The Rehire Wave

In 2023, IKEA didn’t fire 8,500 call-center workers when its AI went live. It retrained them into a new revenue line that now does €1.3 billion a year.

Three years ago, when Ingka Group deployed a customer-service AI called Billie, it had a decision to make about the roughly 8,500 people whose inbound-inquiry work Billie was designed to eat. The rest of the industry was already writing the script in ink — fire the humans, book the savings, applaud the margin — and no one would have been surprised to see IKEA follow the script.

Instead, Ingka retrained the whole cohort into what it called a Remote Interior Design Service, aimed at customers whose questions were less “where is my order” and more “what should this room look like.” The retraining took months, not slides.

Fast-forward. Billie now handles about 74% of inbound inquiries — up from 47% in year one — for around €15 million a year in operational savings. The Remote Interior Design Service generated €1.3 billion in FY22 sales, or 3.3% of Ingka’s total revenue. Customer-happiness scores rose from 60 to 89. That is not a slide-deck outcome; it is a P&L outcome.

Meanwhile, everyone else

While Ingka was quietly running the math, most of the market wasn’t. Robert Half’s April 2026 survey of U.S. hiring managers found that 32% who eliminated a role citing AI have since rehired the same or a similar one — sometimes spending more on the rehire than they saved by cutting. Finance is leading the reversal at 44%; HR at 35%; tech at 32%. A separate Inc. survey put the leaders-regret number at 55%.

The named cases are getting concrete. Ford has rehired 350 veteran engineers over three years after automation quietly failed on quality. IBM has committed to tripling U.S. entry-level hiring in 2026 after realizing you cannot promote a mid-level engineer you never hired as a junior. Gartner has since found that 22% of firms cut entry-level hiring for AI in the past year — the very talent pipeline the rehire wave is now scrambling to rebuild. Klarna, familiar from earlier columns, quietly rehired humans after its 700-agent AI swap sagged on satisfaction.

The math they skipped

The difference between Ingka’s outcome and the 32% club is not virtue. It is arithmetic.

Ingka measured two things when Billie went live: the money Billie would save, and the money the redirected labor could earn on a different demand curve. The 32% club measured one. When you count only the labor-cost line, AI adoption produces a single number and it always looks like a win. When you count both, you find out whether the humans you eliminated were sitting on a business you never sold.

Interior-design consulting is not, on its face, an obvious pivot for a call-center agent. Ingka didn’t stumble into it; it looked at what customers were already asking for that Billie couldn’t do, and moved the humans there. That is the operational move: identify what AI absorbs; identify what the market wants that AI can’t yet supply; connect the two.

Two honest complications

None of this makes IKEA a saint. Ingka has, in the last year, made a round of unrelated 2026 layoffs — in different business units, not touching the reskilled cohort. The reskill program worked on its terms; the parent group has other cost pressures. Both statements are true, and readers should keep them separate.

IKEA also did not offshore. It kept the whole cohort onshore in the U.K. Everyone in the OA reader universe should notice the caveat. The lesson is not “onshore rather than offshore”; it is that whoever runs your customer-service delivery needs to be doing Ingka’s math, not the 32% club’s — regardless of where they sit.

What this means for you

For BPO buyers, this is the buy-side version of TP’s 500,000-seat AI-enablement pledge three weeks ago: the operator is claiming to run the two-column math instead of the one-column math. Your job is to test the claim.

The right question at your next renewal is not “will AI reduce my contact-center bill.” It is: “when your operator retires the volume AI absorbs, where does the labor go and what does it earn?” A provider who cannot answer that is a provider running the 32% club’s math on your account.

The question for your business

Is your AI transformation running two ledgers — one for costs, one for the revenue the freed capacity could unlock?

Read more thought leadership articles here:

• • The Warning Arrived Late
  • The Uninvestible Files Back
  • The Great Call Center Gold Rush