Closing the AI Value Gap: How Fashion Can Grow Its Technology ROI

Fashion’s demand signals can change on a dime. With quick trend cycles and consumer behavior shifts, a style that seemed in-demand during planning and production can flop by the time it reaches retail. On the flip side, a design with a shallow buy could unexpectedly take off, requiring a quick read on interest to ramp up production and avoid missed sales opportunities.

Artificial intelligence has the potential to drastically improve how executives tackle the complexities of the fashion business. When integrated with an organization’s systems, the technology can autonomously search for insights on product performance that can be acted on during the selling season.

Supply chain software company Aptean worked with technology market researcher Vanson Bourne to survey 245 apparel and fashion executives about AI. The results show the industry is investing in automation, and a significant portion are choosing fashion-specific solutions that understand the nuances of apparel operations.

Sourcing Journal spoke with Alain Tessier, director of product management at Aptean, about some of the survey’s findings. Here, he shares why data management is critical for successful AI implementation, where companies should keep people in the driver’s seat and how fashion firms can get more out of AI tools.

Sourcing Journal: The survey showed that among the sectors studied, apparel organizations were most apt to have adopted an industry-specific AI solution. Why do you think that is?

Alain Tessier: General AI tools don’t understand the most critical pieces of the apparel industry—such as what a season cancel date is—and in this business, that date has real financial impact. The time between a designer sketching an idea and that item hitting a store shelf is eight to 12 months, sometimes longer. In that window, you’re committing budget, locking sourcing capacity and betting on a color or a print you may only be able to buy once. A generic system can crunch numbers, but it can’t reason across a style matrix or tell you that this particular floral only comes around once a season. Fifty-three percent of apparel organizations in our survey have already adopted industry-specific AI, and I don’t think that’s a coincidence. It’s the first time the tools actually match the way this business works, instead of the other way around.

Sourcing Journal: Another key finding was that apparel is having more success at gaining a competitive advantage from AI than other industries. How have you seen your clients strengthen their positioning through AI?

A.T.: It usually comes down to catching a problem while there’s still margin left to save, not after. Take a footwear brand’s planning team: They spend most of a Monday morning pulling sell-through reports before their merchandising review, and by the time they’ve found the styles in trouble, the cancel date is already too close to do anything but discount. Give that same team a faster way to see which styles are sitting on high inventory with low sell-through and a shrinking window, and they can spot three problem styles in minutes instead of hours. They cancel two open purchase orders and set up a targeted promotion on the third before the meeting even starts. That’s the real competitive edge apparel is getting from AI right now. It’s not some abstract efficiency gain, but the ability to act while a decision still matters instead of writing a postmortem for it. Forty-one percent of apparel respondents in our survey say they’ve seen this kind of gain over the past year.

Sourcing Journal: As agentic AI is being rolled out more broadly, how should companies set safeguards to ensure there are guardrails for the agents’ autonomous decision-making?

A.T.: You put a human approval step in front of anything that actually changes your business, full stop. It makes sense for a multichannel brand to set up a system that pulls together sales, order status and inventory across wholesale, direct-to-consumer and retail every Monday morning, automatically. That part can run entirely on its own; it’s just assembling information. But the moment that system wants to write something back into the ERP, a person has to say yes first. That’s the line: for research and analysis, let the AI run. Anything that touches an order, a shipment or a customer commitment, a person signs off. Skip that step and you’re one hallucinated purchase order away from a real mess. The companies most confident about autonomy are usually the ones who decided, in advance, exactly where that line sits, before they ever turned the agent loose.

Sourcing Journal: What considerations do companies need to make in terms of data management and governance? What are the risks if they don’t have their data in order?

A.T.: Most companies don’t think they have a data problem until the moment they realize they don’t actually know where their data went. I sat with a customer a few weeks ago who mentioned they’d been using a general AI tool for a few tasks, so I asked if they had an enterprise or private tenant set up. They didn’t know, which most likely means whatever they typed in was helping train someone else’s model, not just answering their question. That’s the risk nobody budgets for. It’s not that the AI gives you a wrong answer, it’s that your product data, your cost sheets and your customer information end up sitting somewhere you can’t see and can’t control. Three-quarters of apparel organizations in our survey say they need real data governance in place before they’d trust AI with a bigger decision, and honestly, that number should probably be higher. The businesses I’d worry about are the smaller ones who haven’t even asked the question yet.

Sourcing Journal: The survey found that most organizations haven’t yet reaped the full value of AI. How can companies get more out of their AI investments?

A.T.: Most of the companies stuck in neutral didn’t fail because they picked the wrong AI. They failed because they layered it on top of the same messy mix of systems and spreadsheets they’d always had and expected a different result. I talk to a lot of clients about moving to the cloud, and it’s rarely really about the cloud itself. It’s about making styles, inventory and sales data actually connected and reachable, because AI sitting on top of disconnected data just gives you a faster wrong answer. The businesses getting real value usually started smaller than you’d expect: They picked one specific, painful problem, like a purchase order process or a planning report nobody enjoyed doing, fixed the data underneath it and only then let AI touch it. It’s not glamorous. But that’s the difference between a pilot that gets abandoned in three months and one that’s still running a year later.

Sourcing Journal: What steps should the fashion industry take next to advance its use of AI?

A.T.: Get your data foundation in order before you go shopping for more AI, and start building your governance rules now, not after something goes wrong. Almost half of apparel organizations, 46 percent, are already planning to move at least some systems to the cloud in the next year, and that’s the right instinct, because it’s hard to do anything meaningful with AI when your product and inventory data live in three disconnected systems. After that, I’d tell most brands to resist the urge to solve everything at once. Pick the highest-risk, highest-cost problem, usually something tied to markdowns or inventory exposure. Get AI working there first, prove it out, then expand from a position of confidence instead of hype. The industry’s ahead of most others in adoption. It’s not yet ahead in capturing the value, and that gap is where the next 12 months get decided.