Two years after brands rushed to replace customer service teams and creatives with AI, the promised savings aren't materialising. Companies like Klarna, Commonwealth Bank and Ford cut staff for algorithms, then quietly rehired many of the humans they let go - often at double the cost.
The corporate backtracking comes as the industry confronts what some investors have been warning for months: running large AI models at scale is commercially unsustainable for everyday business tasks. Analysis from Sequoia Capital puts global AI infrastructure spending at $600 billion annually, with revenue and productivity gains failing to match that investment.
The AI boomerang
The pattern is now common enough to have a name: the "AI Boomerang." Data from Robert Half shows 34% of professional services companies that cut staff for AI have already reinstated those positions. Among those employers, 75% found the combined cost of AI implementation plus rehiring and retraining erased their initial savings entirely.
Klarna froze hiring and replaced customer service agents with an AI chatbot, then watched satisfaction scores plummet. By mid-2025, its CEO publicly admitted the AI lacked nuance and empathy, forcing a pivot back to human agents.
The Commonwealth Bank of Australia's attempt to replace dozens of service roles with an AI voice bot ended in higher call volumes and mounting pressure on remaining human teams. Ford, which replaced veteran engineers with automated systems, saw product recall rates climb. This year, it quietly rehired hundreds of human experts.
The economics don't favour automation for everyday work. Many AI pricing models charge per token or per query, and costs spiral at scale. Across software subscriptions, cloud infrastructure and human oversight, running the AI often costs more than the original workforce.
Targeted tool, not blanket replacment
Enterprise AI success stories rarely involve a full business overhaul. They are specific: transcription, level-one customer service triage, basic code generation, photo and video editing shortcuts. These efficiencies are real, and they're incremental.
That's why some companies are adopting a "controlled automation" model. Yango, a global tech company, uses AI for repetitive tasks - data processing, media buying execution - but treats it as an operational assistant, not a decision-maker. In many operational scenarios, a bespoke AI solution is more expensive and less reliable than a standard SaaS platform or a human expert. Marketing managers interested in what this looks like in practice can check the AI Learning Path for Marketing Managers.
"Technology scales, but thinking does not." That is the lesson, and it explains why brands are gravitating toward off-the-shelf tools and smaller, task-specific models rather than massive frontier systems. Traditional search engines, for instance, are being replaced by ad-supported AI discovery engines running SLMs - the only way platforms can afford to keep search free.
Why this matters for marketing professionals
AI has become a permanent part of the marketing stack, but it has not become a marketing manager. It still requires a person to set the strategy, judge the output and own the customer relationship. Marketing professionals who succeed will be those who use AI for the heavy lifting and spend their time on judgment and creative work.
For marketing teams assessing their tooling, the takeaway is to make the cost analysis painful. The quotes, workflows and ROI claims that sounded cheap a year ago may be the same ones quietly spiraling on invoices today. And for skills: the demand is shifting to people who can and oversee AI workflows - a role that employers are already rehiring for. For those building that skill set, AI for Marketing resources are a useful starting point.
Your membership also unlocks:
