First came the technology: for the last three years, artificial intelligence has been touted as a breakthrough to help enhance customer experience (CX) on multiple levels at once.
Next came the panic: worried they would fall behind competitors, senior leadership teams have been mandating AI adoption, and urging their teams to deploy it as quickly as possible.
Now comes the reality check: AI may not deliver the CX changes executives had hoped for, which makes the concept of customer agency an important concept to learn and apply.
What is customer agency in AI?
For Elaine Buxton, president and CEO of Cary, NYC CX market research consultancy Confero, this isn’t a difficult term to define.
“Customer agency means the customer still feels in control of the journey. AI should serve the customer, not steer them into pre-determined paths,” Buxton told 360 Magazine, adding that reducing friction should be the primary goal vs. simply deflating calls to a contact center. “CX leaders need to frame AI decisions in terms of friction, not features. The first question should always be: does this tool make it easier or harder for customers to get what they came for?”
Ravi Teja Surampudi, senior manager in business technology, go-to-market at Workday, customer agency also means the customer understands what is going on during an experience as well as affect the outcome, such as leaving or overriding the AI path without being penalized for it.
The “control” aspect should be broken down as assessing whether AI allows for a predictable experience across channels vs. something that feels random, Surampudi said. Then, organizations should determine whether they can make the reasons for the AI suggestion or action understandable to everyone before allowing them to change what happens.
“The biggest mindset shift I’ve seen is this: if we treat agentic AI as ‘automation,’ we’ll build just chatbots, but if we treat it as ‘agency,’ we’ll build experiences that customers actually trust and choose to use,” he said.
What are some examples of customer agency in AI?
The best examples of customer agency demonstrate the use of multiple levels of control, rather than simply presenting a black box to customers, said Brady Lewis, senior director of AI Innovation at marketing consultancy Marketri. For example, brands should offer easy-to-use preference settings and clear opt-out options, which helps to foster a greater sense of trust.
“Over-personalized customer experiences often fail to meet customer expectations if the focus is on maximizing engagement instead of usefulness,” he said. “When the automated system begins to direct a customer’s behavioral choices and no longer responds to what the customer intends, it becomes apparent to the customer.”
From a governance perspective, Lewis said brands must distinguish between data readiness and data activation, with the caveat that just because a signal can be activated does not necessarily mean that it should be. This helps pinpoint where the “human in the loop” needs to happen.
“The objective is not to remove humans from the customer’s interactions,” he said. “The objective is to ensure that humans occupy the position where judgment is still required.”
Kelsey Silver, founder and buyer psychology specialist at ForesightHQ in West Hampsted, NY, said brands ignore customer agency amid the eagerness to deploy AI at their peril.
“I’ve seen a drastic culture shift within the agencies I work with. They are concerned about the skepticism and lack of trust customers have when interacting with AI,” she said. “Customers are savvy. If they even think they are interacting with AI, they often change their behavior.”
For example, when brands are interacting with clients on social media, it’s actually riskier to make automated responses sound human, Silver argued. “Customers clock this immediately. Then, they either mess with the automation on purpose, or lose trust and end the conversation. Either way, trust goes down the tubes.”
How should customer agency in AI be measured?
CX leaders are the obvious candidates to evangelize customer agency, but it may be a process of “managing up.” Nik Kale, principal engineer of CX engineering, cloud security and AI platforms at Cisco, suggested framing it as a risk and trust issue, not a philosophy issue.
“If customers feel like the system is doing things to them, they disengage, complain, or work around it, and then the AI program becomes a brand and compliance headache,” he said. “The simplest leadership message is: if we cannot explain what the system is doing and give customers meaningful choices, we are not ready to scale autonomy. That turns agency into a readiness gate, not a nice-to-have.”
The success of customer agency efforts could be evaluated using traditional CX metrics like CSAT and NPs, but Kale suggested some alternatives. These include is “customer override rate,” meaning how often customers correct or change what the AI suggested. Another is “explanation success,” meaning whether customers report that the system’s rationale made sense.
“I also watch ‘regret rate,’ like cancellations or reversals shortly after an AI-assisted action, and ‘escalation quality,’ meaning when customers do escalate, do they feel heard and get a clean resolution,” he said.
For Buxton, however, asking customers openly about their AI interactions is the strongest signal.
“It tells them their voice matters,” she said. “In business school, we were taught that it is far less expensive to keep a satisfied customer than to attract a new one. That principle still holds true, especially as technology becomes more visible in everyday service interactions.”