AI in Customer Service Should Build Better Jobs, Not Cut Them

Every customer service AI project eventually runs into the same awkward question. If a bot can answer a large share of customer questions, what happens to the people who used to answer them?

A lot of companies answer that question too quickly. They look at the contact centre, count the repetitive enquiries, work out how much volume an AI system can absorb, then start talking about “efficiencies”. The word sounds tidy. In practice, it can mean fewer people taking on harder work, while customers get pushed through more automated layers before reaching someone who can actually help.

That is a poor way to use the technology.

The wrong efficiency question

AI can reduce the weight of repetitive customer service work. It can answer simple questions quickly. It can pull information from help centres, account notes and product pages faster than a person can. Used well, it can make support teams less stretched and make customers less dependent on long queues.

The problem starts when leaders treat that capacity as a reason to cut staff. It assumes customer service is mainly a volume problem, when most service leaders know the harder truth. The easy questions were never the full job. The real work begins when the answer is unclear, the customer is frustrated, the policy does not quite fit, or the system has already failed once.

AI changes the shape of customer service work. It does not remove the need for people who understand customers.

What IKEA’s public example shows

IKEA is one of the better-known examples of AI being used alongside customer service reskilling.

Ingka Group, the largest IKEA retailer, introduced an AI chatbot called Billie to handle common customer enquiries. Billie dealt with questions about orders, products and availability, the kind of work that can fill up a contact centre very quickly. Ingka said that between 2021 and 2023, Billie handled roughly 3.2 million customer interactions and resolved about 47 percent of the enquiries it received. Ingka also reported close to €13 million in savings from the chatbot.

That could easily be read as a standard automation story. A large chatbot handles millions of conversations, the business saves money, and the next discussion becomes staffing.

The stronger part of the example is what Ingka said happened around the customer service workforce.

Ingka said 8,500 call centre co-workers had been reskilled to gain different capabilities, including remote interior design competence, digital retail sales, relationship-building and handling more complex customer enquiries. That is a more careful way to describe the case than saying IKEA simply moved every worker into one new job, or that AI replaced a set number of roles.

These were people who already knew IKEA’s products, customer questions and service pressure. They understood what customers were asking for when they could not explain it neatly. They knew which details mattered during a home design conversation because they had spent years listening to customers describe real problems in real homes.

That experience was useful. It just needed to be moved into work with more value.

With Billie handling simpler enquiries, Ingka described co-workers as being able to play a more value-adding role within remote selling. Customers could still get quick answers when the issue was simple, while staff were used for work that needed judgement and product knowledge.

That shift also had commercial weight. Ingka said its remote customer meeting points reached €1.3 billion in sales by the end of FY22, equal to 3.3 percent of total sales. The safest reading is that AI, remote selling and reskilling formed part of the same operating shift. It would be too strong to claim the chatbot alone created that revenue.

This is the part of the IKEA example that customer service leaders should take seriously. The available evidence shows a customer service operation where AI handled a large share of simpler enquiries, while thousands of co-workers were reskilled into work Ingka described as more value-adding.

That is a much stronger ambition than deflection alone.

Support knowledge is business knowledge

Most support teams already hold a huge amount of customer knowledge. Agents know which product pages are confusing. They know where customers get stuck during checkout. They know which policies sound fine internally but create irritation in a real conversation. They know when a customer is asking for a refund because they are angry, and when they are asking because the company has made the same mistake twice.

That knowledge rarely shows up properly in dashboards. It gets buried inside ticket notes, call summaries and team chats. When AI is added carefully, it can help surface those patterns. It can show where customers need clearer information and where automation is becoming another obstacle.

The people closest to the queue are the ones who can read those signals properly.

This is why laying off support staff after introducing AI is often shortsighted. It can remove the workers who understand what the system is doing to customers. It can also leave the remaining team with the more complex work, usually with less breathing room.

The human queue gets harder

There is a real risk here. When a chatbot answers the simple questions, the human queue can become heavier. Agents receive more of the conversations that could not be solved by the automated layer. Those customers may already be annoyed. Some will have repeated themselves. Some will arrive with a problem that the bot misread or handled too slowly. The work becomes more emotionally demanding, even if the number of total contacts goes down.

A company that cuts staff at that exact moment is creating pressure in the wrong place.

The headline metrics can still look impressive. Containment may rise. Average handle time may appear to improve in some parts of the operation. Cost per contact may drop. Yet customers may feel less supported, and agents may feel as though the business has taken away the easy work while leaving them with the stress.

That is a management choice.

A better approach starts earlier, before the chatbot goes live. Support staff should be part of deciding what the AI handles, where it escalates and how it explains itself. They can identify which questions are safe to automate and which ones need a human earlier than the system might realise. They can help write responses in language customers actually use. They can also test the strange or emotionally charged cases that tend to expose weak automation.

The stronger path

Once AI is live, support roles should evolve around the new work being created. Some agents may move into customer advisory roles, as Ingka’s public example suggests. Others may become AI quality reviewers, knowledge base owners or escalation specialists. Some may move closer to product and operations, using customer conversations to identify the gaps that keep driving contact.

This kind of redesign requires investment. It is easier to announce a chatbot than to build proper career paths around one. It is easier to measure saved contacts than to measure the value of better advice. It is easier to talk about productivity than to ask what kind of service experience the company is actually building.

Still, companies that treat AI mainly as a workforce replacement may end up with a thinner service model. Customers will notice the gaps. Staff will absorb more of the harder moments. Over time, the brand can start to feel the difference.

IKEA’s example works best when it is described carefully. It does not prove that AI prevents job cuts, and it should not be used to make broad claims about IKEA’s workforce decisions across the whole business. What it does show is more specific and still valuable: in this customer service case, Ingka publicly linked chatbot adoption with large-scale reskilling, remote advisory work and a growing remote sales channel.

Customer service has spent years being treated as a place to control costs. AI gives companies a chance to shift that thinking. The technology can take pressure out of the queue, but leadership decides what happens next. One path leaves customers with fewer humans and staff with less security. The stronger path uses automation to clear space for better human work.

That is where customer service AI should be heading. Bots that handle the repeatable work. People trained for the conversations where judgement matters. A service model that does not treat every saved contact as a saved salary.

Sources

Bellé, A. (2026, June 9). How IKEA turned a €13 million chatbot into a €1.3 billion business. CIO. https://www.cio.com/article/4180896/how-ikea-turned-a-chatbot-with-13-million-users-into-a-business-worth-1-3-million.html

Davis, S. (2026, June 23). Instead of laying off 8,500 workers because of AI, Ikea used this radical leadership playbook to grow revenue. Inc. https://www.inc.com/stephanie-davis/layoffs-workers-ai-ikea-leadership-playbook-grow-revenue/91364108

Dhar, P. (2026, April 1). How IKEA turned AI ‘failures’ into €1.3 billion in revenue. Fluent Support. https://fluentsupport.com/how-ikea-turned-ai-failures-into-1-billion-in-revenue/

Ingka Group. (2023, June 29). AI and remote selling bring IKEA design expertise to the many. https://www.ingka.com/newsroom/ai-and-remote-selling-bring-ikea-design-expertise-to-the-many/

Reid, H. (2023, June 13). IKEA bets on remote interior design as AI changes sales strategy. Reuters. https://www.reuters.com/technology/ikea-bets-remote-interior-design-ai-changes-sales-strategy-2023-06-13/