Share this post:
Companies use all kinds of tools to find out what customers think about their products and services so they can improve the overall customer experience (CX). They mine data from social media, conduct focus groups and surveys, and collect feedback from customer service reps.
Eventually, all this information can reveal what customers are thinking and feeling. But businesses are often left wondering why their customers feel the way they do. Their response may be another round of focus groups or surveys to gather even more data. In the meantime, the business continues to miss the mark with its customers because it still hasn’t solved the why issue.
Getting to the why in untapped “dark data” with AI
At cognitiveCX, we believe that the why exists in data that organizations already have. We like to talk about something we call “dark data.” That’s all the unstructured data companies have collected from call center logs, focus groups and surveys, and comments made in emails, letters or on social media. We call it “dark” because it usually lies dormant and untapped. But that’s where the real customer insight comes from because that’s where customers express their emotions and sentiment about a company, its products or services and its employees.
Until now, it’s been impossible to read through or analyze all this dark data and view it in context to generate any meaningful insights in a timely manner. But applying natural language processing and AI technology to the data has changed all that. Today, AI allows people to understand the why behind their data at scale.
When we were ready to launch an AI-driven customer experience (CX) analytics service, we tested several commercial and open source solutions, and IBM Watson consistently outperformed them. We’re now using the IBM Watson Discovery Service solution, hosted on the IBM Cloud, to power our cognitiveCX platform-as-a-service (PaaS) offering.
Using the IBM Watson Knowledge Studio application, we can quickly annotate and train Watson on domain-specific documentation. We then use the IBM Watson Natural Language Understanding (NLU) application programming interface (API) to interrogate unstructured text on an entity-by-entity basis for tone, intent, sentiment, service area and contextual details.
Shining light on the right data
We can get amazing insight by applying AI to our clients’ data. And I mean all their data. Before AI, we didn’t have the capacity to analyze the kinds of data volumes we’re looking at today, and those volumes continue to grow. With Watson powering our cognitiveCX solution, we don’t need to sample. We can look at everything – millions, maybe billions of documents, calls, posts, news articles – it doesn’t matter. We’re no longer looking at 1,000 social media posts out of hundreds of thousands, or at 100 people surveyed in focus groups. We’re looking at every relevant customer comment about a product.
And even with that huge amount of data, we know that the data we’re analyzing is relevant because Watson Natural Language Classifier is so powerful. It filters out posts that are irrelevant to the conversation, such as company posts that are just really advertisements. And it allows us to focus entirely on actual customer responses and social sentiment to get to the why.
Getting the why right
In one project, we did an analysis for a government body in Australia that was considering whether to expand bike lanes and paths in an urban area. Citizen feedback collected through surveys and other traditional means had revealed high levels of anger and frustration concerning cycling. The government thought perhaps that the negative sentiment came from frustrated cyclists who wanted better infrastructure or more paths. But our analysis showed that the negativity emanated mainly from pedestrians who used the shared walking and cycling paths, and were upset at cyclists’ disregard for traffic rules and speed limits. Clearly, then, the problem wasn’t a lack of bike paths.
That’s the difference between typical sentiment analysis and applying AI, with some very smart language filters, to the same data. Instead of investing in new, unnecessary infrastructure, government leaders could directly address the root cause of the negativity surrounding bike paths while protecting citizen safety.
In today’s economic climate, companies, and even governments, can’t afford to miss the mark when dealing with customer or citizen experience. But they have to get the why right before they can get to the how of fixing the problem. Now we’ve built a way to help companies extract value from all that dark data they’ve had just lying dormant. Using AI for customer service enabled us to extract key insights to deliver meaningful value.
Hear Nicholas Therkelsen-Terry and Samuel Irvine Casey discuss how cognitiveCX uses IBM Watson on IBM Cloud to develop the AI-driven customer experience analytics service: