In June 2025, Gartner released a report that should have alarmed every executive banking on AI to replace customer service staff: half of the companies that had committed to eliminating human customer service agents were quietly abandoning those plans.
The research covered 163 customer service leaders. The finding was unambiguous. After months of hyper-investment in AI-driven automation, companies realized they'd made a strategic miscalculation. The cost savings consultants promised weren't materializing. Customer satisfaction was collapsing. And the legal liability was becoming real.
This is the story nobody wanted to tell: the great AI customer service bet is failing for most companies that tried to go all-in on it.
The Promise vs. The Reality
The pitch was simple and seductive. Replace expensive human agents with AI chatbots. Cut per-contact costs from $13.50 (agent-assisted) to $1-3 (AI-native platforms). Eliminate 10,000 positions. Watch profit margins expand. British Telecom planned exactly this, announcing plans to eliminate 10,000 customer service representatives by 2030.
The problem: customers hated it.
A Verizon CX survey of 5,000 consumers found an 28-point satisfaction gap between human and AI support. Humans scored 88% satisfaction. AI scored 60%. That gap isn't noise—it's a customer retention problem. According to Acquire BPO research, 70% of consumers would switch brands after just one poor AI experience.
The contradiction is stark: 88% of contact centers now use AI in some form, but only 25% have actually integrated it into daily operations. Most deployments are experiments that haven't scaled. And among those that have scaled, the results are mixed at best.
The Trust Collapse
Here's what Gartner found about customer trust: 51% of customers trust human agents most. Only 7% trust AI most for resolution.
Yet 62% of customers also express concern that AI will make it harder to reach a human agent. This reveals the actual customer sentiment: people want AI to help, not to replace. They want speed and convenience, but not at the cost of losing access to someone who can actually solve problems.
The data shows a bifurcation emerging. Companies using AI for tier-1 support (simple questions, routing, basic troubleshooting) while keeping humans in the loop for complex issues are seeing 65% of problems resolved without human escalation. That's functional. But companies trying to push AI-only solutions are hitting walls.
Consider the numbers on consumer preference: 61% of new buyers choose faster AI responses over waiting for humans. That sounds like a win for automation. But flip it: 77% find chatbots frustrating, and 88% still prefer humans when it matters. Customers want speed for routine stuff and humans for everything else. Most AI deployments can't tell the difference.
When AI Hallucinates, Customers Leave
The real liability emerged in April 2025 when Cursor's AI support bot "Sam" invented a non-existent policy about one-device-per-subscription limits. The bot fabricated it entirely. Users canceled subscriptions. The incident exploded online.
Cursor isn't alone. Air Canada faced a tribunal ruling in February 2024 that held the company legally responsible for a chatbot's invented refund policy. The precedent is set: if your AI makes promises it can't keep, you're liable.
This is the hidden cost nobody factored into the ROI calculations. One hallucination can trigger brand damage, customer abandonment, and legal exposure. The per-contact cost savings evaporate when you're processing chargebacks and managing reputation damage.
The Winners: Hybrid Models
The companies winning with AI customer service aren't eliminating humans—they're augmenting them. Sephora is the clearest case study. Their AI chatbot resolves 75% of daily inquiries without human intervention. Response time dropped from minutes to under 10 seconds. Cart abandonment fell 18%. Operational costs fell 20%. Response time improved 40%.
But Sephora kept human agents. The AI handles tier-1 (product questions, order status, returns). Humans handle everything else—complex complaints, VIP customers, nuanced beauty advice that requires expertise.
This is the model that actually works. And it's the opposite of what most companies were planning.
The Cost Mirage
Here's why half the companies are backing off: they underestimated total cost of ownership. Gartner analyst Brian Weber put it bluntly: "Customer service and support leaders expect cost savings from generative AI, but they often underestimate the total cost of ownership, making savings difficult to realize."
The math looks good on paper: $1.84 per self-service contact vs. $13.50 for agent-assisted. But that assumes your AI actually works. Training it doesn't. Monitoring it doesn't. Fixing hallucinations doesn't. Managing customer complaints when the AI fails doesn't. Legal fees when the AI invents policy don't.
89% of the $47 billion invested in AI during the first half of 2025 delivered minimal returns. That's not a stat about customer service specifically—it's the whole market. But customer service was a primary use case for that spending.
The Gartner Verdict
Kathy Ross, a Gartner analyst, summarized the reality: "While AI offers significant potential to transform customer service, it is not a panacea. The human touch remains irreplaceable in many interactions."
And Brian Weber added a line that should be framed in every executive office: "Agentless contact center is not yet technically feasible, nor is it operationally desirable."
That's the buried lede. The technology doesn't work. And even if it did, customers don't want it.
What This Actually Means
The AI customer service story isn't ending—it's being rewritten. The companies backing away from elimination aren't abandoning AI. They're repositioning it as a tool to make agents more productive, not replace them.
This pattern mirrors what we've seen elsewhere in the market. As we covered in The AI Labor Reckoning Starts Now, the companies succeeding with AI are the ones treating it as augmentation, not replacement. The ones betting on displacement are discovering that the math doesn't work and the market punishes them.
The next 18 months will separate the companies that learned this lesson from the ones still doubling down. Look at hiring: the ones bringing customer service staff back are signaling that the AI-only strategy failed. The ones still cutting positions are either early enough in the process to reverse course or stubborn enough to run the experiment to its conclusion.
Either way, the great AI customer service replacement bet is over. The market has spoken. Customers want speed and accuracy, not the illusion of both delivered by a system that hallucinates policy and disappears when things get complicated.
The real story isn't "AI transforms customer service." It's "companies learned the hard way that customers still want humans, and that lesson cost them millions in wasted investment and lost customers."
Originally published on Derivinate News. Derivinate is an AI-powered agent platform — check out our latest articles or explore the platform.
Top comments (0)