As the global beauty industry benefits from the rapid growth of e-commerce, new product virality driven by creator-driven content on social media channels like TikTok, and AI-driven product discovery, these tailwinds also pose more complex logistical challenges for manufacturers.
“Shopping for beauty products is no longer a planned, needs-based, replenishment shopping journey,” Biswaranjan Sen, Chief Product Supply Chain Officer at Unilever Beauty & Wellbeing, told BeautyMatter. “It is being replaced by experiential always-on shopping behaviors, leading to very high levels of volatility in demand.”
The global beauty market has increased by 10% year over year, according to an April report published by NielsenIQ (NIQ). But 87% of beauty product discovery happens through social channels, according to Sen, who added that “With beauty trends moving quickly, production needs to stay tightly aligned with what consumers are buying.”
This may explain why major beauty companies, including Unilever and Estée Lauder, have embraced a wide range of AI-enabled use cases in manufacturing, including physical AI applications that leverage advancements in AI and computer vision technologies embedded in physical machines.
These AI-driven machines can more precisely perceive, understand, and perform tasks in real-world settings. Experts say that AI advancements and declining robot manufacturing costs have helped fuel demand, but so too has necessity. Labor shortages are rampant across the sector, and for beauty brands and retailers, e-commerce and AI-driven product discovery have accelerated demand for faster fulfillment.
At Unilever’s manufacturing site in Hefei, China, the consumer products giant is currently piloting advanced, humanoid AI systems to complete workplace tasks. The site has already deployed nearly 120 robots across its logistics operations, reaching around 90% automation, according to Sen. Overall, 31 different digitally enabled use cases have resulted in orders being fulfilled 75% faster, while logistics costs are down by 24%.
Other AI use cases at Hefei focus on equipment setup, and algorithms trained to learn from past malfunctions that can proactively alert team members when new issues may arise. Beyond manufacturing, AI is also helping Unilever better sense shifts in demand and then use that data to inform fulfillment and supply chain planning.
“As beauty portfolios continue to expand and trend cycles shorten, physical AI allows factories to respond faster without compromising on quality or cost,” Sen said.
Pierre Dupreelle, Managing Director and Partner at consulting giant Boston Consulting Group (BCG), said he advises beauty industry clients to always pilot physical AI technologies at a small scale to ensure they understand the potential financial impact. “Some of them require a large investment,” Dupreelle told BeautyMatter. “You measure what the impact is, depending on what you’re optimizing for, and then once you say it’s working, you can scale it to a larger portion of your business.”
Florent Guillet-Caillot, Vice President of Supply Chain at Estée Lauder, told BeautyMatter about a new AI tool now online in two factories called “Ella,” which stands for “Estée Lauder Line Assistant." Ella helps operators address equipment issues and speeds up when the manufacturing line switches between product runs, and it will roll out to additional factories over time.
There are also AI-enabled vision systems in place that check the quality of glass and other packaging components, and even monitor quality control for beauty products like lipstick. Previously, Guillet-Caillot said, “We had to check every single product with a limit to what a human could potentially catch. Now we guarantee that quality, and we do it much faster and much more efficiently.”
Guillet-Caillot said these technology investments are a critical component of Estée Lauder’s “Beauty Reimagined” turnaround plan, launched mere weeks after Stéphane de La Faverie became CEO in January 2025. After three consecutive years of declining sales, Estée Lauder is in the early stages of a comeback, with organic sales for the current fiscal year projected to increase by 1% to 3%—and one way it envisions restoring sustainable sales growth and gains in double-digit adjusted operating margins is by speeding up product innovation.
Estée Lauder has had to reimagine the workflows that go into turning a marketing brief into a product on the manufacturing line. This is where the company’s investment in digital twin technology comes into play: a virtual representation of the physical manufacturing floor. With this capability, Estée Lauder can plan modifications virtually, facilitating a more streamlined back-and-forth between product designers and manufacturing operators.
“In the past, that's really an area where we were losing time and losing money, because we had to run all those trials, which also takes time,” said Guillet-Caillot. “For those trials, you need components, formulas, and man-hours until you reach the optimal setup. That’s where AI, through the digital twin, is helping us make a positive impact on the design-to-launch workflow.”
Robotics startups are also assertively pitching their technology solutions to large beauty brands. These startups have recorded $27.6 billion in venture capital investments across 1,009 deals in 2025, according to financial data provider PitchBook. Globally, 542,000 robots were installed in factories in 2024, more than doubling from a decade before, according to the latest annual figures published by the International Federation of Robotics.
Locus Robotics, founded in 2014, says its autonomous warehouse robots have “picked” 7 billion items, and that around 600 million units, or 8%, were for beauty products. One of its biggest beauty customers is Ulta Beauty, where the company’s LocusBots were deployed to fulfillment centers in South Carolina and Illinois in 2023 and 2024, respectively. At both sites, the startup’s robotics work in a 130,000-square-foot building, handling e-commerce and retail store orders.
“The whole idea is to be able to collaborate with humans, to be able to get from point A to point B, and eliminate that walking for the people,” Jasmine Lombardi, Chief Customer Officer at Locus Robotics, told BeautyMatter.
Despite President Trump’s vow to bring more manufacturing jobs back to the United States, many manufacturing experts say that modern workers don’t want to perform physical tasks that historically involve lifting and moving boxes and carts weighing hundreds of pounds across large warehouse floors with poor temperature control. By 2033, the US manufacturing industry is estimated to need an additional 3.8 million workers, according to Deloitte and the Manufacturing Institute.
“It can be very difficult to find workers,” Lombardi said.
At Ulta Beauty, because Locus Robotics is handling many repetitive tasks autonomously, the startup claimed it has reduced the time required to train workers by 75%, from 4 hours to 15-60 minutes.
But even with all the touted benefits, Lombardi admitted Locus Robotics has work to do to ease worker apprehension that their jobs may be at risk when robotics are deployed. “People are skeptical; ‘Oh my God, is the robot going to take away my job?’” She asked rhetorically. “Once they get the hang of it, they understand it's more of a coworker.”
Meanwhile, in California, Unilever’s skincare brand Dermalogica deployed Corvus Robotics’ aerial inventory drones in just 16 days. “We can even recognize when the system says a product expires on the 31st, but the palette says it’s six days earlier,” Tony Esquith, Head of Enterprise Sales at Corvus Robotics, told BeautyMatter. “That mismatch we will flag.”
Corvus Robotics’ system is capable of image-scanning Dermalogica’s entire floor 52 times per year, representing a 600% increase in inventory imaging frequency compared to manual processes.
“If you can find a product before it’s set to expire, and you can get it out the door and get it on the shelf, it’s an instant value,” Esquith said.
Unilever’s Sen, meanwhile, sums up all the necessary ingredients to scale physical AI: clean and connected data, consistent standards across multiple manufacturing sites, and tight integration between enterprise software systems and factory-floor technology.
“What’s key is to redesign ways of working so people and AI-enabled systems operate efficiently together, with clear guardrails, safety controls, and accountability,” said Sen. “There is tremendous opportunity in scaling physical AI across our manufacturing network to turn proven use cases into repeatable capabilities for daily operations at scale.”