Tech’s Countdown to Zero-Touch Customer Service
July 23, 2020
July 23, 2020
Imagine a mobile coaching app for customer service. A support engineer receives a help ticket from a tech customer. But instead of attempting to piecemeal a response with ad hoc documents and random forum discussions, the agent makes a single click to an Alexa-like app. An Alexa with an engineering degree, who is always up-to-date on your product’s technology and an unparalleled expert on how to use it. She can instantly and perfectly problem solve and coach. With the utmost ease, she can give your agents and customers the answers they need now.
In our new era of Enterprise 4.0, this is how Peritus plans to disrupt the future of customer service for tech enterprises (and beyond). The ambitious startup is aggressively building an AI-powered platform for “expert” customer service — automating specialist knowledge for tech support teams.
In today’s “new normal” with Covid, we can all too easily relate to the concept of zero touch and the increasing need for digitization. In the world of customer support, the reality is that customers feel frustrated and exhausted by interactions with service agents. And they expect more, especially from tech providers, who they believe should be leveraging the latest in technology and the Cloud. Tech consumers want a zero-touch support experience that’s automated, seamless, and personalized.
In actuality, when a customer encounters a problem using a tech product, it takes an average of 14 days (or longer for a complex problem) to get an issue resolved. And when it comes to customer questions posed on product forums, a staggering 65-80 percent go unanswered. And if they actually get answered, the average response time is 8-14 days.
It’s no wonder that opening a technical support ticket has become a customer’s last resort. According to Peritus founder, Robin Purohit: “When a customer has a problem with a tech product, they try to solve the problem themselves, poking around on the website, researching vendor support resources and going to multiple forums to answer their own question. But when they ultimately can’t find an actionable answer, they finally open a case, only to wait for Support to hopefully resolve it over a couple of weeks.”
Purohit explains that the root cause of the problem is a lack of consolidated expert knowledge in customer support. “Answers to issues are in somebody’s head, random documents, or even archived tickets.” He continues, “there’s no real way of automatically going across all sources of knowledge, for a customer service agent to be radically more effective at resolving a customer’s problem or for a customer to be more self-sufficient with actionable advice.”
Support teams try to find answers with crude search methods and random resources. Even though their performance is measured on response time for closing tickets, they rely on personal knowledge, internal chat specialists, and old customer records. And if they just can’t solve the problem, it gets escalated to someone else “more qualified.” Purohit describes it as“an ad hoc process with a lot of scrambling.”
On top of that, technology is moving fast. There’s constantly a lag between people on the front lines and their understanding of the technology. For Purohit, something is fundamentally broken in the relationship between support and product teams. “We don’t do a good enough job as product teams, to empower our support teams fast or well enough about new products we put out in the market.”
When you introduce a new tech product to the market, customers will inevitably have questions testing and figuring out how to use it. But if support teams aren’t equipped to smooth over any friction with a customer’s adoption of the new technology, customer experience deeply suffers.
In addition to leveraging training sessions and product guides, companies are likely to put their smartest product people in Support for the time being. But throwing more people at the problem is a very expensive and “labour-centric process.” According to Purohit: “Companies have too many cases and that’s draining their best people from working on high difficulty problems.”
Indeed, all of this comes at a substantial cost for tech enterprises. The cost for handling support cases ranges from tens of millions to hundreds of millions of dollars. A typical enterprise vendor will have up to one million cases come in per year. The average cost in the industry is roughly 400 dollars per case. For a large vendor, that amounts to 400 million dollars. “With these kinds of costs in customer support, if you can move the needle, even incrementally, you have a great ROI success story,” Purohit surmises.
Increasingly more companies are looking to customer service infrastructure as a strategic revenue driver. For the SaaS industry, a focus on customer experience has proven to increase recurring revenue and company valuation. Clearly, a customer’s experience with Support, greatly impacts their level of satisfaction with a tech product.
When a company launches a new product into the market, less than satisfactory customer support can significantly slow product growth and its perception in the market. The stats show that it can slow a product’s initial revenue curve, by as much as 6-18 months. “If you don’t get customer service right with a product launch, you can seriously miss your forecasts,” Purohit warns.
