The Power of AI In Customer Service
Think about the last time you hung up the phone and cursed. If you’re like us, it was likely the result of a poor customer service experience that may have included any (or all) of the following—an interminable wait, a long authentication process, poor triage/classification of a business issue, or an ineffective response. Although it may seem like these experiences are here to stay, as they have been over the last few decades, help is on the way in the form of a new wave of intelligent software solutions designed to let companies take customer service to a new level of effectiveness.
Within five years, the vast majority of customers will be able to get instant, effective support, and those that don’t will be quickly and efficiently routed to the right person for their needs. Essentially great customer service will be a common occurrence—companies interacting with active customers to ensure continued satisfaction, correct existing issues, or answer questions that may arise.
A main driver for this shift is that customer service is now a paramount consideration with the explosion in consumer choice, the increased visibility into companies through social media or forums, and publicly available NPS (Net Promoter Score). Entrenched industries increasingly have come under attack from upstarts—Dollar Shave Club taking on Gillette, Netflix replacing the cable operator, or Uber reducing car ownership—giving customers more choices and no reason to have to put up with a bad experience.
Today 66% of customers will change providers as a result of bad service, up dramatically from the 17% of customers that would switch back in 2005 according to Accenture. Unhappy customers have more leverage than ever with the ability to leave a scathing review on social media (United Airlines is the latest most newsworthy example.)
Companies are already implementing various techniques to ensure a quicker and more accurate resolution. These efforts include everything from customer service bots, knowledge bases, and real-time pop-up windows (a la WalkMe). Historically, nearly all customer service requests came in through voice or email, and companies could predictably scale up to meet demand. However, the complexity of touch points has increased meaningfully today, with requests arriving through text, live chat, email, social media, bots, and more. As a result, companies increasingly leverage a sophisticated multi-channel backend system that allows them to manage disparate requests, stitch together a unified view of the customer, and respond with a better knowledge of the customers’ past interactions.
In all of these areas, AI and ML are uniquely positioned to play a powerful role. First, the explosion of online data combined with advances in NLP (Natural Language Processing), voice recognition, and systems capable of processing unstructured data provides an opportunity for intelligent algorithms to learn in a way that simply wasn’t possible when the voice of the consumer was trapped, in a call center. Second, the nature of a customer service interaction is well suited for training AI systems, as it’s fairly easy for customers to rate whether or not a certain approach resolved their issue with a short feedback cycle. Contrast that with sales efforts that may drag on for months, and lack clear resolutions. In addition, while there is certainly a long tail of unique customer requests, a large percentage of issues cluster around particular problems (for example up to 30% of IT service desk volume can be traced back to password problems). While AI certainly won’t be able to solve all customer service problems, it is well suited to create a taxonomy of frequent issues and determine the best response which will put a giant dent in the CS burden. CS agents are likely equally unhappy answering the same questions on a daily basis.
The business imperative to improve customer service is very real. Unlike product or engineering teams that gain leverage as a business grows, or sales teams that benefit from an existing base, customer service costs scale almost linearly with top-line growth. With call costs often exceeding $5 (according to customer service consultancy F. Curtis Barry & Co.), a large company handling millions of calls annually would push millions of dollars directly to the bottom-line by deflecting even 10-20% of inquiries. However, despite the enormity of this number, it may just be the tip of the iceberg. Given the recurring nature of customers in many businesses (especially software), it’s likely that a company will incur the most costs from losing a customer due to bad service by losing a revenue stream that would otherwise last for many years (and having to reacquire a new customer).
The other advantage of adding AI to the CS workflow is to provide real time visibility into what CS issues are occurring and increasing/decreasing. These can be early indicators of product issues, as well as a great source of what things need to be designed back into the product to improve the customer experience. This capability moves CS from a firefighting role to really helping close the loop to create a better product and a happier, more profitable customer.
The long-term potential for disruption of customer service with AI is significant, but we are in the early innings of the transformation. Best practices today revolve around ensuring that the problem is addressed as quickly as possible using a variety of tools that create a digital workflow (e.g., Zendesk, Pagerduty), easy to use knowledge base, and a set of scripts to choose from. A manual process has been digitized in many cases, but there is a huge opportunity for real time intelligence and automation beyond that. With the help of AI, a new era in CS will feature smart, automated replies which will both alleviate the pain of a large cost center, help companies maintain good reputations in a highly visible world, and focus their resources on their core mission of delivering a great product or service.
There are some powerful emerging companies with key areas of innovation being multichannel CRMs, customer service deflection, and modern messaging platforms. There’s great potential in AI here. The best is yet to come.