A Twilio survey of delegates at Customer Contact Week Australia and New Zealand found that 85% of marketing and customer experience leaders say tech fragmentation makes managing AI agent productivity harder, iTWire reported. The findings, corroborated by TelcoNews Australia, paint a picture of enterprise agent deployment that is widespread but not yet delivering on its promises.
The numbers tell a clear story about the gap between deployment and value:
94% of respondents said their organization would achieve greater financial gains with better AI agent productivity, with close to half estimating the potential benefit at 50% or more. Yet 25% of respondents report agents consume half or more of their teams’ time in management overhead. Another 41% say agents take up a quarter of team time. Only about one-third of respondents have multiple agents in production driving core business goals, according to Twilio’s Howard Fyffe, Director of Australia and New Zealand.
Deployment Without Strategy
The survey reveals a maturity problem. Nearly a quarter (24%) of respondents use AI agents on an ad hoc basis with no central strategy. Another 31% are still in pilot experiments and not yet scaling. 13% remain skeptical and have not deployed agents at all.
“We are in this transient period right now, where most organisations are deploying AI Agents, but not yet using them strategically, or seeing the full ROI,” Fyffe told iTWire. “Many teams are also still spending as much or more time managing these AI Agents than they are saving.”
Most respondents said greater agent productivity would lead to faster innovation (66%) and improved customer satisfaction (52%). But 65% said security and data privacy would not improve with better agent performance, signaling persistent concern about agents’ impact on data protection.
The Integration Tax
The survey’s central finding is architectural, not technological. The primary barrier to agent ROI is not model capability or agent sophistication. It is the fragmentation of existing tool stacks that agents must integrate with. Contact centers not built for AI-era workflows, disparate marketing platforms, and siloed data systems all create an “integration tax” that consumes the productivity gains agents are supposed to deliver.
This mirrors the legal profession’s parallel discovery that AI validation overhead can negate net productivity gains. In both cases, agents perform their core task effectively but the surrounding organizational infrastructure extracts most of the benefit.