Event season is in full swing, and with it comes the annual flood of industry reports into inboxes. Some will be useful. Some will mainly confirm what the market already wants to believe.
The more valuable reports are the ones that reveal where contact center AI is colliding with operational reality. This week, three of them point in the same direction.
Five9 gives the contact center angle, showing that consumers are open to AI when a person remains within reach. Arkose Labs gives the AI angle, showing how agentic systems are starting to appear across protected digital journeys, forcing contact centers to think beyond their own bots. ContactBabel gives the UK operator’s view, where AI investment is rising quickly but accuracy, accountability, measurement, and agent readiness remain unresolved.
My read is simple: AI is not being rejected, but the market is becoming much less forgiving about vague promises.
Five9 Shows Why Human Access Still Defines AI Trust
Five9’s new 2026 Consumer CX Report, developed with Hanover Research, lands on a practical point that should shape every contact center AI roadmap. Consumers are embracing AI and asking brands to make AI recoverable.
The research surveyed 3,000 consumers in the US, UK, and Germany who had interacted with customer service in sectors including healthcare, financial services, travel, retail, higher education, sales, telemarketing, customer service, and outsourcing.
Five9 found that 57% of consumers say customer service has improved over the past decade, while 47% find AI useful. Yet only 20% use AI most of the time or always when seeking service. That gap should make contact center leaders careful. Positive sentiment toward AI does not automatically translate into default preference.
Five9 says trust in AI rises from 26% when no human option exists to 55% when a person is reachable. Meanwhile, 71% of consumers consider it very or extremely important to know when they are interacting with an AI agent. Amit Mathradas, Chief Executive Officer at Five9, positioned the finding as a case for humans and AI working as one service system:
“Consumers aren’t asking companies to choose between AI and people; they’re asking for both, done right.”
He added that Five9’s vision is for AI to absorb routine work, while people remain available for moments that need judgment and empathy.
This is important because contact center leaders often talk about escalation as a cost or exception, whereas customers experience escalation as reassurance. A visible route to a person signals to the customer that the company has not trapped them inside automation.
Five9 also found that 92% of consumers give AI less than ten minutes before escalating to a human. Forty-six percent say the top reason they abandon AI is that it repeatedly misunderstood them, and roughly a third say they have to repeat information already given to a bot once they reach a person.
Those findings reinforce a point CX Today has already made in its coverage of AI escalation strategy and customer trust: escalation only protects trust when it transfers context. If the human agent starts cold, the customer views the escalation as proof that the AI wasted their time.
AI performance must now be judged by misunderstanding rates, time-to-human, context transfer, repeat contact, and whether customers knew they were dealing with AI in the first place.
Arkose Labs Turns Agentic AI Into a Service Risk
Arkose Labs’ State of Agentic AI Trust Report for Q3 2026 significance for contact centers comes from what it says about AI moving into the customer journey before service teams have designed for it.
The report analyzed production traffic from July 1 to September 30, 2026, across protected actions including login, registration, checkout, password reset, and one-time passcode flows. By the end of Q3, 65.2% of Arkose Labs customers had seen AI agents on protected flows.
The standout figure is the disclosure gap. Arkose Labs says 97.7% of legitimate AI agent sessions carried no verified disclosure. Only 2.3% presented a verified Web Bot Auth signature.
For a contact center leader, that creates a new question. What happens when a customer delegates part of their journey to an AI agent before they ever reach the contact center?
An AI agent may check an order, initiate a password reset, query account information, change a delivery slot, or attempt a refund. If the task succeeds, the contact center may never see it. If it fails, the customer will arrive through voice, chat, or email expecting the organization to understand what happened.
That makes agentic AI a service continuity issue as well as a digital security issue.
Arkose Labs also shows why simple allow-or-block policies will not work. The company reported that 70.5% of registration sessions at protected endpoints were fake-account attempts, 33.5% of sign-in sessions were illegitimate, and 17.7% of checkout sessions involved card testing or payment fraud. Mitchell Davies, Security Research Manager at Arkose Labs, framed the distinction clearly:
“Every agent we track shows signs of automation somewhere. This was never the question. Rather, the real question is, what did the agent come to do? What is its intent?”
That is the contact center translation. Automation signals will not be enough. Service organizations will need to understand authorization, intent, customer consent, risk, and escalation across digital and assisted channels.
The customer’s AI agent may be legitimate. A malicious agent may look similar. A legacy bot-control policy may block both or miss both.
That means contact center AI strategy has to extend beyond owned virtual agents. Service leaders will need tighter coordination with digital, identity, fraud, and security teams. They will also need clear rules for what happens when an agentic interaction triggers an exception, creates ambiguity, or hands off to a human.
