Personal AI agents now contact customer service teams

NiCE has published survey findings showing that personal AI agents are already contacting customer service teams. The responses came from 118 practitioners across more than 70 customer organisations.

Six in ten respondents said they are already dealing with AI agents that contact companies autonomously on behalf of consumers or partners. The survey covered organisations in Australia, North America, the UK, Europe and India, with respondents drawn from frontline, operations, IT, workforce management and leadership roles.

The findings suggest a shift in how customer service operations are managed. Rather than using AI only to assist human staff, many organisations now describe AI systems interacting directly with other AI systems to complete tasks such as identifying customer intent, retrieving account information, routing cases and triggering follow-up actions.

More than half of respondents said they are using a multi-agent ecosystem. These setups often involve several AI agents working together within one organisation, while a separate group of personal AI agents contacts the business from the customer side.

In Australia and New Zealand, the shift is already affecting digital and call centre operations, according to Ben Hancock, APAC Lead, Forward Deployed Engineering, NiCE.

"Across Australia and New Zealand, we're seeing how AI is creating new opportunities for businesses to transform customer experiences and improve operational efficiency. In Australia, a major insurer is seeing a significant increase in personal AI agent traffic on its quote and purchase pages, creating new challenges for its digital team in distinguishing genuine customer prospects from activity initiated by personal AI agents. Rather than simply blocking or restricting these interactions, NiCE is working with the customer to explore how our platform can support and accommodate personal AI agents, helping businesses adapt to this emerging trend.

"In New Zealand, a leading telecommunications and utilities provider has developed a Voice AI Concierge in collaboration with NiCE Labs' Forward Deployed Engineering team. The AI agent allows customers to explain their needs in plain language, verify their identity, troubleshoot internet issues and, when necessary, transfer the conversation to a human agent with the relevant context for a more seamless customer experience. With an estimated 90-120 seconds saved in average handling time per call, the project is progressing faster than more traditional AI initiatives because of its expected return on investment.

"We're also seeing momentum in the healthcare sector, with another New Zealand organisation preparing to go live with Cognigy.AI by Christmas. The deployment will make it one of the first organisations in the country to implement live Voice AI at scale and the first NiCE customer to use Cognigy.AI for X-ray concierge services, routing and appointment booking. These examples show how organisations across the region are moving beyond experimentation to apply AI to real-world customer needs and measurable business outcomes," Hancock said.

The survey suggests these interactions are no longer limited to internal automation. Respondents reported personal AI agents scheduling appointments, while service agents worked with logistics partners' agents to manage requests without direct human involvement at every stage.

Oversight focus

Governance emerged as a central issue in the responses. More than half of participants cited oversight, auditability, human-in-the-loop controls or unified interaction logs as the factors that enabled them to expand automation while retaining visibility over what AI systems were doing.

That emphasis reflects the operational risks involved when software agents act independently across customer-facing systems. Organisations said they were putting orchestration, identity checks and intent-verification controls in place to handle AI-to-AI interactions more securely.

Several respondents described measurable effects on efficiency. A propane distributor used a customer-facing agent to resolve delivery and billing queries by linking directly with telemetry and routing systems, while a healthcare technology provider said AI-to-AI handoffs had cut most after-call work and improved documentation quality and case closure speed.

A respondent from the transport sector described a direct effect on routine work volumes.

"We are increasingly seeing customer-side personal AI agents interacting directly with our front-line virtual agents to negotiate appointments, clarify account details or initiate service requests without a human present on either side. This has dramatically reduced call hold times and human agent handling time for routine queries," said an Engineer at a Global Transportation Company.

The same respondent also outlined the operational controls being used.

"NiCE helps us manage this transition by maintaining strict intent-verification workflows, ensuring API security at the routing layer, and providing unified interaction logs so our human supervisors can monitor agent-to-agent dialogue in real time," the Engineer said.

Other sectors reported similar changes. A Head of Customer Experience at a UK energy utility said automation had risen from no fully automated interactions to almost 35% of interactions being contained within an AI agent, while a Non-Clinical Performance Manager at a UK urgent and integrated care provider said AI agents were moving from support roles to completing end-to-end tasks with other agents.

A Customer Service Planner at a Global Building Technologies Company said visibility across those exchanges was a key factor in adoption.

"NiCE provides visibility and control over these interactions. It centralises customer engagement data, tracks automated actions and ensures AI-driven processes remain auditable and aligned with business rules," the Customer Service Planner said.

"This gives me confidence that automation is improving efficiency without losing oversight or customer experience," the planner said.