Most businesses are investing in AI. Far fewer can prove that it is actually paying off.

An MIT study found that just five percent of generative AI projects are delivering real returns. Gartner predicts that 40 percent of projects will be canceled by 2027 because of unclear business value, while BCG says only five percent of companies worldwide have reached the point where they are consistently generating substantial value from AI.

Those numbers paint a pretty bleak picture. Yet some companies are already using AI to cut costs, increase revenue, improve customer experience and solve problems that would have been extremely difficult to tackle in other ways.

So what are they doing differently?

From Walmart and IKEA to Bank of America and UPS, here are eight companies generating measurable value from AI, and the lessons their success offers to everyone else.

The US retail giant deployed a generative AI tool to its product team in 2024. Its aim was to improve the quality of its product catalogs; internal databases used for everything from helping customers search available items to managing logistics and inventories. This involved creating or updating 850 million data points . Completing this manually would have involved expanding the size of the team tenfold. Thanks to their AI, the task was completed on time, delivering improved customer and employee experience, in-store and online.

UK energy supplier Octopus has rolled out a number of generative AI initiatives through its Kraken platform. They include Arlo , an autonomous agentic platform built for handling routine customer service inquiries, credited with lifting customer satisfaction rates from 73 to 76 percent. The success of Arlo and other genAI projects prompted the company to spin out Kraken as an independent entity, providing AI services to other energy companies and credited with generating £422 million in revenue in 2025.

Latin America’s largest e-commerce provider used tools built around OpenAI’s GPT-4 to screen online product listings for potential fraud and counterfeiting. Its system checks every product listed against more than 5,000 variables to determine their authenticity in under a second, protecting customers from online scams or buying fake products. This has enabled it to achieve fraud detection accuracy of almost 99 percent. Another genAI system enabled it to increase the speed at which it can catalog products by 100x over a two-year period.

The largest food delivery platform in the US worked with AWS to build a voice-activated solutions portal for customers, merchants and delivery drivers. A specific problem they aimed to solve was the need for drivers to access help while on the job without having to stop making deliveries. This was achieved by cutting the number of calls requiring transfer to human agents by 49 percent. The Customer Connect system now handles hundreds of thousands of calls every day through a self-service portal credited with achieving $3 million in year-on-year operational savings.

The US banking giant was one of the first businesses to launch a dedicated customer service chatbot, way back in 2018 . Erica helps customers manage their accounts and get quick answers to questions instead of being left waiting on hold. In 2025, it surpassed three billion client interactions and was credited with cutting the number of service desk calls by 50 percent, meaning BoA’s human agents could spend their time on trickier or sensitive issues requiring an expert human touch.

When the Swedish furniture designer deployed its own AI customer assistant, Billie, it was able to autonomously resolve 47 percent of incoming queries. However, rather than downsizing its human workforce, it retrained 8,500 customer service agents as remote interior design consultants. By offering this new service to customers, it was able to generate $1.4 billion in new sales.

Radisson used AI to create and optimize digital advertising for its hotels, automatically adapting images and messaging for different languages, markets and cultural preferences. This cut time spent on creating ad material by 50 percent while also driving a 22 percent increase in ad-driven revenue and improved ROI on ad spend by 35 percent.

Changes to global tariff regimes in 2025, specifically around US imports, have forced many delivery and logistics companies to adopt new approaches in order to avoid long delays. UPS’s response to this challenge was to integrate agentic AI into its brokerage and documentation systems . In June 2026, it announced it had increased the number of small packages clearing customs in one day without manual intervention from just 21 percent to 97 percent, delivering solid customer experience benefits.

So, What Did They Do Differently?

The common thread running through these success stories is surprisingly simple: they started with a business problem, not with AI.

Walmart needed to improve hundreds of millions of product data points. DoorDash wanted to reduce the burden on customer support. Mercado Libre needed to detect fraud faster. UPS had to respond to a sudden increase in customs complexity.

In each case, AI was applied to a clearly defined challenge where success could be measured in revenue, cost savings, productivity, customer satisfaction or speed.

That is an important lesson at a time when so many AI initiatives are struggling to demonstrate value. The companies getting the strongest returns are not asking, “Where can we use AI?” They are asking, “What important business problems could AI help us solve better?”

For business leaders, that may be the most useful place to start. Find the problem, define what success looks like, and then decide whether AI is the best tool for solving it. That approach will dramatically improve the odds that your AI investments end up among the success stories rather than the 95 percent that fail to deliver.