Today's customer service organization is expected to do more with less: more high-value customer engagement, less effort. But there haven't always been the tools, technology, and budget to succeed. It's been the heroic efforts of service reps and managers that have kept customers happy day in and day out.
Now, with the influx of both internal and external data, support is sitting on a treasure trove of customer knowledge. That's an opportunity to better capture the voice of the customer and fully understand their needs and a burden, requiring heavy lifting to figure out how to track every interaction, understand all the pieces of the engagement, and act on them.
Approaches like skills-based routing, self-service IVRs, and crowdsourced replies are enabling customer service organizations to create new efficiencies. Also emerging as an automation solution of choice is machine learning.
What is machine learning?
At its core, machine learning analyzes past customer and agent behavior to understand what actions and data led to what outcomes (whether that's a highly satisfied customer or a long time to first response), then interprets each new interaction and recommends how best to engage with each individual customer to achieve success. All the while, machine learning is continually getting smarter, updating its recommendations based on the latest outcomes to remain relevant without requiring manual changes and inputs.
As an intelligent layer sitting on top of existing customer-facing solutions (e.g. CRM, marketing automation, support systems), machine learning automatically extracts patterns from complex data. This augments current capabilities, allowing companies to be systematically data-driven to increase productivity and capacity while contributing to a high-performance customer service organization.
Unlocking productivity with machine learning
Basic automation often gets a bad rap since historically there's been a trade-off between efficiency and customer satisfaction. However, by creating self-learning systems through machine learning, the ability to drive greater efficiency and deliver personalized, high-touch customer interactions are not mutually exclusive.
Are you tasked with a hire-and-ramp plan to address increasing ticket volume? Shooting for a bonus based on your ability to increase productivity? Or are your customer satisfaction scores in need of a boost? In this environment, self-learning systems that augment human decision-making are your new best friend. Machine learning can help agents work faster, freeing up time for them to work more tickets or devote more time to higher-value customer issues.. They can also help agents work smarter. By watching how your best agents do their jobs and intuiting what makes the customer experience great, machine learning can help automate and share that approach throughout the organization.
For example, machine learning creates the opportunity to automatically read and evaluate a customer inquiry and select from a library of responses using the same, or even better, judgment than a human. Machine learning can look for consistencies in the way your people are replying (towards a good outcome) and recommend that these be part of the official library. It is an opportunity to create consistency and a faster ramp-up for new or novice agents to reach the level of the experts.
<p">As a result, machine learning can also be the difference between a reactive and a proactive service organization, automatically discerning predictive patterns in complex data that goes beyond typical levels of human interpretation.
Consider a freemium business model, such as online gaming, where the majority of games are free but additional games cost virtual currency. A business like this is often faced with an overwhelming number of support tickets from free players; but they want to focus efforts on the requests from players looking to upgrade to paying users. Machine learning can automatically analyze data, such as the text of email support requests, ticket submission data (e.g. date, channel), product data, as well as customer demographic and activity data, to find contextual patterns related to billing issues and the purchase of that virtual currency. It can then help them be forward-looking to make predictions about who might have trouble paying, efficiently prioritize those higher-value customers, and ultimately head off any issues before the customer has to make contact.
Making machine learning work for you
As you're looking to apply machine learning to turbo-charge your customer service processes, use the following checklist to get started.
- Know your business processes. Machine learning is best suited to repetitive processes that are relatively mature and stable, with a solid history of engagement and outcome data, such as routing or responding with templates and macros.
- Know your data. Machine learning can work with your existing systems and any data type (even text) or volume. Take inventory of your data, including where it lives and how it’s being used to make decisions; that, combined with understanding your business processes, will help you prioritize where to layer on machine learning.
- Get crystal clear on the outcomes. The outcome is the nucleus around which machine learning operates, so be sure you have a clear focus on the outcome for which you're striving (such as Who is the best agent to work on a particular customer issue? or What templates or knowledge base articles are best suited to solve a customer's question?) By understanding past outcomes, machine learning can tie your data together in meaningful ways to drive better future outcomes.
Augmenting current approaches with machine learning can lead to smart, efficient processes that directly align with customer needs, all in the name of high-quality customer service.
The series conclusion next week will share advice on how to take advantage of machine learning to extend the value of customer service across the entire business, enabling customer service as the hub of customer success.
Jeff Erhardt is CEO of Wise.io, which delivers predictive applications built on machine learning technology. Previously, he was chief operations officer at Revolution Analytics. He holds an engineering degree from Cornell and master's degree in business administration from the Wharton School.