CX Today: Gartner warns GenAI customer-service costs may outgrow savings

CX Today: Gartner warns GenAI customer-service costs may outgrow savings

CX Today reports Gartner’s prediction that gen AI customer-service cost per resolution will exceed $3 by 2030, higher than many B2C offshore human agents. Patrick Quinlan, a Gartner analyst, says organizations can overlook total cost of ownership when they focus on low per-unit consumption costs. He points to specialized AI talent, higher-than-expected usage, and infrastructure expenses. The article describes several possible cost pressures: some large language model vendors subsidize prices; newer models may use more tokens per interaction; and data centers require compute, electricity, water, and specialized chips. The article says these pressures could affect vendor prices, but does not establish that they will. Gartner also forecasts that half of companies that cut customer-service staff because of AI will rehire people for similar functions by 2027. A Gartner survey from October 2025 found that only 20% of customer-service leaders had reduced agent staffing because of AI. Quinlan distinguishes automating resolution from using GenAI to support agents: he recommends considering triage, summarization, note-taking, and intent classification, and says GenAI resolution is unreliable. The article’s argument is that expected savings may disappoint, while using AI to engage customers and provide an immediate connection may be a better fit for some organizations. What evidence in

Publication: 347bd340-c704-4eb9-8386-d6fc0625d9b3

attribution — pass: The predictions are correctly attributed to Gartner. Quotes and views are correctly attributed to Patrick Quinlan, Senior Director Analyst. The reporting is attributed to CX Today. The October 2025 survey and the rehire forecast are correctly identified as Gartner's.

Patrick Quinlan , Senior Director Analyst at Gartner , explained that the issue goes well beyond the per-unit consumption costs

summary — uncertain: The summary ends with a stray truncated fragment, "What evidence in", which reads as broken text. Remove that fragment. The draft also flattens Quinlan's conviction. He says organizations expecting big savings "will be disappointed" and that subsidies are "going to have to change", but the draft renders this as "may disappoint" and "could affect". Restoring his stated certainty, while keeping it attributed, would repair this. Central terms are otherwise understandable, with tooltips for tokens, B2C and intent classification.

I think organizations who expect big cost savings from GenAI will be disappointed

date — pass: The originalPublished date of 2026-02-03 is consistent with the publisher image path (2026/02) and with the article's reference to an October 2025 survey as past research. No conflicting date appears in the text.

support — uncertain: Most claims match their excerpts. However, the draft says "some large language model vendors" subsidize prices by up to 90%. The source instead says LLM vendors generally subsidize, with the figure qualified as coming from "some estimates". The draft moves the qualifier from the estimate to the vendors. The smallest repair is: "LLM vendors subsidize services by up to 90%, according to some estimates." The question and limitations are otherwise supported.

Large language model vendors are currently subsidizing their services by up to 90%, according to some estimates

media — pass: The selected media is the verified direct link to the original CX Today article, which is the most direct access to the source.