With business spending projected to grow to $47 billion by 2020, Artificial Intelligence is swiftly embedding itself into all sorts of industries, but for Customer Experience professionals the benefits are particularly heightened - data is power after all. With the promise of deeper understanding through broader customer data, CX professionals are embracing the opportunities AI offers - but with this improved technology we also have begun to see the limitations of existing widely-used methods and technologies.
The CX software industry has certainly provided breakthroughs for organisations and CX professionals in automating, collecting, organising and disseminating customer data at scale. While they may carry a range of labels; ‘text analytics’ capability is now considered a commonplace feature and a must-have tool for analysts. But now with the advent of customer-focused AI we can start to reflect on the limited effectiveness that text analytics offers in directing actions to the right problems.
The key limitation of text analytics is right there in its name - it’s only the analysis of text. Vendors in the world of text analytics attempt to provide context to text data usually by introducing sentiment into the equation, but the technique consistently reveals more questions than it does answers. The outcomes of text analytics become part of subsequent analytical efforts and a potentially long tail of professional services as organisations struggle to apply often generalistic insights to make decisions and take action.
The truth is that just like manual coding based methods, text analytics can do some of the heavy lifting but still struggles to convey the real ‘meaning’ and context behind customer conversations. Real meaning is often very different to how customers actually articulate themselves (verbally or in writing) or how they describe experiences and interactions. It differs again in how this meaning might be recalled and recorded by customer facing staff. Real meaning is also completely unique in the context of each organisation and their products, channels, services and processes.
The more you understand what’s going on inside text analytics the easier it is to see the gaps that make it inherently untrustworthy and how this might lead to poor CX decisions. Probing further, text analytics is one of many forms of ‘rules based’ analytics, relying on pre-existing understanding of the types of patterns and concepts that might already exist to create a set of rules that text is matched against. In general terms, you need to know what you’re looking for to find it. Patterns, concepts and insights that are ‘truly unknown’ are undetectable to rules based analytics.
The ability of AI to define real meaning was demonstrated in a cognitive analysis undertaken by a mobile telecomms operator. Struggling to understand a sharp upturn over a few days in billing-related complaints to its contact centres, this analysis used AI in the form of Touchpoint Group’s cognitive analytics product Ipiphany. Drawing on both structured and unstructured data, Ipiphany interrogated call notes and customer transcripts simultaneously with specific customer and contact centre data. It easily confirmed that the source of the issue was customer’s unexpectedly high mobile data consumption, but the real insight was uncovered beyond the reach of text analytics. The root cause of the billing issues was the viral gaming phenomenon now familiar to the world as Pokemon Go, but being completely unknown at the time, rules could not exist to detect that “Pokemon” and “Go” may belong together as a text pattern. By finding and defining this root cause, the telco was however able to take confident, precise action ahead of its competitors, communicating with and proactively offering data plan changes to relevant customer segments.
The immediate identification of previously unknown meaning and the speed and surgical precision of subsequent market-facing action illustrates one of the clear differences in real world value between text analytics and AI. Take Ipiphany; as well as leveraging industry specific ontologies for Customer Service, Retail Banking and Telecommunications to name a few, it uniquely automates natural language processing and text analytics with semantic understanding and machine learning to uncover root causes and underlying meaning from customer data. Providing insight that is contextual, measurable and fast, Ipiphany aims to transform how, when and where customer insights are used in organisations and in decision making.
Contrasting text analytics with AI it is easy to see how CX decision-making supported by text analytics based insight is vulnerable. At best, decisions and actions will be less timely, less effective, significantly less precise and consequently inefficient. At worst they are the misunderstood response to a misunderstood problem. As Nicola Morini, Managing Director of AI at Accenture recently said “AI is about data and organization. The more data that is accumulated then applied, the higher the barrier against competitors becomes.”