Kogan Deploys AI Agent to Handle Customer Service Queries

What happened

Australian online retailer Kogan has introduced an AI agent to manage parts of its customer service operation, according to iTnews. The move sees the e-commerce group hand frontline customer queries to an artificial intelligence system rather than relying solely on human support staff.

Details of the rollout — including the scope of queries the agent handles, the vendor behind the technology, and how it is being phased in alongside Kogan's existing support channels — are limited in current reporting. What is clear is that Kogan is positioning AI as a direct operational tool for customer-facing service, rather than as a back-office or analytics function.

Why it matters

Kogan's move reflects a broader shift among online retailers: AI agents are moving from pilot projects and chatbots with scripted flows into systems trusted to handle real customer service workloads. For a high-volume e-commerce business, where order queries, returns and delivery questions drive much of the contact centre traffic, deploying an AI agent at this layer can materially change cost structures and response times if it works as intended.

It also signals how retailers are rethinking the division of labour between human agents and automated systems — not simply deflecting simple queries, but potentially having AI own entire service interactions end to end. That raises the stakes on getting the experience right: a well-designed agent can lift consistency and availability, but a poorly tuned one can just as easily erode trust at the exact moment a customer is already frustrated.

The Renascence take

The headline risk in stories like this isn't whether the AI can answer questions — it's whether the handover between AI and human, and the customer's sense of control throughout, has been designed with the same rigour as the model itself.

Most retailers treat an AI customer service launch as a technology decision; the ones that actually win trust treat it as a service-design decision first. The real test isn't resolution rate, it's what happens in the 10% of cases the AI gets wrong — how fast a customer can reach a human, how much context is preserved, and whether the experience still feels like Kogan rather than a generic bot. Operators considering a similar move should be measuring effort and trust alongside efficiency, not efficiency alone.