When Klarna laid off 700 customer-service staff and replaced them with AI, the cost savings looked compelling – until customer frustration began to climb. This summer, the Commonwealth Bank of Australia followed a similar path, letting go of 45 roles only to rehire weeks later as call volumes surged.
These case studies are early signs of a wider phenomenon called the “boomerang workforce”, where companies shed roles in the race for AI-fuelled efficiency, only to find themselves rehiring once service dips and costs creep back in.
The rush to “AI everything” is reminiscent of the 1990s offshoring wave. Back then, banks, telcos and utilities scrambled to cut costs by moving call centres overseas. Within a few years, many reversed course, using on-shoring as a badge of superior customer experience. Today’s race to automate risks repeats the cycle, but only faster and with higher stakes for brand trust and regulatory exposure.
AI hiring and firing
Reducing headcount to boost efficiency looks clean on a balance sheet, but the true cost to brand and reputation often appears months later. As dissatisfied customers flood the channels, and unhappy users are more likely to leave poor reviews or share their frustration publicly, this amplifies the reputational impact. If companies scramble to bring talent back at a premium, recruitment costs start climbing too, and those they rehire may be less loyal than before.
Internally, culture suffers the most. Laying people off only to rehire shortly after shows management’s indecision and breaks employees’ confidence, both in leadership judgment and in the technology meant to replace them. It creates a sense among the workforce that, if the tech improves further, they’re definitely ‘out’ next time, making future hiring and retention harder. The distrust and anxiety can linger for years, harming employee engagement and reducing the willingness of staff to contribute to future transformation initiatives.
However, the problem rarely lies in technology alone. It is often the missing groundwork that causes the failure. Many organisations switch off human support too quickly, with no transition period where humans and AI operate side by side. AI systems frequently go live with messy or incomplete knowledge bases, meaning answers are usually inconsistent or incomplete. With missing scaling paths, customers are left in endless loops when they need a human answer.
Eat your greens before AI
Avoiding a boomerang workforce starts with full due diligence, or “eating your greens”. Companies should get their foundation policies, processes and data in order before unleashing automation at scale. They need to communicate changes clearly and early, engaging employees in the transition and offering reskilling opportunities rather than defaulting to redundancy. Otherwise, the business experiences the worst of both worlds: the cost of redundancy payouts and change programmes, followed by the cost of rehiring once service metrics slip.
It’s important to adhere to the human-in-the-loop models. When a query is sensitive or urgent, customers still expect to reach a person, and businesses that remove that option risk long-term loyalty. Therefore, maintaining hybrid teams, even at reduced capacity, can help provide a safety net during the learning curve. Organisations should carefully phase the AI rollouts, piloting the initiatives in a controlled environment, measuring them rigorously and refining them before reducing headcount, if necessary. This dual-running period may feel expensive, but it is significantly cheaper than reversing a failed deployment months later.
A 90-day C-suite plan
Boards and C-suites can act now to avoid the next headline about rushed layoffs. The first step is to pilot AI deployments in a small segment of the business and track key metrics such as response times, first-contact resolution and customer satisfaction. They can then share data from these pilots across the entire leadership team so that decisions are informed by actual outcomes.
The next step is to set clear guardrails for where automation is appropriate and where a human must step in. Teams should clearly define escalation policies and test them with real customers. Leaders must then iterate based on employee and customer feedback, building confidence in the new operating model before scaling further.
Finally, executives should model future talent needs, including the cost of rehiring and retraining. This workforce planning process should also consider reskilling pathways so that displaced employees have an opportunity to move into higher-value roles, reducing disruption and preserving operational knowledge.
In short: slow down to speed up. AI can deliver better and faster service, but only when it is introduced with discipline, clear governance and a plan that puts customer trust first.
Learning from the past
If the offshoring case study taught organisations anything, it was that the cheapest service is rarely the best service. Those who recognised this early used it to differentiate and win market share. The good news is that organisations have the same opportunity now. To avoid the boomerang effect, businesses should balance automation with human expertise and invest in reskilling rather than simply cutting. This will help build loyalty in the process.
AI is here to stay, but so is the need for empathy, judgment and trust. AI should not be used as a means to reduce headcount but as a tool to augment their people to be more efficient. Only then can we ensure that the next wave of transformation creates a more capable, loyal and resilient workforce.
Martin Colyer is the director of digital and AI at the HR consultancy LACE Partners
Read more: Reducing the burden on SOC teams requires context, context, and more context
