Kogan deploys Salesforce AI agent for customer support
The retailer says its agent can handle 10 types of support requests, including order tracking, voucher redemption and identity verification.
By Quang Dinh
Australian online retailer Kogan has put an AI customer service agent into production following about six weeks of development, with 10 support use cases now available, chief technology officer Goran Stefkovski said.
Speaking at Salesforce’s Dreamforce conference, Stefkovski said Kogan wanted to make support requests less demanding for customers and deliver quicker responses. Customers had previously relied on email and web forms, while an earlier deterministic bot resolved about six per cent of incoming inquiries.
Kogan rebuilt the bot using Salesforce’s Agentforce platform, working with the vendor’s forward deployment team. Initial prototypes were developed during an internal hack day. The agent uses four services: NextGen Authoring, Agent Script, Observability and Testing Center.
The agent’s functions include retrieving order status from SAP for customers, personalised product recommendations across 16 million SKUs, membership support, mobile voucher redemption, investigations into unfamiliar charges, troubleshooting and identity verification.
Stefkovski also described an emerging use case involving customer photographs of damaged products submitted for returns. Kogan commonly requests photos when goods are damaged in transit. He said generative AI could interpret those images without Kogan having to build a model-training dataset.
He said the technology could also identify images submitted repeatedly across claims, helping detect attempts to misrepresent product damage. A July blog post said Kogan was gradually extending the agent’s coverage rather than using it for every product type.
