In 2021, Ingka Group, the company that runs most IKEA stores, switched on a chatbot named Billie. The goal was ordinary. Billie would field the questions that pile up at any retailer: is this in stock, when does it ship, where is my order.
It worked. Over the next two years Billie handled 3.2 million conversations and closed out roughly 47 percent of them on its own. That saved the company around 13 million euros and took a load of repetitive work off the call center floor.
A machine was now doing the work of 8,500 people. So what happens to the 8,500 people? The math practically writes itself, and most companies let it. You point at the savings, trim the team, and move on to the next project.
IKEA went the other way.
Instead of staring at the 47 percent Billie could answer, they looked hard at the 53 percent it couldn't. When they sorted through those failed conversations, they found something a better model was never going to fix. People weren't calling to check whether a sofa was in stock. They were calling because they wanted to know if it would fit their living room, and they wanted a person to tell them.
That is not a support problem. That is a sales conversation nobody at IKEA had been set up to have. So the company retrained those 8,500 call center workers as remote interior design advisors. The thing that had been a help desk became a sales channel, run by thousands of people over video and phone. In its first full year it brought in 1.3 billion euros, about 3.3 percent of IKEA's revenue, and the company has said it wants that figure near 10 percent by 2028.
Put the two numbers side by side. Looking inward for efficiency saved 13 million. Looking outward for an opening earned 1.3 billion. Roughly a hundred times more, which is why IKEA barely bothers to mention the original savings anymore.
The number nobody asks for
Most AI projects get measured on the AI. How many tickets did it close, how much did each interaction cost, how many hours did it hand back. Those are the numbers leadership asks for and the vendor is happy to supply. They are also the numbers that keep you looking in exactly the wrong direction.
Not what the tool handled, but what it exposed. For IKEA, the chatbot quietly surfaced a want customers had carried for years and the company had never named: they wanted design help, and they would pay for it.
You can find a version of this almost anywhere. The pattern travels; the specifics change by industry.
Automate the support desk
Take the first line off the queue and you start seeing the deeper problems that were buried under password resets.
Hand off routine review
Give contract review to software and a legal team suddenly has hours for the business questions it never got to.
Spotting it is the easy half
Finding the opening is one thing. Deciding to act on it is where almost everyone quietly bails.
Cut and bank it
Trim the staff the tool made redundant and book the savings. No committee argues with it. It is sitting right there in the report.
Redeploy and invest
Move 8,500 people into a service that barely existed, and pour money into it - straight against the savings on the page. That is a real bet.
What made it possible at IKEA was not a lone visionary who saw the future. It was a setup where the question could even come up.
The head of technology and the head of people were at the same table, not filing reports to two separate committees. Plenty of companies have an AI task force. Very few have that.
Where this leaves the people running it
For anyone leading technology in a company, this is the fork.
Do what's expected
Ship the tool, measure it, report the savings, line up the next use case. Reasonable, and forgettable.
Keep asking what it turns up
What need shows itself once the software takes the routine work. What skills sit unused in your own staff because nobody ever asked for them.
You don't need a new committee to start. You need to raise the question in the next meeting that's already on the calendar. The first approach treats AI as a technical job. The second treats it as a business one that happens to land on the tech leader's desk before anyone else's.
IKEA's story tends to get told as a clever save, a cost center flipped into a profit center. The more useful read is smaller and harder to copy. They built a room where someone could notice what the machine had missed, and then they were willing to act on it.
“The tool found the gap. The people decided what to do with it. That second part is still the whole game.
What is your AI quietly exposing?
The savings are the easy number. The opening the tool reveals is the one worth chasing - and acting on. Get a free expert review of your AI idea or roadmap, read by our Chief AI Officer, focused on where the real value is.
Get your free expert report