Australian retailer Kogan has put a Salesforce-built AI agent into production for ten customer-service use cases, including order tracking, product recommendations and returns. The practical detail is that the system went live after about six weeks of development, while the company is still expanding its reach rather than applying it to every product type.
Watch Desk analysis
What happened
Kogan developed the agent with Salesforce’s Agentforce technology and its forward-deployed engineering team, iTnews reports. The retailer says the agent can handle ten use cases, from answering “where is my order?” questions using order information in SAP to membership, voucher redemption, troubleshooting, identity checks and warranty or return issues.
The article says order-status requests account for roughly 60% of Kogan’s inbound customer requests, citing a company blog post from July. Kogan is also exploring image analysis for damage claims. Chief technology officer Goran Stefkovski told iTnews that the system can identify when the same damage image has been submitted in multiple cases. Those are company-reported capabilities, not independently measured results.
Why it matters
This is a concrete example of an AI customer-service project moving from prototype to live use, with a short reported build-to-production timeline and a defined set of tasks. It also shows why the first useful measure is not a grand claim about replacing support staff, but whether the agent can resolve common requests and handle awkward cases without making customers start over.
There is an important comparison in the account: Kogan’s earlier deterministic bot resolved about 6% of incoming enquiries. That figure describes the earlier bot, not the current AI agent. The report gives no resolution rate for the new system, so the ten use cases and six-week launch are evidence of deployment, not proof of how well it performs.
Our read
The interesting part is the operational shape: connect order data, handle repetitive requests, and try image analysis on a specific returns problem. That is more useful than another claim that AI will transform customer service by Tuesday. Kogan’s reported experience is a promising deployment example, but the missing number is how often the new agent actually resolves a request correctly, without handing the mess back to a person.
What to watch
- Whether Kogan publishes resolution rates for the new agent, and how they compare with the earlier bot.
- How many of the ten use cases are fully automated rather than assisted by staff.
- Whether image checks reduce repeat or fraudulent damage claims without misclassifying genuine ones.
- How quickly Kogan expands the agent to other products and support tasks.
Discussion spark: Should retailers roll out customer-service AI once it can handle a defined set of tasks, or wait until they can show that it resolves those requests reliably?
Sources and evidence
- Source update (13 October 2026, 00:00 UTC)
- Source update (13 October 2026, 00:00 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.