IKEA retrained the people AI replaced, and they became a €1.25bn sales channel

During the pandemic, IKEA’s customer service call centres were handling up to 3 times their normal volume, and most of it was the same handful of questions. What time does the store open? Is this sofa in stock? Where’s my order? Can I return this?

In 2021 IKEA brought in Billie, an AI customer service assistant built to answer exactly those questions, around the clock, without a person on the line. In its first two years Billie could deal with 47% of the customers who used it. It now handles 74%. The people it took work from were IKEA’s customer service agents, the staff who had been answering those calls. In most companies, that’s the point where the call centre gets smaller.

IKEA did something else. Ingka Group, which runs most IKEA stores, retrained around 8,500 call centre staff over 2 years, some to handle the complicated queries Billie can’t, and many to work as remote interior design advisers, helping customers plan a kitchen or a living room by phone or video. Fortune covered the programme in July.

That remote design channel is now IKEA’s fastest-growing way of selling. It made €1.25bn last financial year, up from €1.08bn, and it’s growing 15% to 20% a year. Its 24 remote sales centres helped 10 million customers, and IKEA’s own customer satisfaction score went from 60% before Billie to 89%.

Why do most companies cut instead?

I think it comes down to short-term cost pressure. Redundancies show up as a saving in the next set of accounts. Retraining shows up as a cost, and the revenue it might create is a year or two away, if it arrives at all.

Most companies also don’t have the people or the time spare. To do what IKEA did, you need the people to run the training, the volume of staff to make a new department worth building, and the patience to trial it. It takes a lot of training, and not everyone is going to succeed in a new role. Plenty of people would rather stay in their lane, and that’s fair.

And most businesses simply don’t have the options IKEA has. A furniture company with a customer base that size can open a design service and know there’s demand for it. A business with one product and one revenue stream has fewer places to move people to.

So I don’t think every company can copy this. But I do think most companies never ask the question, and that’s the bigger problem.

What should companies do differently?

The difference at IKEA wasn’t the chatbot. Lots of companies have one. It’s that someone decided, before the savings were banked, what the freed-up people and hours were going to be used for. The time Billie gave back went into the one job a bot can’t do well, which is sitting with a customer and helping them picture their home.

That decision rarely gets made because nobody owns it. AI arrives through IT, the saving lands with finance, and the people affected hear about it from HR. I think every business should have an AI committee, or an agentic committee if they’re building AI agents, with someone from the board on it. The question of where the freed-up capacity goes should be one of the first conversations it leads with, not something it gets to after the redundancy plan.

It’s the same trap I wrote about in the AI pilot that never lands: the tool works, nobody changes the job around it, and the gain disappears. And it fits what I’ve seen about where AI pays back first, which is speed rather than profit. Speed only becomes profit when you decide what to do with it.

IKEA isn’t a pure example. Ingka and Inter IKEA both cut hundreds of corporate roles this spring, though Fortune reports none were in the remote sales centres and none were put down to AI. The lesson still stands.

If the only plan for AI is the saving, that’s all you’ll ever get from it. IKEA got a channel worth €1.25bn a year.