Integrating AI with marketing is an emergent disruptive force. AI powered conversational bots, or chatbots, will have the ability to know customers better then humans and automate most customer services interactions. Chatbots can offer a surprisingly compelling personalized experience by predicting customer intent and helping users engage with products and brand.
Chatbots, also called Conversational Agents or Dialog Systems, are algorithms designed to have human level conversation capabilities. Several companies are using this technology, either as personal assistants or as conversational algorithms for language understanding. The goal of bots is to achieve a level of natural conversations indistinguishable from human – thus capable to pass the Turin Test.
There are two types of bots, Retrieval-based, that use a repository of predefined responses and some heuristics to pick an appropriate response based on the input and context and Generative models that generate automatically responses from past experience and context. Most of the latter rely on Deep Learning technology.
However, good algorithms and lots of domain-relevant data is needed. For the first time we have both: deep learning and big data. Millions of chat sessions of data are commonly available and data sources are proliferating as the ability to connect all the steps of a customer’s journey from search to purchase and beyond is becoming more sophisticated.
The key aspect of this automation is to link search and advertising data with revenue transactions and service transactions. As clients are moving to end-to-end interaction, everything from search and personalization to customer care could be used to analyze their behavior. If a valuable potential customer is spotted searching the web, the system will bid high to win it, and then is able to fast-track it by determining the intent of the search — for example, specific interactions that indicate they are searching for a particular phone model. The search then brings the consumer to a personalized landing page, which offers them an opportunity to chat with a human customer service agent or a chatbot. Segmenting out frequent buy-and-return customers allows companies to identify customers early on the journey, and make sure a live chat option is never offered to that kind of person, but they are instead redirected to a self-service area.
For instance, Chatbots are commonly used by the travel meta-search engine Skyscanner. As customers become more familiar with the medium, they go from being conversational to being very specific in what they want, and return to use bots more frequently. Those using the chatbots treat them in a very ‘human’ way – they’ll ask for the bot’s name, send an emoji or sticker of appreciation. Skyscanner wants to deliver complex answers from simple questions such as ‘who flies cheaply from here to there’, but has matured to the point it can now offer simple answers to complicated ones. The chatbots are fully automated with no human supervision, although there is an option for users to ‘speak to a human within Messenger bot which transfers them through to one of our community managers. The bots use natural language processing, and Skyscanner trained the voice recognition systems with its own employees’ 50 different accents.
The vision should be on how you want to interact with customers and then looking at whether you could do this more efficiently with technology. A completely automated customer service using chatbots on is still impossible, but it could drastically reduce the work force in handling most routine operations.
Going a step forward, the next generation (of humans, not bots) should learn an etiquette of talking to computers. Weired conversations may occur as people ask the bot’s name, if the bot will have sex with them or help them find a girlfriend, sending nude photos or inappropriate animated gifs…