Conversational AI in the times of Covid

Guest Column: Niraj Ruparel, Head of Mobile & Emerging Tech at GroupM India & Head Of Voice, WPP India, writes on Conversational Experiences to help brands reduce dependencies on manual processes

e4m by Niraj Ruparel
Updated: Sep 24, 2020 12:27 PM
AI

Consumer behaviour in the last nine months has changed drastically across the globe and people are spending more time on messaging platforms than ever before. Not that conversational commerce has become relevant only in COVID times, but indeed, COVID has accelerated the digital transformation across industries. It is propelling brands to rethink their e-comm strategies and automate processes to reduce dependencies on manual processes. 

Secondly, when COVID hit, most of the businesses depended on e-commerce sites for their online business. And these aggregators take up roughly 10 to 30% of the margins from brands. And while it may work for volume sales, it does impact the brands' revenues greatly. Hence Direct-to-Consumer works much better. COVID just cemented the fact that offline retail cannot be depended upon forever. Things have to move to digital and c-commerce is the only way to replicate that experience on digital channels. 

Traditionally, in the chat-based solutions, chatbots were more or less used for half-baked experiences. In India, we are enabling brands to ditch traditional chatbots and enable full-scale conversational experiences. Key ideas would be to go omnichannel (WA, FB Messenger, iMessage etc) and using a different approach for different use cases (Automated Chat, 1:1 Human Chat between Brand Advisors and customers, or a Hybrid approach) to power the entire customer lifecycle. 

Conversational Experiences are not just for Customer Support Automation or redirecting to websites anymore. They are bigger and better when it's full-scale with the 360 media surround. We are helping brands strategize conversational commerce and the approach. While we use partner tools to enable these experiences, a 360-degree strategy has to be coupled with the tool. There is content, marketing strategy, etc. involved. 

Unlike campaigns, conversational commerce is driven as part of the business strategy by the brand teams. It's an integral part of the digital transformation of a brand. And to help brands achieve optimal results and ROI from these channels, we play a vital role in the customer experience. 

Clients across industries are gunning to double or triple their revenue using these channels to complement their e-commerce website. Traditional e-comm websites have conversion rates of less than 1%. Conversational Commerce gives you much higher conversion rates. 

Conversational AI engine is designed keeping in mind the behaviour patterns of consumers while making a purchase. The Sales Engine effectively replicates the behaviour of a Sales Assistant and is able to guide customers through their buying journey. It is built to deliver the most relevant recommendations to the user with the least friction and is optimized for a superior Customer Experience over Voice or Text. 

Let's understand this better by deep diving into these five key components, each of which helps in creating a deep understanding of products and consumers, of the kind that exists in the mind of any sales expert. 

  1. Domain Knowledge 

Much like a human sales assistant, the Sales NLU models are pre-trained to acquire all the knowledge of the domain they are built for. The NLU models learn all common concepts and relationships across products, reviews and customer questions. 

Within any industry, there are various types of products. Each of those products has different features, attributes, qualitative aspects, quantitative aspects, price ranges, dependencies, etc. among a host of other factors. Furthermore, the questions by customers in any domain fall into a variety of types. The engine learns to understand all these concepts and their synonymous representations using publicly available data such as open webpages, forums, Wikipedia articles etc. It also adds to this domain knowledge once it integrates with the business inventory data by learning new concepts present in the inventory. 

  1. Information Extraction 

The domain knowledge becomes the basis for the engine to extract structured information from natural language product descriptions. The engine syncs with the inventory data of the business and is able to extract structured product features and numerical attributes from any unstructured form of text using advanced semantic extraction techniques. This is similar to how a human sales assistant would read a product’s description, understand the information in a structured manner by making the required deductive conclusions and store that information for later use. 

This domain knowledge is also central to understanding any incoming messages from the customer. It allows the engine to quickly extract the domain concepts present in the message and match them reliably with known customer query patterns.  

  1. Sentiment Analysis of Reviews 

Since reviews today form a critical component of the customer purchase journey in many domains, engine too relies on creating a deep understanding from reviews about overall product quality, as well as for specific use cases. In reviews, different types of consumers talk about different attributes and what aspects of the product have specifically resonated with them or left them dissatisfied. The engine maps this in order to be able to have this context while recommending products.  

  1. Recommendation Engine

Since the effectiveness of any sales expert is primarily based on the quality of their recommendations and personalized guidance, this is a critical piece of our sales assistant engine. The recommendation engine takes into account not only information learned from user reviews and product descriptions, but also uses deductive logic and reasoning to come up with personalized recommendations based on the customer’s use case and needs. Further, it can be easily configured by the business to prioritize new releases, specific brands or products. 

  1. Context Manager

 Once a consumer has started interacting with the IVA, they often keep refining their search further based on preferences. The Context manager helps maintain the various parameters that a consumer has already mentioned over the course of the conversation. It keeps in mind factors such as objective of discussion, stage of discussion, known preferences and criteria, products that the customer has shown interest.

Use cases across categories :

  1. All Industries (1-way Conversations): Sending alerts & notifications on WhatsApp to customers (Much like receiving a ticket from Makemytrip/Goibibo etc) 
  1. For consumer brands : 3 different approaches (Automated Conversations, Assisted Brand Advisor led Conversations, Hybrid model)

Use Cases: Marketing Chatbots/ Conversational Commerce/ Customer Support/ Product Recommendation/ Connecting Offline Stores/ Sampling/ Product Consultation (Beauty advisors, Style advisors, Nutrition Experts etc.) 

  1. For Banking, Automobile, Insurance: Automated Customer Experiences, Lead Generation, Assistance using Relationship Managers, Wealth Managers etc, Connecting to branches, Showrooms, booking appointments etc

Disclaimer: The views expressed here are solely those of the author and do not in any way represent the views of exchange4media.com.

For more updates, be socially connected with us on
WhatsApp, Instagram, LinkedIn, Twitter, Facebook & Youtube