Most companies are pouring money into artificial intelligence, yet only a tiny fraction can prove it works. A study from MIT found that just 5% of generative AI projects deliver real returns. Gartner predicts that 40% of AI projects will be canceled by 2027 due to unclear business value. And Boston Consulting Group reports that only 5% of companies worldwide consistently generate substantial value from AI.
The gap between hype and reality is wide. But a small group of companies, spanning retail, energy, e-commerce, food delivery, banking, furniture, hospitality, and logistics, is showing what disciplined AI adoption looks like. Their common thread is simple: they started with a business problem, not with the technology.
The Problem-First Approach
Successful companies did not ask what AI could do. They asked a sharper question. Bernard Marr, a contributor at Forbes, framed it this way: "What important business problems could AI help us solve better?"
That question changes everything. Instead of hunting for use cases, these firms identified pain points where success could be measured in revenue, cost savings, productivity, customer satisfaction, or speed. Then they applied AI to those specific challenges.
The approach dramatically improves the odds of success. Companies that skip this step often end up with pilots that never scale or tools that nobody uses. The eight companies profiled here avoided that trap by tying every AI deployment to a concrete metric.
Retail and E-Commerce: Walmart, Mercado Libre, and DoorDash
Walmart, the US retail giant, deployed a generative AI tool to its product team in 2024. The task was massive: improving the quality of its product catalog. The AI created or updated 850 million data points.
Doing that manually would have required expanding the team tenfold. Instead, Walmart let the AI handle the heavy lifting while employees focused on exceptions and quality checks. The result was a faster, cheaper way to keep millions of product listings accurate and complete.
The lesson is not about the technology. It is about the scale of the problem. Walmart had a catalog that was too large for humans to manage efficiently. AI was the only realistic answer.
Latin America's largest e-commerce provider, Mercado Libre, used OpenAI's GPT-4 to screen product listings for fraud. The system checks every product against over 5,000 variables in under a second.
The accuracy is striking. Mercado Libre achieved fraud detection accuracy of almost 99%. That means nearly every fraudulent listing gets caught before it reaches a buyer.
The speed gain was even more dramatic. Mercado Libre increased its cataloging speed by 100x over two years. Fraud screening that used to take human reviewers hours now happens instantly, at scale, without slowing down the marketplace.
DoorDash, the largest food delivery platform in the US, built a voice-activated solutions portal with AWS. The system, called Customer Connect, handles hundreds of thousands of calls daily.
The impact on support operations is clear. Customer Connect reduced calls requiring transfer to human agents by 49%. Fewer transfers mean shorter wait times and lower staffing pressure.
The financial result followed. DoorDash achieved $3 million in year-on-year operational savings. That is a direct, measurable return on an AI investment, not a vague promise of future efficiency.
Energy and Banking: Octopus Energy and Bank of America
Octopus Energy, the UK energy supplier, took a similar path. The company rolled out generative AI initiatives through its Kraken platform, including an autonomous agentic platform called Arlo. Arlo handles routine customer service inquiries, and the impact shows up in the numbers.
Customer satisfaction at Octopus Energy rose from 73% to 76% after Arlo was introduced. That is a modest shift, but it happened while the company was handling growing volumes of customer contacts.
Kraken itself became a business. In 2025, the platform generated £422 million in revenue after being spun out as an independent entity. Octopus Energy turned an internal tool into a product that other companies can license, which is a different kind of AI payoff.
Stay ahead of the AI curve
The most important updates, news, and content — delivered weekly.
No spam. Unsubscribe anytime.
Bank of America launched its Erica chatbot in 2018. That was early for a major bank, and the bet has paid off. Erica helps customers manage accounts, answer questions, and complete routine tasks.
By 2025, Erica surpassed three billion client interactions. The scale is enormous, but the more important number is the cost side. Erica cut service desk calls by 50%.
That is a permanent reduction in operating costs. Every call that Erica handles instead of a human agent is money saved. The bank did not need to hire thousands of additional service representatives to keep up with customer demand.
Furniture, Hospitality, and Logistics: IKEA, Radisson, and UPS
IKEA, the Swedish furniture designer, deployed an AI customer assistant called Billie. Billie autonomously resolves 47% of incoming queries. That is nearly half of all customer contacts handled without human involvement.
But IKEA did something unusual. Instead of laying off the displaced workers, the company retrained 8,500 customer service agents as remote interior design consultants.
That decision created a new revenue stream. IKEA generated $1.4 billion in new sales from the new interior design service. The AI did not just cut costs. It freed up people to do higher-value work that generated new business.
Radisson Hotel Group used AI to create and optimize digital advertising. The hotel group cut time on ad material by 50%. Campaigns that used to take weeks now come together in days.
The revenue impact followed. Radisson saw a 22% increase in ad-driven revenue. The return on ad spend improved by 35%.
These are not marginal gains. They are the kind of numbers that justify an AI budget line item. Radisson did not need a grand AI strategy. It needed better ads, faster, and AI delivered that.
UPS, the delivery and logistics company, integrated agentic AI into its brokerage and documentation systems. The trigger was external: changes to global tariff regimes in 2025, specifically US imports, forced logistics companies to adapt quickly.
The AI now handles customs changes automatically. The results are dramatic. In June 2026, UPS increased the share of small packages clearing customs in one day without manual intervention from 21% to 97%.
That is a fourfold improvement in a process that directly affects customer satisfaction and operational cost. When tariffs change, the AI updates documentation and clears shipments without waiting for human review.
What Separates Winners from Losers
The pattern across all eight companies is consistent. They found a specific problem, defined what success looked like, and then applied AI to that problem. They did not start with a shiny model and look for a use case.
Success was measured in hard numbers. Revenue, cost savings, productivity, customer satisfaction, or speed. Every company in this group can point to a metric that moved because of AI.
The broader market tells a different story. Most AI projects fail to deliver value, and the cancellation rate is expected to climb. Gartner's prediction of 40% cancellations by 2027 suggests that many companies will keep making the same mistake.
The advice for business leaders is straightforward. Find the problem first. Define success in measurable terms. Then decide if AI is the best tool for the job. Sometimes it will be. Sometimes a simpler solution will work better.
The eight companies here prove that AI can pay off when it is aimed at the right target. The technology is not magic. It is a tool, and tools work best when they are used for the job they were built to do.