The promise of "free" customer service automation is running into a $3 reality check.
In my work as a CX Expert, the narrative for the last two years has been dominated by a single promise: total automation. But as we move into 2026, the "unseen" details of the balance sheet are starting to tell a different story.
According to recent Gartner research, the cost per resolution for Generative AI in customer service is projected to exceed $3 by 2030. To put that in perspective, that is often higher than the cost of a high-quality offshore human agent. The subsidies are ending, and the bills are coming due.
The End of the Subsidy
For the past 18 months, LLM vendors have been subsidizing consumption by up to 90% to gain market share. As these companies pivot toward profitability, that "cheap" API call is becoming a premium line item. When you factor in the specialized AI talent required to maintain these systems—talent that commands significantly higher salaries than traditional staff—the math changes.
Infrastructure is the New Bottleneck
Beyond the software, there is a physical reality. Data centers are hitting power grid limits. We are seeing states pass laws to prevent infrastructure costs from being passed to consumers, meaning vendors have to absorb the cost of electricity and cooling. This "operational friction" is something many companies ignored in their initial AI pilots.
The Rehire Prediction
Gartner predicts that by 2027, half of the companies that cut staff for AI will end up rehiring. Why? Because AI is non-deterministic. It’s great at triage, but often unreliable for resolution.
The real winners won't be the companies that replace their humans, but the ones who use AI to give their humans the "backstory" of a customer's problem before they even pick up the phone.