IKEA let a bot take the routine queries and moved 8,500 staff to sales

Ingka's Billie chatbot resolved about 47% of enquiries from 2021 to 2023. The people it freed now sell remote interior design. What a smaller retailer can copy.

Ingka Group, the largest IKEA retailer, put an AI chatbot called Billie in front of its customer contact centres and reported in June 2023 that it had resolved approximately 47% of the enquiries it received between 2021 and 2023, about 3.2 million interactions, saving nearly EUR 13 million. Rather than cut its 8,500 call-centre co-workers, it retrained them to sell remote interior design, a channel that produced EUR 1.3 billion of sales by the end of its 2022 financial year.

That is one of the most detailed written records a large retailer has published of what happens after a bot takes the routine questions. It matters this week because the software industry has just told every retailer that the bot is about to become the front door.

Why this is the question now

At Dreamforce this week Salesforce presented AIforce, which it describes as a layer that lets people reach its system from whatever AI interface they already use rather than from Salesforce's own screens. Strip away the platform strategy and the message to a retailer is simple: more of your customers' questions will be answered by software, in channels you do not design. The question an operations manager actually has to answer is not whether to buy that. It is what the people who answer those questions today will do instead.

Ingka is a rare retailer that has put its answer in writing, with numbers, in its own newsroom.

What Ingka actually did

Billie is a chatbot that answers customer questions, gives product information and helps with planning, at any hour. On its own that is unremarkable. The decision that made it worth writing about was the second one: the contact-centre co-workers whose routine work Billie absorbed were reskilled rather than released. Ingka says 8,500 of them were trained in remote interior design, digital sales, relationship building and the complex problems a bot cannot close.

Those people now staff what Ingka calls remote customer meeting points, where a customer books time with an adviser to plan a kitchen or a room remotely. By the end of FY22 that channel accounted for EUR 1.3 billion of sales, 3.3% of the group total, and Ingka said it aimed to take the share to 10%.

The bot saved Ingka about 13 million euros. The channel its retrained staff work in sold 1.3 billion.

Those two numbers are not directly comparable: the savings figure covers roughly two years of bot activity, and the sales figure covers the whole remote channel, not only the work of the retrained staff. But the order of magnitude is the point. The cost saving was the small prize.

What did it cost, and what did it save?

Ingka has not published what Billie or the retraining cost. What it has published is the saving: nearly EUR 13 million across about 3.2 million resolved interactions. Our arithmetic from those two figures puts the saving at roughly four euros per interaction the bot closed. That is a useful benchmark for anyone pricing a customer-service agent against a contact-centre cost per contact.

The governance cost is visible too. In its FY25 annual summary Ingka says co-workers must be trained on responsible AI use before they can use its internal AI portal, that it extended AI literacy training to a further 6,886 co-workers, bringing the total to 11,062 since FY23, and that it treats some uses the EU AI Act classes as high-risk as prohibited inside the company. None of that is free, and a retailer copying the model should budget for it.

A second, faster version: Falabella on WhatsApp

Grupo Falabella, the Chilean retail group, shows what the same idea looks like when it is built on a vendor platform in 2025 rather than several years earlier. According to the Salesforce customer story published in December 2025, in which named Falabella managers are quoted, its service agent on WhatsApp answers common questions and gives live order status by connecting to the e-commerce system. It resolves 60% of WhatsApp inquiries without a person. It went from build to production in nine weeks, and a quarter of its conversations happen outside business hours.

The part worth copying is the hand-off. When a request is too complex, the agent opens a case and passes the full conversation to a human, so the customer does not start again. Falabella's e-commerce services director makes the point in the case study that returning a refrigerator involves far more policy than returning a T-shirt. That is exactly the work that should stay with people.

Does this apply to a 200-person retailer?

Partly. The bot half now applies to almost any retailer with an online order flow, because the tools are sold off the shelf and Falabella's build took nine weeks. The redeployment half only works if you have a higher-value job for freed staff to move into, the way Ingka had remote design selling. Without one, the saving is just a smaller contact-centre bill.

For a chain of twenty stores that might mean moving two or three service staff onto outbound work: quotes for larger orders, delivery and assembly booking, or following up abandoned baskets. The volume is smaller than Ingka's, but the logic is the same.

What it does not mean

It does not mean Ingka has ruled out job cuts. In March 2026 it said around 800 roles in its group functions may become redundant, while it plans up to 20 new stores and 500 new roles. Its own announcement frames the move as simplification and does not cite AI, so we do not read the cuts as a verdict on Billie either way.

It also does not mean a bot will lift your satisfaction scores on its own. Neither company's published figures isolate the bot's effect on sales, and yours will not either unless you measure it.

What to do next

Before you buy an agent, write down three lists: the questions your team answers most often, the questions that need judgement, and the higher-value work your people would do with the time. If the third list is empty, the project is a cost saving and should be priced as one. If it is not, you have Ingka's business case.

Correction policy: if any figure here is shown to be wrong, we will correct it at the top of this piece, dated.