Some of you may have already heard the IKEA story.
I went back to it recently as it’s one of the earliest AI transformation success stories, but I think we moved on from it too fast.
In 2021, IKEA deployed an AI chatbot called Billie to handle customer inquiries. Billie eventually took over nearly half of all incoming calls. That meant 8,500 customer service agents whose jobs no longer existed.
But IKEA didn’t lay them off. They recognized a $1.4 billion opportunity.
Instead, they studied what Billie couldn’t do. Customers kept asking for help with home planning and interior design. So IKEA reskilled all 8,500 agents as interior design consultants, equipped with generative AI to build tailored room layouts on demand.
A brand-new business line, built entirely from reskilled talent. And those agents had a natural head start: years of deep IKEA catalog knowledge made them a credible fit from day one.
It’s an amazing case study. But how can others replicate it, and actually pull off a reskilling effort at that scale?
How do you move thousands of people into a completely new role, across markets, at speed, and know whether the new skills are translating into real performance on the floor?
The training playbook most companies still use was designed for a different era.
For decades, reskilling meant one thing: take people off the floor, put them in a room, run them through a module, certify them, and send them back. (And, may I add, hope for the best.)
That model was designed for a world where skills changed slowly, products launched annually, and policies updated once a year.
That world is gone:
Skills now have a shelf life of months, not years. Products, policies, tools, and customer expectations shift every quarter.
AI has entirely changed the knowledge problem. Finding information is no longer the hard part. A frontline agent can get an answer in seconds. What they can’t do in seconds is apply the right judgment, in a live customer conversation, under pressure.
There’s no time for long sessions. Frontline teams now learn between calls and shifts. Learning has to fit into the flow of work.
The gap has moved. It’s no longer between agents and information. It's between information and behavior.
So what does the new model actually look like?
First: motivation before content. Before an agent engages with new material, they need to understand why it matters to them personally. Think about the “WIIFM” framework: what's in it for me. It’s the foundation of whether learning sticks or gets ignored.
Second: microlearning in the flow of work. 3-to-5-minute bursts between calls, not hour-long modules off the floor. The format has to fit both the shift and the people working it. If your floor is staffed by a Gen Z workforce raised on TikTok and YouTube, a 60-second interactive clip will land better than any slide deck.
Third: practice, not just certification. Passing a quiz proves nothing about what someone can do under pressure. AI role-play simulations let agents rehearse real customer conversations and get immediate, specific feedback. The gap between knowing and doing closes through practice and repetition, not testing.
And fourth: performance signals decide the lesson. When signals from the floor show where the gap is, the right intervention reaches the right person at the right time. The gap—not a compliance checklist—decides the next lesson.
Learning becomes a loop instead of a milestone.
Continuous reskilling also enables what I call “frictionless mobility.”
When agents build skills continuously, you accumulate a real-time picture of who’s ready for what.
Imagine this: surface the open role, let the agent preview what the job actually involves, and give them a way to try it (even with an AI simulation). For example, an agent considering a move from customer service into sales or back-office claims can rehearse interactions before applying.
This is the operational commitment most companies underestimate. The IKEA story gets celebrated for the boldness of the decision. But the decision was just the starting point.
The real work is building the infrastructure to move people into new roles continuously, not once. To know, in real time, whether new skills are translating into the right behaviors on the floor. To close the gap between what someone learned on Tuesday and what they’re doing on Wednesday.
That gap — between the announcement and the actual behavior change on the floor — is where most reskilling efforts quietly fall apart.
Did you find this article helpful? Share The Future of Work newsletter with a colleague!