2020 will be forever defined by COVID-19. The profound impact of the pandemic will leave lasting impressions on how we feel, how we behave as consumers, and how businesses respond in kind. According to Emerj, a leading analyst firm covering AI in industry, many businesses are shifting spend to prioritize risk reduction, operational efficiencies, building competitive advantages, and accelerating digital transformation. Budgets that remain, and as it relates to AI, are laser-focused on cutting costs while maintaining productivity. Nowhere is this more true than in customer service centers. Artificial intelligence can provide substantial impact by augmenting and automating functions of contact centers. Capabilities like natural language processing, sentiment analysis, virtual agents, and text mining are proven capabilities that provide meaningful business impact. As a result, we are seeing a big uptick in requests for call center AI requests. Though it sounds like a good idea on paper, many businesses still struggle to quantify the benefit that justifies spend on call center automation.
AI-enabled conversational agents, for example, are expected to handle 20% of all customer service requests by 2022 (Forbes, study)
In general, spending on AI in a COVID-19 climate seems like a moonshot. However, there can be strong business justifications made for augmenting this costly business function whereas AI can increase productivity and lower costs. Let’s explore how applying artificial intelligence to customer service centers can provide both strategic and financial benefit. Strategic v. Financial ROIGenerally, there are two ways to measure ROI - strategic ROI and financial ROI - when considering AI projects. Strategic ROI measures the degree which a business can differentiate their capabilities or accelerates in the achievement of a long-term, strategic objective. Here are a few examples of Strategic ROI measures:
Strategic ROI has more qualitative benefits. This is a better tool to prove benefits for emerging technology initiatives. Strategic ROI is easier to calculate which is useful when the speed to solution is an important drive. Financial ROI in some instances may be impossible to measure or take months to calculate effectively. So if you need to find a quick win, this is the way to go. By measuring Strategic ROI, it does not mean you are foregoing Financial ROI. In fact it is a precursor or early signal you are on your way to proving long-term, financial impact. On the other hand we have Financial ROI. This is the more universally accepted way of proving benefit where we measure business gained or costs saved by implementing the solution. Here are a few examples of Financial ROI:
While the qualitative nature of financial ROI makes it extremely valuable, it’s the most challenging to calculate. Most businesses lack formal processes to build a business case and doing so can be very labor-intensive. It requires many business units to provide input and agree on the outputs. In the context of machine learning projects, this means project initiation can be slowed months while the business tries to quantify the value. Where much of the market wants to find quick wins with ML projects, justifying the spend on a quick win can be at odds.
So that is why we believe in getting started with new artificial intelligence projects begins with finding Strategic ROI. As you parlay AI projects to programs you can begin to build Financial ROI models and track over time. This is especially the case when applying AI in areas like customer service and call centers.
Business Impact in Customer Service
Due to the recent effects of the current pandemic, much of the focus on remaining AI dollars is pointed towards back-office efficiencies, customer retention, and experience. It is not surprising that we are seeing increased attention and spend being pointed towards where they all converge. Call center (or customer service) AI has proven to be effective for augmenting agents and providing enhanced customer experiences.
There are several areas where strategic ROI can be readily found. One area of machine learning to explore it is natural language processing and text analytics. For example, you can convert audio recordings into text and run an analysis of the exchanges and outcomes. If the text transcripts are labeled by effectiveness or outcome, machine learning can be trained to classify (even predict) the quality of past and even real-time engagements. This capability would be essential to understand the quality of call center agents and understand where coaching opportunities may exist. You can even analyze aspects of the call like customer sentiment, time to resolution, and outcomes to understand what likely leads to successful events. This level of analysis may lead to the creation of engagement quotients which could provide objective measurement for how successful individual agents are.
Another strategic ROI opportunity exists within customer understanding. Similarly to using data and machine learning to provide insights on agent behavior, you can analyze customer data. Using natural language processing and Word2vec embeddings to understand trends in what makes customers frustrated or happy about your solution. This data is extremely important for the business and can augment everything from sales, marketing, customer support strategy, sales engagements, and even product development.
Though difficult to accurately measure, Financial ROI can be achieved with machine learning in call centers. One key area is call center automation. ML can be a critical tool to automate a subset of agent workflows that occupy a large percentage of time. First, you must know where the majority of the workloads and bottlenecks exist. If there are areas where 20% of the customer service requests occupy 80% percentage of agents time, those are likely target-rich for automation. Those workflows can be automated using a chatbot that can analyze customer requests and resolve low-level issues or triage to the appropriate agent. This was the case for Vitality, a health and wellness partner for the insurance industry. Vitality’s call center is using AI to prioritize their calling activity. They have done this by using AI to route agents’ calls. In doing so, Vitality has been able to increase conversion rates by 2X. Additionally, they have also reduced the cost of each new appointment by 24%. Financial ROI is achieved with this level of automation increases agent productivity or reduces headcount.
Humana Case Study:
For companies like health insurance giant Humana, overloaded service agents was a huge challenge. In 2016, for example, Humana call centers were flooded with 1 million calls every month from doctors and administrators. Yet 60% of those calls were simple queries for basic information, such as whether a patient’s policy covered a specific procedure or treatment. Callers wasted valuable time, and Humana paid more to outsourcing companies for all those agent-staffed phone calls.
To fix the problem, Humana partnered with IBM’s Data and AI Expert Labs to create a solution that could quickly identify and deliver the specific information that callers needed. IBM’s natural language understanding (NLU) software—using seven language models and two acoustic models—now translates more than 90% of spoken sentences it “hears” from customers. It can also understand the unique vernacular of particular insurance-related topics, such as distinguishing between a claim and a referral.
Rather than just get routed to another queue, the caller might get an answer such as “the co-pay for chiropractic visits is $100.” Since the system was rolled out in 2019, the percentage of callers who use the AI-enabled system has doubled, and the cost of running it has dropped by two-thirds. Health providers calling in today can complete their initial inquiry in less than two minutes—and don’t wait to talk to a live agent. (Forbes, 2020)
AI can be a powerful tool to enable the agent to be more effective in their roles. A chatbot can also exist as an internal tool that monitors the conversation and gives agents advice on how to resolve and de-escalate issues. This is not science fiction! One startup, Scale.ai, is using ML to augment inside sales teams. As the bot listens to conversations with interested prospects, the AI monitors to make sure all talking points are covered and gives suggestions on how to handle objections. AI can be used to monitor customer sentiment in real-time or make predictions around customer needs. Here, return will be measured in increased satisfaction, retention, faster time to resolution, and increased call volume per agent. These bots even learn and grow more effective over time as they observe more and more interactions and outcomes.
If you are interested in artificial intelligence but unclear how to measure its effectiveness for call centers, begin with an assessment of current performance. Try to calculate agent performance today and total costs of the program. Then set goals on what improvement is worth investment, regardless of technology. AI may not be the answer to completely accomplish every goal. But if it can incrementally solve some of the challenges, that can mean massive return potential.