Seven Lessons from the AI Frontier
What Retail Leaders Can Learn from OpenAI's Enterprise Report
The question isn't whether AI will transform retail—it's whether your company will lead that change or scramble to catch up. OpenAI's new "AI in the Enterprise" report shows exactly how companies are already winning with AI. We're talking about real results: 20% performance improvements, $40 million profit gains, and 60% efficiency boosts.
These aren't pie-in-the-sky promises. Companies like Morgan Stanley, Indeed, and Klarna are seeing measurable returns right now. For retail leaders, their playbook offers a clear path to AI success that goes way beyond just buying new technology.
Start with Measurement, Not Dreams
The best AI projects begin with boring spreadsheets, not grand visions. Morgan Stanley proves this point perfectly. When they wanted to help financial advisors work better, they didn't launch a company-wide AI revolution. Instead, they picked three specific things to measure: how accurately AI translated languages, how well it summarized documents, and whether its responses matched expert-level advice.
This careful approach paid off big. Today, 98% of their advisors use AI every day. Document access jumped from 20% to 80%. Client follow-ups that used to take days now happen in hours.
Here's how a grocery chain could use the same approach for food waste. Pick one problem item—say, strawberries that spoil quickly. Set up a test with some stores using AI to predict demand and others sticking with the old system. Measure everything: spoilage rates, stockouts, and forecast accuracy. Only expand to other products after you prove it works.
The key insight? Ask "How will we know if this AI actually helps?" before you ask "What cool AI stuff can we do?"
Make AI Part of the Customer Experience
The biggest wins come when AI isn't a separate tool but becomes part of what customers already do. Indeed learned this with job matching. Their AI doesn't just suggest jobs—it explains why each job makes sense for that person's background. The result? 20% more people applied for jobs, and 13% more actually got hired.
Think about a home improvement store that builds a "Project Helper" into their app. A customer planning a deck doesn't just get a generic guide. The AI helps them design it, shows what it would look like in their yard, creates a shopping list that updates when they change materials, and gives step-by-step instructions just for their project.
This isn't about adding AI features. It's about using AI to make the whole experience better. Customers aren't just buying materials—they're getting help with their entire project.
Start Now Because Results Build on Each Other
Here's the thing about AI: it gets better the longer you use it. Klarna's customer service bot now handles two-thirds of all customer chats, cutting response time from 11 minutes to 2 minutes. That's a $40 million profit improvement. But it didn't happen overnight—it got better through constant testing and tweaking.
Even better, 90% of Klarna's employees now use AI daily. This creates a snowball effect where the whole company gets smarter about AI, which leads to better ideas and faster execution.
Stitch Fix shows what happens when you start really early. They built their entire business around AI from day one in 2011. Now they have more than a decade of data on what people like to wear. Every purchase, every return, every customer review has made their AI smarter. That's a competitive advantage you can't buy—you have to build it over time.
(more: Tacticone, Stitch Fix)
The lesson? Companies that start now will have years of experience and data when AI becomes essential in their industry.
Customize AI to Fit Your Business
Generic AI is like a store-bought suit—it's fine, but custom is better. Lowe's worked with OpenAI to fine-tune AI models for their specific products and customers. The result: 20% better product tagging and 60% better error detection.
“Excitement in the team was palpable when we saw results from fine-tuning GPT 3.5 on our product data. We knew we had a winner on our hands!”
Nishant Gupta
Sr Director, Data, Analytics and Computational Intelligence at Lowe’s
Using another retail example, here's why this matters for a clothing retailer: a generic AI might suggest products, but it doesn't understand your brand's specific fits, fabrics, or customer preferences. When you train AI on your data—measurements, materials, purchase patterns, return reasons—it becomes an expert on your specific business.
Instead of saying "customers also bought this," your AI can say "Love our Sculpted Sleeve Blouse? Try the Origami Mini Dress—it has similar shoulder details customers often pair together." That's the difference between generic algorithms and branded experiences.
Let Your Experts Build AI Solutions
Your best people know your business better than any tech team. BBVA figured this out when they gave AI tools to all 125,000 employees instead of just IT. In five months, employees built over 2,900 custom AI tools. Their credit team uses AI to assess loans faster. Legal answers 40,000 policy questions a year with AI help. Customer service automatically analyzes survey feedback.
For retailers, this means empowering your category experts to build AI tools. Imagine your running shoe specialist can create training scenarios for store staff without writing code. They upload product specs for new trail running shoes, and AI generates realistic customer conversations that teach associates how to match shoe features to customer needs.
The people closest to your customers and products often have the best ideas for how AI can help. Give them tools to build solutions, and you'll get better results than top-down tech projects.
Remove Development Bottlenecks
Most companies are held back by how long it takes to build new capabilities. Mercado Libre solved this by creating an AI platform that lets their 17,000 developers build applications using plain language instead of complex code. Security and rules are built in automatically.
The results speak for themselves: they can now catalog 100 times more products, detect fraud with 99% accuracy, and automatically generate review summaries. Each project that used to take months now takes days or weeks.
For retailers, this means thinking about AI as infrastructure, not individual projects. Instead of building one AI tool at a time, create a platform that lets different teams solve their own problems quickly and safely.
Think Big About What You Can Automate
The most successful companies don't just improve existing processes—they reimagine them entirely. OpenAI's internal automation handles hundreds of thousands of tasks monthly by connecting AI to existing workflows. Support teams can instantly access customer data, write responses, and trigger actions like account updates or support tickets.
This works because they asked "How would we design this process from scratch?" instead of "How can we make this slightly better?"
Retailers can apply this thinking to inventory management, customer service, merchandising, and supply chain coordination. Most of these processes involve routine work that keeps smart people from focusing on strategy, relationships, and creative problem-solving.
Your Next Steps
These seven lessons work together as a system:
Start with clear measurement
Integrate AI into customer experiences
Begin now to build capabilities
Customize models for your specific needs
Empower domain experts
Create development infrastructure
Think big about automation possibilities
While some companies have a head start, AI is still early enough that smart moves today can create lasting advantages. Your competitive position tomorrow depends on the measurement frameworks, customer integrations, and organizational learning you build starting now.
The framework exists. The technology is available. The question is whether you'll use these lessons to lead your market or follow companies that started sooner.