The bottom line is tech’s exponential growth in complexity and innovation is outpacing the availability and actionability of expert knowledge. Digital enterprises don’t have an automated process, to capture and distill product knowledge for actionable customer support. The traditional approaches of creating tickets and escalations, or even waiting for just the right expert, no longer work.
Peritus is imagining a world beyond tickets — with automated, proactive, expert customer service. Where infrastructure is continuously learning how to solve customer problems better and faster. The aim is to transform front-line tech support into experts, and to reduce the burden of escalation on engineers, so they can innovate.
Named after the Latin word for “expert,” Peritus leverages AI to automate expert knowledge for customer service and its infrastructure. They do this by consolidating data across public and private data sources, using knowledge graphs (the same internet skill learning and relationship building technology as Google and Facebook). With these algorithms, they build a model specific to a customer’s specific product, to predict the best potential responses in any support case— from 1:1 problem solving to public forum answers.
The first step is empowering the AI engine with novel data sets and proprietary data. Peritus is doing this by partnering with early adopter marquee customers, who share their vision of transforming the tech industry. “We’re training our AI knowledge engine, by collaborating with the biggest and most strategic vendors, the who’s who of the tech industry,” Purohit explains.
As a startup, leveraging big data and new data sets has also been crucial to building trust with early adopters. “We start by processing a company’s public data (from public forums and documents), to demonstrate our initial value. And that usually incites them to share their private data, to get even better end results,” he explains.
And so far, the results are impressive. In production pilots with early adopters Peritus has already improved support team productivity, by an unprecedented 20 percent.
While AI powers the Peritus platform, humans are still at the heart of the process. As a human-in-the-loop machine learning system, the platform recognizes and understands human concepts and behaviors, by using the collective knowledge of subject matter experts and capturing user feedback.
Purohit explains: “To make sense of the data the machine is processing and following, it needs human annotation to understand the meaning of things in different contexts. AI labelling helps the machine understand the context of the data.”
Peritus also collects recommendations from subject experts, enlisting them to continually review, augment, and rate the machine’s automated responses. The aim is to constantly improve the content, tone, and voice of the machine’s support advice, so it can deliver the highest quality responses for customer cases and public forums.
The AI engine is also always learning from the feedback of user behavior. “When a support engineer or end customer receives our recommendations, we track their behavior. Every time they act on a recommendation, give kudos, or act by clicking on something, these ‘signals’ come back into our algorithms and refine the way we make recommendations.”
Purohit explains, “It’s a staged process, from curating all the data to validating the algorithms, and then refining predictions for the most accurate and useful recommendations.”
Peritus is embedding AI recommendations into today’s most popular tools for customer service and IT agents, including Salesforce and Lithium. Support agents see automated recommendations in browser plugins, as answer bots in forums and sidebars in case management consoles, as well as in an early mobile coaching app. “We don’t want people to have to learn a new system to get value out of what we’re doing initially. We’re working with existing legacy systems. We want to drive adoption, by injecting seamlessly into the existing user experience paradigm,” says Purohit.
In the next stage, they plan to deliver a brand new user experience model for a higher-value workflow, completely powered by AI applications. “The future is a brand new, much more dynamic problem solving user interface.” he explains. The plan is to launch a mobile digital app (fully integrated with case management systems), to closely coach support engineers and IT workers.
Peritus is imagining the equivalent of an Alexa mobile app with high-level technical expertise, updated in real-time on a product’s latest technology. She will be able to translate complex issues into simple and actionable solutions. And she’ll even ask users her own questions, “because algorithms get smarter from the smartest people.” Closing the machine-human loop, Peritus will gamify and solicit knowledge from app users, to make automation increasingly more intelligent.
Purohit concludes: “This is a great example of human augmentation. We’re bringing them proposed answers, they’re augmenting them, and in the process, automating our algorithm.The magic is that it’s not just a simple bot, but a knowledge engine that’s highly trained at understanding what your product is and helping problem solve in the right context.”
Alexa, how can a tech enterprise automate expert knowledge for zero-touch customer service? The answer looks like Peritus.