“Whom you often long to speak to during a customer service call.” This clue from a recent New York Times mini crossword puzzle gave me nary a pause, thanks to the CX analyst in me. I knew “human” had to be the answer … despite the industry’s best efforts to insert AI agents into customer interactions.
In studying business and consumer perceptions about AI’s use for CX, Metrigy has found several areas in which the two don’t match up. Here’s one example: Most businesses (75.9%) using AI feel their customer service has improved in the last year, according to our global “AI for Business Success 2025-26” study with 1,104 companies. Consumers, however, are divided in their opinions on the quality of their customer service interactions. In an April 2025 study of 503 consumers in North America, only 35.6% reported that customer service had gotten better. The others were fairly evenly split between not seeing any change and seeing a degradation in customer service.
Additionally, while the majority (almost 70%) of consumers anticipate AI will have at least some (44.1%) if not extensive (25.6%) impact on their future interactions with companies, not everyone feels that impact will be good. A significant percentage, 41.1%, said they anticipate AI’s impact to be negative.
One potential ding might come as consumers encounter the AI text and voice agents which many companies have deployed to facilitate customer self-service. While consumers who use AI agents give a nod to improvements they’ve seen with them, the overall customer experiences lack consistency. 26.4% of consumers note that they’ve experienced “significant variability” using AI agents, and 55.3% have encountered “some variability” in AI-based interactions.
And AI agents aren’t yet a go-to service option for many consumers. Thirty eight percent of consumers said they avoid using AI text agents compared to only 19.5% of consumers preferring AI text agents; 26.5% of customers avoid voice AI agents while 30.5% of customers prefer voice agents. The primary frustrations leading to customers avoiding both text and voice AI agents are the agents’ inability to understand needs, and customers getting stuck in an AI-agent interaction loop without a clear path to a human. Conversely, time savings and convenience are the top reasons for using AI agents. The takeaway for companies is that AI implemented to improve efficiency will create more frustration than convenience if it doesn’t adequately understand consumer needs or provide an easy escape route to a live agent.
While generative AI is in widespread use, our research finds that trust remains limited among consumers. Only 21.0% fully trust generative AI, compared to 60.8% who trust AI only in select circumstances or the 18.2% who don’t trust it at all. And as much as companies talk about making AI more and more autonomous, 28.4% of consumers indicated they will never trust AI to complete tasks on their behalf. The primary concerns consumers have about generative AI include data privacy (57.6%), ethical use of data (47.3%), malicious use of the technology (46.4%), and the accuracy of information (40.9%).
Building trust will take transparency and delivering practical value. According to consumers studied, companies should be open about when AI is being used (46.1%), ensure human oversight of AI’s content creation and decisions (41.2%), and provide assurance that personal information is not being used in ways that compromise privacy (39.5%).
Beyond trust, companies can enhance the value of AI in CX by aligning with consumer preferences. For example, 39.2% of consumers appreciate at least some personalization during their interactions. This includes the use of preferred channels and conversational tone.
CX leaders are rightly enthusiastic about AI’s potential, but many consumers are more cautious. Building trust and demonstrating tangible benefits will serve CX leaders well in winning over more consumers to AI’s court.
