Conversational commerce: Your bot needs to understand human personality better

Personality conversational commerce
istock
Share this content

When contact centre interactions feature similar interaction styles between caller and agent, there is higher satisfaction and better business outcomes. So what does this mean for bots?

The broadcast era is decades dead; today, conversations count – interactive, iterative commerce, driven by personality, opinion, and emotion. Today, the customer journey spans a spectrum of consumer-brand touchpoints, both in person and automated. Can your bot keep up?

Mattersight makes call centre products – telephony, routing, and analytical systems. The company has been applying analysis-derived insights to bots. These insights centre on personality with the aim of generating higher satisfaction scores, decreased churn, improved sales conversions, and lower costs.

Andy Traba, vice president of behavioural and data science at Mattersight, makes this happen, and it's a great story for the 2018 Sentiment Analysis Symposium. I interviewed Andy in the run-up to the symposium, which takes place March 26-27 in New York, to get some insight into his topic:

Andy Traba, Mattersight

"Your Bot Needs to Understand Human Personality Better"

Seth Grimes. Andy, according to your bio, you run the team that is "responsible for generating algorithms that turn freeform conversations into data." Let's unpack this. What conversations? What data do you extract? And what classes of algorithms?

Andy Traba. Mattersight’s typical customer is a Fortune 500 business who also operates large call centres for service, sales or retention operations. So generally we are analysing the conversations between a consumer and an agent, more or less the types of conversations you are regularly having — calling your healthcare company to inquire about benefits, your bank to dispute a charge, or media company to purchase internet or TV services.

Over the course of a decade, we have built millions of algorithms to analyse every second of every call for these companies. Examples of these algorithms are classifying the personality of the caller, whether they are having a positive or negative experience, and the behaviours of the agent that are supporting business objectives. The data extracted is used to drive real business value by making more intelligent routing decisions, improve business processes and driving smarter talent management programs.

Seth. You're using AI to model intention and personality. What aspects? (How) do you match intention and personality to outcomes?

Andy. We’re analysing for the communication preferences of the consumer and cataloging their interactions with the enterprise. Each of us have a preferred way to be communicated with and this is advertised by the way we speak. Over simplifying it for illustrative purposes, but specific language choices like “I believe” vs. “I think” or “We could” or “You should” provide advertisements of our communication preferences.

Algorithms are good at capturing these nuances. After analysing billions of conversations in the call centre arena what we’ve found is that when there is similar communication styles between a caller and agent business outcomes are improved. Specifically, conversations are more efficient or effortless, sales are higher and loyalty is stronger.

After analysing billions of conversations in the call centre arena what we’ve found is that when there is similar communication styles between a caller and agent business outcomes are improved.

Seth. Do you mine and model the conversational flow, that is, the paths between a statement and a response?

Andy. Yes, the personality model which inspired all of our work is called the Process Communications Model. So how we communicate, the tone, tempo, syntax, word-choice, triggers, responses, etc. are all considered under the “process.” The interaction between a caller and agent is very important. We not only measure the paths between a statement and response, but also analyse the time of occurrence during the interaction (e.g., beginning of a call or end of a call) as well as the magnitude or degree of the response.

For example, we’ll be able to identify that an agent reviewing a policy with a customer led to high-levels of customer distress that was not resolved by the time the call ended. Having the capability to analyse every second of every call and associate it to business outcomes is a very exciting and impactful capability. We see this capability enabling premier brands to have personalised conversations with each of their millions of consumers.

Seth. What's in your software toolkit?

Andy. We have a healthy mix of homegrown and open-source programming. At an architecture level, we have to capture voice and meta-data at very large scale, millions of calls per day, then transcribe speech to text, and then perform computational linguistic analysis to extract over 1,000 features per call. From a data and behavioural science aspect, we have a number of proprietary technologies but also leverage R, Python, and Apache applications within our technology stack.

Seth. Your Sentiment Analysis Symposium talk is titled "Mind if AI get Personal? Your Bot Needs to Understand Human Personality Better." Mattersight aims "to improve every interaction with every customer every time." How are insights and models deployed? At which customer touchpoints, and how do clients quantify results?

Andy. One of the key aspects of Mattersight’s culture is we’re value-driven. Our objective is to ensure the products and services we provide drive tangible business value, and fair to say we’re a little obsessed with that. Today we primarily engage in the call centre space, both inbound and outbound, but our vision is to help enterprises across all communication channels like chat, email and virtual agents or chatbots.

Ultimately, we’re analysing these conversations to determine which aspects are positive and which are negative. Once determined, we build products to help increase the positive and reduce the negative aspects. This can range from an intelligent routing solution that naturally pairs people of similar communication preferences to very sophisticated predictive modeling to identify at risk high-value customers. At the end of the day, we help our customers quantify results by ensuring that their business KPIs are improving and they can easily associate those improvements to our impact.

Seth. What's next? What will you be focused on a year hence, or two years or five?

Andy. I see a very large and untapped market around intelligent routing. Today, we rely on pressing buttons in an IVR to determine whom the caller can speak to and then randomly assigning that caller across a pool of call centre agents to best service, sell or retain them. Given the general state of technology that just seems outdated. So we’re focused on improving that experience. Given the amount of data that we can capture and the analytics we can run, we can predict with high accuracy who you are, why you are calling and the most appropriate experience, including the best agent, to route you to. I think that is a game changer in the market and improves the experience for the consumer when improving business outcomes.

Seth. Thanks Andy!

Meet Andy Traba at the Sentiment Analysis Symposium, March 26-27 in New York.

About Seth Grimes

Seth Grimes

Seth Grimes is the leading industry analyst covering natural language processing (NLP), text analytics, and sentiment analysis technologies and their business applications. He founded Washington DC based Alta Plana Corporation, an information technology strategy consultancy, in 1997. Seth created and organizes the Sentiment Analysis Symposium. He consults, writes, and speaks on business intelligence, data management and analysis systems, text mining, visualization, and related topics.

Replies

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.