
I think we learned a lot of things, not just about AI… The thing that we really thought about was we’ve got to do what’s right for the consumer, and we’ve got to do what’s right for us. …I want to remind you the core premise of Lalaland was born out of the observation around size and fit. Like any e-commerce channel, we have a return rate, which means that some customers bought a pair of our jeans, they got them on their doorstep, they tried them on, and they didn’t fit to their liking. And so the whole goal of Lalaland was to help consumers with that core size and fit question, and help them have confidence in what they were buying. If you look at what has since happened in the past three years, you look at Gemini, you look at Nano Banana, you look at ChatGPT, these capabilities now generate clothing-draped models. There are lots of examples out there.
So the La La Land project was ahead of its time? That’s what you want to say there?
It was an experiment, and I think we learned and we will keep experimenting.
…The core question for us isn’t whether Levi’s adapts, but…we’ve got to do that while staying true to who we are.
In terms of AI, you are in the middle of a revolution here. When you look at the period of time since generative AI was introduced and the changes you’ve been involved with, what are your thoughts on how you assess all of that change?
I think the core question for us isn’t whether Levi’s adapts, but how deliberately and on whose terms, and we’ve got to do that while staying true to who we are. That’s what I think we’ve learned, and that’s what we are still trying to figure out.
Featured image by Coolcaesar – Own work, CC BY 4.0
You must be logged in to post a comment.