Why an H&M Alum Launched a Talent Company for AI Models

Subscribe to Tech Mode by Marc Bain to go deep on the most intriguing developments in AI and how technology is reshaping the fashion industry.

Welcome back to Tech Mode, your monthly guide to how AI and other technologies are changing the fashion industry.

There’s been a lot of great AI coverage from my colleagues here at BoF recently. This month, we did a whole package on how the technology is changing marketing, a topic I’ll be speaking about at a special event we’re hosting in London this evening. Here are all the stories, ICYMI.

• Marketing correspondent Haley Crawford wrote a case study on how fashion is using AI at every step of the funnel. That’s one you’ll definitely want to check out if you haven’t already.
• AI is rewriting the job requirements of marketing roles in fashion companies.
• Luxury brands using AI for campaigns are learning whether they succeed or fail is less about the technology and more about how they use it.
• AI agents can trigger online advertisements, meaning as their number grows they could skew metrics and distort fashion’s ad economy.

Separately, I wrote a story yesterday on why people hate AI so much, and what that means for fashion brands. One stat that caught my attention: 55 percent of Americans think AI is likely to do more harm than good in their day-to-day lives, according to a Quinnipiac University poll.

Let’s dive in.

A New Talent Company for AI Models

Big idea: One of the minds behind H&M’s foray into using digital twins of human models has struck out on her own with a new company that’s pushing the idea even further.

Alva, which had its soft launch earlier this month, describes itself as a talent company that creates, licenses and safeguards digital twins. Its aim is to produce these AI avatars not just for models but also artists, actors and athletes in premiere sports such as soccer and racing — basically anyone who is their own brand and might want to use their twin for a photoshoot while they’re busy doing something else.

Backstory: The company’s founder and chief executive is Louise Lundquist, who was previously a business developer at H&M — and one of the people I spoke to last year while reporting on the retailer’s plan to use AI models. H&M is Alva’s “founding client,” according to Lundquist. (Ellen Svanström, H&M Group’s chief digital information officer, contributed a quote to the launch press release.)

“There’s room for an independent player to solve what needs to be solved for the ecosystem to continue to expand into an AI future,” Lundquist told me when we caught up recently.

Alva wants to be the infrastructure layer: It provides the technology platform and works with the talent to create the twin, which Alva holds the exclusive license to. The talent, meanwhile, maintains ownership rights of their data.

Twinning: When BoF first reported on H&M’s plan to use digital twins, not everyone was a fan of the idea. A number of photographers, stylists and others whose labour is typically required for a photoshoot worried about what would happen to their jobs if the process were to be digitised. Those fears haven’t gone away, while AI has only continued to improve. It can produce images and video that are nearly indistinguishable from reality already, and it’s not hard to imagine more models, actors, athletes and others wanting to create AI twins of themselves that they can license out to bring in more cash. (On a related note, celebrities such as Taylor Swift and Matthew McConaughey are applying to trademark their images to prevent AI-generated infringements.)

Alva is already working with models such as Vilma Sjöberg, and Lundquist said it has some high-profile sports clients it can’t yet name. The company is also collaborating with top-tier modeling agencies as it seeks to grow. The bottleneck at the moment, according to Lundquist, is brands upskilling their teams and familiarising them with this new production method so they can scale its use.

Retail’s AI Invisibility

Hide and seek: Adobe Analytics released new data this month showing large portions of retail websites are not visible to LLMs, even as traffic to e-commerce from AI platforms continues to grow. Using its own diagnostic tool, Adobe analysed how much of US e-commerce is readable by AI. A score of 50 percent, for example, meant half the content on a page was visible to AI, and half was not.

For some types of pages, a significant share was invisible, including:

• 25 percent of retail homepages
• 26 percent of category-level pages, such as “men’s apparel”
• 34 percent of individual product pages

Why it matters: AI traffic to US retail sites is still relatively limited, but it’s scaling fast. By Adobe’s count, in the first three months of 2026, it surged 393 percent year over year.

That traffic is now driving sales, too. In March, AI traffic converted 42 percent better than from non-AI sources such as paid search or email marketing, according to Adobe’s data. That’s a major reversal from a year prior, when by comparison it was 38 percent worse. Adobe cited rising trust in AI’s recommendations as a factor.

The suggestion is that e-commerce has been slow to adapt to AI’s growing influence on product search and discovery. “Retailers have thousands of SKUs, and our data shows that much of the content is currently invisible to LLMs,” Adobe said.

Leaders and laggards: These sorts of averages, of course, obscure the differences in performance hiding in the data. Adobe looked at that, too, and pointed out that, in the case of US retail’s top performers, just 17.5 percent of their sites were unreadable. For the worst performers, that number rose to 45.8 percent.

“This gap between the two sets of brands shows that some retailers have moved on this trend with greater speed, updating content across their webpages,” Adobe stated.

Psychological Barriers to AI Agents

A new study from the Wharton School at University of Pennsylvania and Science Says, an AI and marketing insights firm, argued that what’s holding back adoption of AI agents is less technology these days and more psychology.

“Giving an AI agent control over fundamental parts of your life is a completely different proposition than just asking a question to a chatbot,” the researchers wrote. “It requires believing the agent can do the job, trusting it will do it right, and being willing to hand over control.”

So what: The findings have implications for fashion brands hoping to launch their own AI shopping agents, and the study offered some tips that could smooth the way to wider adoption, which could also help mitigate customers feeling like the AI is a cost-cutting tactic. There are too many to list in full, but here are a few:

1. Signal competence: People were less likely to use AI that sounded cheerful and friendly compared to AI that clearly explained its reasoning and cited the criteria it used to make decisions.

2. Show added value: Users simultaneously weigh the risks of using an AI agent, like privacy concerns when handing over data, against its benefits, meaning the benefits need to be obvious and outweigh the risks. When using the tech to book travel, people valued AI that was convenient, functional, always available and could personalise the experience based on their past preferences.

3. Explain the process: It also helps for the agent to show the steps it took and sources it used to perform an action. Reliability is a top concern when people use AI, and they’re more apt to think the AI is reliable if they understand what it’s doing.

4. Give it a supporting role: There was less resistance to AI when it was presented as supporting human experts, rather than being an expert itself.

5. Disclose its limits: When people were told upfront about the limitations of an agent and where it was likely to fail, they trusted it more and worked more effectively with it.

Yes, but: I don’t blame anyone for hesitating to hand over tasks to an agent, let alone their credit card details, when LLMs still frequently make mistakes. One psychological factor at work here is the normal friction involved in adapting to new technology, but another is the distrust that comes from people’s firsthand experiences with AI, as the researchers themselves acknowledge.

Until people’s main uses of AI, which include things like getting answers to questions and writing emails, are free of errors, they may hesitate to rely on agents, even for low-stakes tasks like shopping. Building trust will take time and effort, but it is happening.