Amazon’s automated customer chat is becoming more sophisticated.
The e-tail giant has begun testing two neural-network-based automated customer service systems. One can handle common customer service requests automatically and one helps customer service agents respond to customers more easily.
As opposed to standard text-based online customer service systems featuring automated agents that can handle simple requests by following rules, Amazon’s new neural network-based agents can handle a broader range of interactions. This enables live customer service representatives to focus on tasks that depend more on human judgment.
In randomized trials, Amazon has been comparing the new neural agents to its existing rule-based automated customer service systems, using a metric called automation rate. Automation rate combines two factors: whether the automated agent successfully completes a transaction without referring it to a customer service representative, and whether the customer contacts customer service a second time within 24 hours. According to automation rate, Amazon says the new agents significantly outperform the old ones.
At the same time, Amazon is also testing a variation on the system that suggests possible responses to customer service representatives, with the goal of saving them time. Amazon trained separate versions of each model for two types of interactions, return refund status requests and order cancellations. As an input, the order cancellation model receives not only the preceding customer dialogue but also some information about the customer’s account profile.
In addition to the context and the profile information, the response ranker receives a candidate response as input. It also uses an attention mechanism to determine which words in which previous dialogue are particularly useful for ranking the response.
“It is difficult to determine what types of conversational models other customer service systems are running, but we are unaware of any announced deployments of end-to-end, neural-network-based dialogue models like ours,” said Jared Kramer, machine learning manager at Amazon, in a corporate blog post. “And we are working continually to expand the breadth and complexity of the conversations our models can engage in, to make customer service queries as efficient as possible for our customers.”