Patrick Nguyen, Chief Technology Officer
When COVID-19 forced offices worldwide to shut down, contact centers had to quickly adapt. Millions of agents were suddenly forced to go home, and even in regions where contact centers were designated as essential services, many employees have been unable or unwilling to come to work due to transportation shutdowns and health concerns.
Before the pandemic, the agent deployment model relied on physical offices and centers, with agents all stationed next to each other, and supervisors standing nearby. Remote or at-home agents represented a small fraction of the total agent population (around 15% in the US, but much lower in outsourcing locations such as India or The Philippines). So most companies were unprepared for the sudden shift and had to scramble, both technologically and culturally to get their agents up and running and able to work from home.
Thankfully, companies eventually made the pivot. But as they were doing so, customer experience took a hit. And in many cases, that experience has failed to return to previous levels. At first, customers faced unbearable wait times as companies struggled to handle a surge of calls with a depleted agent pool. Then, when balance was restored between call volumes and agent capacity, customers were impacted by the limits that many companies imposed on the activities of at-home agents. So what best practices can we draw from companies that have successfully navigated this transition?
When a crisis increases call volumes while at the same time reducing agent availability, basic indicators of customer pain such as wait times can go through the roof. In the early days of COVID-19, wait times rose precipitously as more and more callers were held in queue to be served by an agent workforce that was just ramping up remotely. The urgent priority was to get those times down again and provide a minimally acceptable level of service rather than simply hang up on callers.
Any rapid increase in at-home agent capacity requires a concerted effort across technology, security and human resource teams. Computers need to be set up and shipped to homes, and networking, security and agent management systems have to be established. In areas with less reliable public infrastructure, additional equipment is potentially required, such as mobile hotspots and uninterruptible power supplies.
Increasing the ability for customers to self-serve is also part of the solution. Calls that are fully or partially automated in an IVR reduce the load on human agents. While traditional IVR systems may require weeks to update, some modern systems allow new functions to be deployed in just hours or days. Companies with newer systems were able to quickly insert call flows to automate new call types, such as questions about COVID-related cancellations, deferrals or refunds.
Deflecting calls to digital channels is another important strategy. Digital conversations can be conducted in live chat, which occur in a single session with short agent response times, or in messaging (through services such as Apple Business Chat, Google Business Messages, Facebook Messenger, WhatsApp, etc.), which can be paused and resumed over hours or days depending on the availability of the customer or the agent. Both live chat and messaging can support customers through a blend of bot automation and human assistance.
Digital channels have a number of advantages compared to the phone channel in dealing with spikes in customer demand:
Once the wait time is flattened, customers can actually get through to agents working from home. From that point, however, many companies fail to return to "normal" levels of customer satisfaction and operational effectiveness due to challenges with supervising and supporting a large, distributed workforce of at-home agents. To solve these problems, automation and collaboration technologies can be deployed to reach levels of security, productivity and employee engagement that are comparable to those in a traditional call center.
In a call center, many security measures are in place to minimize inappropriate behavior. Video monitoring, biometric access, device restrictions (e.g., no recording devices or cameras), and the proximity of supervisors and other team members reduce the risk of illegal activity. For agents working from home, companies may not have equivalent measures. As a result, these agents are excluded from performing sensitive transactions such as collecting credit card and financial information. These constraints prevent first contact resolution and increase customer effort (forcing customers to take extra steps elsewhere to complete their transaction).
To overcome these limitations, AI-powered monitoring can verify workspace and transaction compliance, two areas of heightened security risk for remote agents. The agent’s physical workspace should comply with restrictions equivalent to those at the office: no personal communications devices, no recording devices, no writing materials and no other individuals. Using an attached camera, images or video of the agent’s work environment can be analyzed by computer vision to detect unauthorized objects or individuals, and alert human supervisors of potential violations.
Transaction compliance ensures policies are followed and transactions are reconciled to customer requests. Using automated transcription, a phone call can be converted to separate text lines for the customer and the agent side of the conversation. Natural language processing can then be applied to detect the customer intent, extract the entities associated with the customer’s request (e.g., product names, order quantity, dollar amounts), and classify each step performed by the agent. Potential policy and transaction anomalies are flagged and sent to a human auditor for verification. Using manual compliance, a major retailer was only able to audit 20 agent interactions per auditor hour. With AI-based compliance, that retailer was able to monitor nearly 200,000 interactions per auditor hour.
After security, the biggest challenge is supporting and developing remote agents. At the office, an agent can easily lean over to get help from a fellow agent or supervisor. Using AI, an agent working from home can receive comparable on-the-job training through automated recommendations on the best response or action. While an agent is talking to a customer, the call can be transcribed and suggestions can be retrieved from knowledge bases or generated from deep learning models. By training these AI models on a very large set of highly-rated conversations, at-home agents can benefit from the experience of the best agents throughout the organization, thus accessing a support network that extends beyond their immediate work group.
Agent development is all about replacing the collaboration that is commonplace in the office. To be effective, remote agents need to feel part of a larger organization and culture that is committed to clients, and to each other. Collaborative tools such as text and video chat enable agents to ask each other questions, conduct team huddles, and seek supervisor approval. Additionally, there needs to be social engagement. Even something as simple as a happy hour or talent contest can really help to boost the morale of agents working from home.
The era in which 85%–95% of agents worked in a contact center is over. Work-from-home has become an important part of the agent mix. To prepare for continuity of customer service when the next pandemic hits, companies need to invest in automation and collaboration technologies that make remote agents more effective, productive and engaged. These investments will accelerate contact center transformation by not only making companies more resilient to future shocks, but also provide the building blocks for lasting improvements in customer experience.
Patrick is Chief Technology Officer at 7.ai. He joined the company in 2011 with the acquisition of Voxify, where he led product and platform development teams in his role as CTO.