How Levi Strauss & Co. uses AI: the data programme and its lessons

KEY TAKEAWAYS Summary by the editors

1. Levi Strauss & Co. ran an eight-week machine learning bootcamp in 2021 that produced 101 graduates from existing staff in its first year, most of whom returned to their original departments.
2. In March 2023 the company faced a backlash after announcing AI-generated models from Lalaland.ai as a way to increase model diversity, and clarified that they would supplement, not replace, real shoots.
3. In November 2025 Levi Strauss & Co. announced a Microsoft Teams based super-agent on Azure that coordinates sub-agents for IT, HR and operations, with deployment from early 2026.
4. Its Outfitting feature recommends looks in the Levi's app in the US, Canada and European markets, and its Stitch assistant supports staff in 60 US stores.
5. The company has not published audited financial results attributable to AI; a 2021 award case study claimed 76 million US dollars a year in value from digital, data and AI.

Levi Strauss & Co. built its AI programme around people as much as technology: it trained its own employees as data scientists, then applied AI to inventory, operations and design, and since 2025 has deployed AI agents for staff and shoppers. Its record also includes a public misstep with AI-generated models in 2023. Audited financial results attributable to AI have not been published.

What has Levi Strauss & Co. built?

The programme has three phases in the public record. The first was capability building. In 2021 the company launched an eight-week machine learning bootcamp in coding, Python and statistics for employees without a technical background. By November 2021 it had produced 101 graduates across two cohorts, with participants from more than 20 locations and over 450 employees applying for places. Use cases named in a 2021 case study published by Constellation Research included inventory prediction for store managers, equipment failure prediction in distribution centres, quality inspection, design generation and production cost optimisation.

The second phase was generative imagery. On 22 March 2023 the company announced a partnership with Lalaland.ai to test AI-generated models, saying they would increase the number and diversity of models shown. The third phase is agents. In November 2025 it announced:

• a super-agent built with Microsoft inside Teams and powered by Azure, coordinating specialised sub-agents across IT, human resources and operations, with deployment starting in early 2026;
• Outfitting, an AI styling feature in the Levi's app that recommends looks based on preferences, purchase history and trends, live in the US, Canada and European markets;
• Stitch, a mobile AI assistant that gives store employees product information, procedures and training, deployed to 60 US stores after a pilot.

Why did Levi Strauss & Co. invest in AI?

In 2021 the stated rationale was talent. Katia Walsh, then chief strategy and AI officer, said investing in people would yield the biggest return from the digital transformation. Training employees who already understood supply chain, retail and wholesale was a way to put data skills where decisions are made.

By 2025 the rationale had become operating model change. Chief executive Michelle Gass called AI a key unlock as the company rewires how it works, from stores to corporate offices. The company reported 6.4 billion US dollars in net revenues for 2024 and describes an ambition to become a 10 billion dollar retailer. Our reading: that ambition implies more direct selling and therefore more demand for personalisation and efficient store operations.

Chief digital and technology officer Jason Gowans framed the 2025 initiatives around both sides of the business, saying AI allowed the company to rethink how it works and how it engages with its fans. That pairing of internal efficiency and customer experience is visible in the choice of a staff super-agent, a store assistant and an app feature launched together.

Read also

How Inditex uses AI and data: RFID and a demand-driven supply chain

How does it work (data, models, process)?

The bootcamp model is a process innovation: graduates return to their own teams and apply machine learning to local problems, supported by the central strategy and AI function. According to the Constellation case study, 75% of the 2021 graduates returned to their original departments and 25% joined the strategy and AI function. The agent architecture is different: a central super-agent routes requests to specialised sub-agents, so that employees use one interface in a tool they already work in. Outfitting draws on customer preferences and purchase history and updates daily, according to the company.

What results has Levi Strauss & Co. reported?

The only financial figure in the public record comes from the 2021 Constellation Research case study, which states that digital, data and AI work at the company delivered 76 million US dollars a year with a 76% gross margin. It is a self-reported award submission, not an audited disclosure, and it covers digital and data work broadly. For the 2025 agents, the company has published deployment facts (60 stores for Stitch, markets for Outfitting) but no productivity, conversion or cost results.

What are the limits and open questions?

The following is editorial analysis. The 2023 episode is instructive. Critics argued that AI-generated models were no substitute for hiring diverse people, and on 28 March 2023 the company said the project should not have been portrayed as a diversity initiative and would supplement rather than replace live shoots. The lesson is that framing and stakeholder impact matter as much as technical feasibility.

Other open questions concern sustainability of the talent model after the departure of its original sponsor, measurement of agent productivity in knowledge work, and data quality: agents for IT, HR and operations are only as reliable as the policies and records they draw on.

Read also

How Uniqlo and Fast Retailing use AI and data in the Ariake Project

What can other fashion companies learn?

1. Train domain experts in data skills; they know which problems are worth solving.
2. Send trained staff back to their departments so capability spreads instead of concentrating centrally.
3. Test the message of customer-facing AI with affected groups before announcing it.
4. Put AI where employees already work, such as collaboration tools and store devices.
5. Define how value will be measured before rollout, so results can be reported beyond deployment counts.

Frequently asked questions

What was the Levi's machine learning bootcamp?

It was an eight-week programme launched in 2021 that taught employees coding, Python and statistics so they could apply machine learning in their own roles. It produced 101 graduates in its first year, most of whom returned to their original departments.

Why was Levi's criticised for AI models?

In March 2023 Levi Strauss & Co. said AI-generated models from Lalaland.ai would increase the diversity of models shown online. Critics said the company should hire diverse human models; it then clarified that the AI would supplement, not replace, live shoots.

What is the Levi's super-agent?

It is an AI orchestration layer built with Microsoft inside Teams and powered by Azure. It coordinates specialised sub-agents for IT, human resources and operations, with deployment starting in early 2026.

How does Levi's Outfitting feature work?

Outfitting is an AI styling feature in the Levi's app that recommends complete looks based on a customer's preferences, purchase history and current trends. It is live in the US, Canada and European markets and updates daily.

GuideThe complete guide to AI strategy for fashion companiesRead the complete guide