Levi’s Chief AI Officer Talks Raising the Bar (in Everything) with Machine Learning
It may be hard to find a more passionate AI evangelist in consumer goods than Levi Strauss & Co.'s Katia Walsh — which is not especially surprising given her title of chief global strategy and AI officer.
“I often say that the great thing about AI is you can do so much with it,” she concedes in a recent interview with CGT. “And the bad thing about AI is that you can do so much with it, which is where focus matters.”
Indeed, Walsh says her ultimate dream is for AI to support the entirety of the Levi’s enterprise — everything from HR and finance to design and product sourcing: “Every single thing we do [would be] AI-enabled.”
[See also: Special Report: Decoding AI]
While the company isn’t quite there just yet, the technology’s profound impact on the business is undeniable, as are the innovative opportunities it’s unlocking along the way. Prioritizing just where to use AI has become a balancing act of notching quick wins with strategizing for longer-term transformative impacts. “It's not just about the financial benefit,” Walsh notes. “It's also about the way in which it can lead to disruption in the company or the industry.”
It’s a vetting process that begins with the Levi’s executive team’s priorities and cascades all the way down to the store associate, with the goal of providing flexibility to jump on compelling use cases when they arise.
Walsh, a former journalist, analyst, and fintech consultant, walked through some of the most significant ways the company is currently leveraging AI and machine learning, including consumer engagement, product development, and pricing — much of which are marrying together promisingly to advance their sustainability goals.
Pricing, Promotions, and Planning
Levi’s is automating as much as it can when it comes to pricing and promotions, which Walsh says has enabled them to avoid deep discounting during store closures thanks to data showing consumers were willing to pay the full price based on brand recognition. While the company will occasionally discount in order to attract new customers, she notes the strategy has been a key factor in maintaining healthy margins in the last two years, as they’re able to set an optimal price from the start.
When applied during end-of-season sales and promotional events, the company is seeing success using AI with adaptive pricing. For example, an unexpected influx of warm weather during Singles Day in China prompted them to lower the price in real time on a down jacket and increase the price on a lighter jacket.
[See also: Levi’s Doubling Down On AI to Power Demand Forecasting & CX]
It’s also transforming how it does merchandise planning. A previously manual task that was very spreadsheet-driven, they’ve now created a livestreaming repository of data to use in creating machine learning models and algorithms to predict the demand for every type of product in various channels in every location in the world.
“[This] is really important because that determines the investment we make in making and buying our products for the next season or the next year,” she notes. “And that helps us not only price right and buy the correct amount, but also eliminate waste, which has a sustainability benefit.”
Sustainability Potential
Walsh is a strong believer that AI can save fashion, an industry plagued by waste. “With AI, you can actually have optimized creativity and profitability and sustainability at the same time, because you minimize waste. You manufacture exactly what you know people want to buy.”
Though it’s still too early to quantify the impact, she says the tech is helping them achieve their sustainability goals. More work is needed in training the models and increasing accuracy across markets; the company has seven clusters globally, and the model performs differently depending upon the market.
[On Demand: 3 CIOs Talk Today's Sustainability Priorities]
The company’s laser finishing, however, is reducing the company’s use of water and chemicals. It connects neural networks with lasers to create patterns in two minutes what previously took two days.
The company’s RAMBO program leverages a proprietary, patent-pending algorithm that determines where to ship an e-commerce order. Developing during the retail store closures, the recommendation engine predicts whether a product will need to be discounted in the future, which in turn determines whether it’s more profitable to sell the item at full price and ship from a store — even if the cost of shipping will be higher than from a distribution center.
It also determines how to minimize the number of shipments from different places, further bolstering both customer service experiences and sustainability efforts.
Consumer Engagement Cues and Collection
Levi’s recommendation engine, first developed in 2019, works in real time to understand consumers’ online behavioral cues to provide personalized shopping recommendations. The tech is not only increasing its mobile app downloads, but it’s also raising average order value.
Its Red Tab consumer loyalty program, meanwhile, is collecting first-party data from its more than 5 million members that extends beyond straight demographics — such as interests, desires, and connections. Levi’s then applies machine learning to predict the most relevant benefits for each member so it can serve up exclusive offers, such as access to events and new products.
Boot Camp Marches On
Demystifying artificial intelligence across the enterprise is a common challenge for today’s consumer goods companies, and Levi’s is no different. To combat this, they’re taking steps to increase education and access to show what’s been accomplished and what’s possible.
Its machine learning boot camp is probably the most visible evidence of these efforts. Launched last year and open to all employees, it’s now graduated 101 members and is set to enroll 60 more in the spring. The program has digital upskilling portfolio with two dimensions — audience and skill set — with hands-on education in such writing scripts in Python and the upcoming session now adding statistics.
Not only does it pull people out of their daily responsibilities for eight weeks to focus on their education and work on real-life use cases, but it also offers the opportunity to work with people they may not have otherwise met. The program culminates with a presentation to the Levi’s executive team.
While boot camps are not necessarily a new concept, this one differs in that it’s open to all employees, says Walsh, who describes the application process and screening as very rigorous. “It’s hard to find people who will make it through the boot camp and apply their skills, but don't have any kind of coding or statistics skills as a prerequisite, [so] the application process screens for problem solving.”
In measuring its success, Walsh touts more than just the 100% graduation rate: About half of the graduates apply the skills they’ve learned least a quarter of their time — and this includes the associates who work in retail stores, not just Levi’s corporate.
For example, a retail store manager in a Denver premium outlet store created a neural network that identifies the optimal way to bundle products in the best outfit. Now when helping customers, she can make recommendations that draw from her previous experience and the data- and ML-powered predictions.
Another employee developed a model that predicts the likelihood of equipment failure in her distribution center — a particular pain point for that associate. She also created a streaming app to make the tech usable to other distribution center employees, enabling them to proactively dispatch technicians when they receive a warning the tech may malfunction. While they’re not yet using it other distribution centers, Walsh said it is something they can scale.
Yet another example: Ron Pritipaul, a Levi’s design coordinator and boot camp student, built a neural network into an existing algorithm to identify such things as button placements and pockets for use in design images. He then layered in information from iconic works of art, including Van Gogh’s “Starry Night,” to develop thousands of design options with just one click.
Pritipaul also built a computer-vision-powered app to help with the extraordinary time-consuming process of color-matching threads, displaying results in seconds and eliminating hours of manual work.
“We have countless examples of surprising things that I would have never planned,” she says. “But because we let these employees — who had known the company, had known the industry, had been facing certain problems throughout their tenure — now they have the tools to tackle the solving. And they went about doing it.”
[See also: Getting to Broad-Scale AI in the Enterprise]
The company is also recording greater engagement and retention for both the employees and their managers — crucial measurements during today’s challenged labor market — and Walsh says that pretty much everyone who has completed the boot camp has remained with the company, and that it’s paid for itself.
“We’re also tracking the financial benefits of the work they're doing,” she adds. “We never set out for this to be a money-making enterprise. It was purely for education — we wanted to invest in our employees — but [we’re] seeing when these people go back to their jobs, and they automate and use predictive models to create better accuracy in their day jobs, it actually has financial benefits.”
These employees are also becoming AI evangelists themselves and converting others to see the value in automation — helping evolve fears of being replaced by technology. Once trained, they’re able to see the tasks being eliminated are repetitive, manual, and not necessarily career-building, another advantage when it comes to retaining employees.
