Complaint handling needs AI treatment

Artificial intelligence and cognitive learning technologies will shape the future of complaint handling says Andrew Anderson, overcoming the problems existing systems have in interpreting unstructured data.

Customer service and complaint handling is a hot topic with utility providers and consumers.

With Ofgem stating in its last formal report at the end of 2014 that over half of those that complained to their utility were not satisfied with the way in which they were handled, it is more important than ever to explore the ways in which failing systems can be improved to meet expectations. The industry is ready for change. British Gas has committed to invest £50 million in improving its customer services, and similar announcement will no doubt follow across the sector.  

Throwing money at the challenge of better complaint handling is not necessarily the answer however.

At the heart of the problem lies our innate human tendency to express dissatisfaction emotively. Whether over the phone or, increasingly, via email, we want to use descriptive words to communicate our unique circumstances and how a particular issue has made us feel.

“As the cost of outsourcing the interpretation of unstructured data mounts in the utilities sector, the need to find better technology solutions becomes more pressing.”

These sorts of correspondences are often referred to as “unstructured” – as opposed to “structured” language, based on set fields or multiple choice options which are easily interpreted in the binary language of computers.

It is the responsibility of companies to handle and respond to these “unstructured” correspondences correctly and in a timely manner, primarily to resolve the customer queries or complaints, but also because doing so will bring reputational and bottom-line value.

The trouble is that, to date, it has been difficult to find technologies which are effective in handling unstructured data.

Established options, such as Customer Relationship Management (CRMs) systems and Robotic Process Automation (RPA) systems, often fall short of the mark. The former, while useful for recording and storing structured information across internal business operations, have little ability to understand unstructured data, and so cannot prompt or deliver actions which respond to these customer communications. Furthermore, call centre information isn’t always readily accessible on central CRM systems and therefore customers find themselves having to fill in the gaps in data when they communicate via different complaint channels.

RPA systems meanwhile, offer the opportunity to automate the handling of structured data between systems. It is sometimes possible to convert unstructured data into relevant structured formats, however doing so requires time and cost intensive manual labour which is intrinsically prone to human error.

As the cost of outsourcing the interpretation of unstructured data mounts in the utilities sector, the need to find better technology solutions becomes more pressing.

Looking to other sectors, utilities could learn a lot from the way in which the retail, rail, logistics and financial services industries are experimenting with artificial intelligence and cognitive learning technologies in order to improve their complaint handling performance.

Automation and specifically cognitive learning technologies enable organisations to consume and understand what their customers are saying regardless of the fact that the content is unstructured and unpredictable. More importantly, the technology is able to learn, so the work force doesn’t need to scale to cope with growth or even unexpected surges in demand. For example the cognitive learning technology and automation would be useful when a company receives a surge of complaints due to a rail strike, delays or even a faulty product. The system implemented could upscale accordingly without the need to revert to the outsourcing of manual labour.

AI will enable organisation to understand large volumes of unstructured customer interactions in real-time, and to respond appropriately. A growing proportion of this correspondence will be dealt with without any human intervention, delivering swift, consistent, yet personal service at a lower operational cost. Using AI will also increase the productivity and effectiveness on issues which do still require human intervention is required.

With the plethora of solutions available it is not always easy to find the right solution for customer service data handling. But what is clear is that utilities companies need to respond much quicker and more attentively to their customer needs and expectations.

Kneejerk reactions will not be sufficient and do not address the root cause of an epidemic of poor customer service within the industry.