Artificial intelligence can be a huge benefit to a retail organization, if used incrementally.
During a recent discussion “State of Retail: Supply Chain, AI and the Future of Fashion,” executives from Infor argued that AI success depends less on the technology itself than on disciplined business processes, clean data and a clear understanding of where automation creates measurable value.
“AI is not just a tool that you can insert,” said Ana Friedlander, senior director of solution strategy for fashion and retail at Infor. “It’s an augmented capability to add to your processes and business models.”
That distinction is important as retailers race to modernize operations amid continuing supply chain volatility, shifting consumer demand and increasing margin pressure. But rather than sweeping AI initiatives, experts recommend a strategic approach that starts with individual use cases.
“Don’t start with an ‘AI strategy,’” she said. “Start where the decisions can be painful, where the decisions are frequent and where the work is manual.”
For most retailers, that means demand forecasting, purchase order allocation and inventory balancing—three areas where teams routinely make judgment calls based on large volumes of data.
She advised to smart small by identifying a single business problem, cleaning only the data required for that use case, deploying AI as a decision-support capability rather than full automation, and measuring results quickly before expanding further.
“The biggest mistake that a lot of companies make is they’re trying to boil the ocean,” she said.
Bad Data, Bad AI
One of the strongest themes throughout the discussion was that poor data—not poor AI—is the primary reason many initiatives struggle to succeed.
Fragmented product hierarchies, disconnected ERP and PLM systems, duplicate SKU structures and spreadsheet-driven workflows all create barriers to effective AI adoption.
“If there’s flawed data, the decisions and recommendations are flawed, so if you’re not prepared with your data, take a step back,” she said, noting that heavy data volume isn’t always the problem. “You might have 4 million skus, but if your data is properly done, [you’ll be okay].”
Despite the rapid, and honestly mind-boggling, advances in generative AI and machine learning, retailers should resist the temptation to remove humans from critical decisions. AI should “augment expertise” rather than replace it.
“Use AI as a decision support tool and not a full automation service,” Friedlander said.
Technology Isn’t a Quick Fix
Jonathan Burch, who leads partnership initiatives at Infor, believes many organizations underestimate the change management required for successful AI adoption.
“Organizations are really complex,” he said. “You can’t just throw AI in and expect it to have results.”
Instead, companies should first identify employees who are already experimenting with AI to improve workflows. Those early adopters often become internal champions capable of accelerating broader transformation.
Burch also encouraged retailers to think about AI less as a novelty, but as infrastructure supporting better business decisions. “Every implementation should begin with a simple question: What business problem are we trying to solve? And is this tool the right application of AI?” he said.
And finally, employees need confidence that AI recommendations are accurate before they fully welcome AI into their operations. Constant validation, measurement and refinement will build the trust to ease change management issues. Without it, users will just revert to manual workarounds.
Driving Product Development
Technology is also reshaping product development. As fashion cycles continue to compress, retailers must balance core products on longer production cycles with trend-driven assortments that respond faster to changing consumer demand.
Successful brands now should separate products into those that are ‘proven and scalable’ (core); ‘proven but constrained’ (leaking margin) and ‘altogether new.’ Meanwhile, AI can incorporate external signals—from social media trends to weather data—alongside historical sales performance for smarter, faster product development.
But it’s not just speed. Brands need to ensure their suppliers, vendors and partners across the supply chain can scale and stay flexible.
“Ordering 20,000 pairs of a trendy item like bedazzled jeans may look appealing, but compared with a proven best-selling product, it could create more risk and less guaranteed profit. Companies must balance proven, scalable products with new opportunities,” said Friedlander. “And this is where all of your questions in AI comes in. It’s product, it’s market, it’s your target market, it’s also your margins as well as your sustainability and capability to grow.”
Technology can also help with disruptions, which are rapidly becoming the new normal.
Burch noted that fashion companies need technology that both helps them plan for the long term while staying agile enough to respond quickly when the market changes. And changes again.
“In the last five years, you’ve probably had to respond in a much faster way than 10 to 20 years ago, and that could just be because of trends in consumer taste, but also the bigger external global factors that you can’t control,” he said.
But despite reasons to update technology, some companies still resist AI due to inertia.
“The biggest misconception is that the way it’s been is the way it always has to be,” said Burch.
Many fashion brands are still operating with outdated processes because they assume that’s simply how the industry works. Or they view digital transformation as something merely to use to exchange data with retailers and avoid compliance fines, even accepting those fines as a normal cost of doing business. Instead, companies should rethink how technology can improve how they run their business.
“When you have solutions that are fast and reliable, they open up the door for a cost center or compliance manager to communicate and onboard with other retailers. You can share and pull in a greater set of information that’s very accurate,” he said.
