LivePerson’s Maven bot platform puts conversational AI in customer agents’ hands

A screenshot from LivePerson’s Maven bot platform.

Conversational commerce provider LivePerson is out with a new platform that counts on customer service agents working with AI to create and orchestrate chatbots.

“We now know,” founder and CEO Robert LoCascio said, “that people in contact centers, who converse with consumers everyday, are the best conversational designers for AI.”

How can customer agents set it up? Called Maven, the new platform employs AI to track ongoing text conversations between users and agents across SMS, Facebook Messenger, WhatsApp and other messaging channels.

A new Conversation Builder lets agents and other users map chatbot conversations for frequent interactions. Here’s a screen showing a conversation built from a previous conversation:

Conversations around frequent topics can mine a brand’s previous conversation transcripts, plus there are pre-built templated conversations — such as for billing — that agents can select and edit. Chief Scientist Joe Bradley noted that this kind of chatbot building does not rely on menus, which can direct the conversation flow in ways not related to the brand.

How do the chatbots get deployed? During a live text conversation between an agent and user, Maven determines intent from the conversation and can recommend next best actions or specific chatbots to take over the conversation.

Bradley added that LivePerson is rolling out various off-the-shelf, special purpose chatbots, such as ones designed to convey order status, complete a purchase or change an address.

For instance, if the agent is explaining order status, Maven can suggest the order-oriented chatbot already built by the brand. (See screenshot at top of this page.) When the conversation veers to another topic, the chatbot will toss the conversation back to the live agent.

When Maven and agent feedback determine that this use of a given chatbot has reached a brand-set level of accuracy in responding to the user, the hand-off to the chatbot can be automatic the next time a live text chat turns to that topic.

Maven’s recommendation of a chatbot in a conversation can include chatbots from other integrated platforms, such Google’s Dialogflow, Watson and Amazon. The conversation from those chatbots will then be available within Maven. Bradley noted that, while LivePerson previously had the ability to integrate with other chatbot platforms, it couldn’t previously recommend their chatbots.

Why you should care. Customer service is now another form of marketing. As a result, customer service agents need to respond quickly, at scale and knowledgeably across a wide range of topics, a task that can be assisted by intelligent platforms and automated chat engines.

LivePerson’s effort to enlist the chatbot-building skills of agents is only the latest effort to make contact centers more intelligent, a trend impacted by recent releases from Google, Genesys and Helpshift.