Why Watson? More accuracy. Less training time. Insanely better CX results
By Nick Therkelsen-Terry | 4 minute read | September 21, 2017
– In the world of customer experience (CX), small improvements can yield big dividends. Standing in the way is just one thing: archaic systems.
– Nearly 80% of contact centers say their current customer service systems won’t meet their future needs.
– Traditional CX data systems classify and organize data for streamlined reporting, often leading to the oversimplification of complex customer interactions, leaving qualitative data ‘in the water,’ dark. Think of it as a “data berg.”
– Max Kelsen is using AI to shine a light on customers’ dark data to help them realize greater value from their entire data berg, not just the tip, creating custom knowledge domains 66% faster and with 97% accuracy.
In the world of customer experience (CX), small improvements can yield big dividends. A simple one percent improvement in first call response can yield hundreds of thousands of dollars in annual operational savings for the average call center according to the SQM Group. American Express reports that one happy customer can deliver as many as nine referrals for new business.
In the future, that will be no different, only more so, with nearly nine out of ten businesses expected to compete mainly on customer experience. Standing in the way is just one thing: archaic systems. Nearly 80% of contact centers say their current customer service systems won’t meet their future needs.
While intelligent chatbots, fueled by AI, are changing how customers interface with businesses, firms must also take steps to access and exploit the dark data they already have and that traditional data analytics systems can’t access and use. According to Forbes, nearly 80% of businesses are essentially blind to what’s happening across their unstructured data.
Think of it as a databerg. In most organizations, only the very tip of the vast amount of data available to an enterprise is visible, and therefore used. Traditional CX data systems classify and organize data for streamlined reporting, often leading to the oversimplification of complex customer interactions, essentially leaving qualitative data ‘in the water’ and unexamined.
Another challenge enterprises often face is that vast amounts of data are spread across the company, residing in disparate and often obscure databases and on different digital platforms. Organized according to different often competing priorities, enterprise content often includes a huge variety of data types, including complaints, product ratings, NPS scores, sales and call logs, verbatims, letters, news and blogs, social media and more.
Getting underneath the surface is well worth the challenge. You could say that this is where things get interesting.
Using AI to shine a light on your firm’s dark data
Our customers approach us because they’re looking for greater value from their entire data berg, not just the tip.
For example, in a typical feedback response, you’ll get a number and then some unstructured text. Traditional analysis will focus on the number: How was your experience? It was a five. And then move on.
What the company loses with that approach is that while yes the service was great, the customer also typed into their comment that … but I also dealt with a really rude sales rep.
This is really important feedback, and it’s entirely addressable without any new programs or extreme effort. But with traditional methods, it’s hidden from view.
We use AI, in particular IBM Watson, to not only find but also analyze that feedback at scale across vast amounts of data, allowing our customers to understand where their most critical CX improvements actually lie.
We had been exploring AI for at least 18 months before we settled on Watson. The factors that mattered most to us were the ability to:
- Train the AI system in custom domains quickly with very high accuracy and using our available resources and customer SMEs who are not all trained AI developers.
- Identify positive-negative sentiment to delineate positive and negative experiences contained in the same feedback comments or other data sources.
Watson, unlike other AI competitors, not only met but exceeded our requirements, allowing us to apply multidimensional analysis that understands emotion, sentiment and relationship on an entity-by-entity basis at a granular level, and create a custom knowledge domain in as little as two weeks.
We use custom knowledge domains because the lexicon of every business, government or subject is different, and we need to teach AI to talk the language of your business, so we can extract more nuanced and actionable insights.
The combination of domain customization and natural language understanding that Watson enables allows us to delineate the complex concepts in feedback and draw very specific insights.
Applying an “Insight Engine” to our Customer Data
Watson is literally a game-changer for us. It’s almost impossible to compare our engagements before and after Watson, the parameters are so different. For example:
- To create a custom knowledge domain manually using 300K words, or roughly 5K documents, would take a team of four with some programming expertise an average of six weeks, or longer depending on the scope of what Watson has to learn. With Watson we can do it in two weeks; that’s a 66% savings.
- Consider that Watson has to learn everything. It is entirely dependent on its trainers. So being able to rapidly train Watson with a small cadre of highly qualified subject matter experts and achieve 97% accuracy levels is frankly remarkable.
With Watson knowledge discovery, it’s like applying an insight engine to our customers’ data. In a matter of weeks, we can produce specific analyses that literally flip their databergs and bring the value of their hidden data to the surface.
Learn more about how our company is shining light on customer data.