3 Steps to Calculate ROI for Customer Service AI & Chatbots

By Allan MacGowan - Nov. 13, 2017

According to a recent report from Forrester Research, artificial intelligence (AI) is already transforming customer service by taking over predictable and highly repetitive tasks, relieving support agents to focus on more advanced functions. The report states that nearly 50% of consumers already engage in automated conversations with intelligent virtual assistants. As I write this blog post, I’m reminded of the chatbot that KLM used to confirm booking information for the flight I recently took from Cardiff, Wales to Toronto, Canada.

Pure AI, in which machine learning is indistinguishable from - or even greater than that of humans - is apparently still years away. Until then, companies are turning to ‘pragmatic AI’, which focuses on specific applications that are commercially viable. A key question for companies, of course, is whether or not the return on investment (ROI) is sufficient enough to justify the cost of these applied AI solutions. In a general sense, the benefits of automation seem clear but, in our experience at CaféX, many customer support organizations still require detailed business case justifications before moving forward.

As such, I wanted to highlight the key aspects of one such ROI model that we’ve developed for CaféX Live Assist® for Microsoft Dynamics 365. In case CaféX is new to you, Live Assist allows customer service agents to engage and assist online visitors through live chat, co-browsing and proactive engagement campaigns - all from within Microsoft’s cloud-based CRM, Dynamics 365. Chatbots can be integrated with customer service interactions as well - acting either as virtual agents or as enablers of specific service tasks with the ability to escalate inquiries to live support agents. 

Step 1: Calculate Number of Reps Needed to Handle Peak Time Interactions by Channel

The ROI model hinges on the ability of Live Assist for Dynamics 365 to redirect customer conversations from higher to lower cost channels. Let’s look at an example: In its current state, the support organization shown below predominantly interacts with customers via email, web portal and voice calls. Using a combination of the Erlang-C algorithm plus industry standard values and customer supplied metrics, the model estimates the number of human agents currently needed (5,000) to process a specific number of conversations at a required service level during peak busy hour.


Step 2: Predict Number of Interactions Initiated by an Intelligent Chatbot vs. a Live Agent

The next step is to predict the future state after Live Assist is implemented, in which a percentage of voice, email and web portal interactions are redirected to live chat conversations, with chatbots acting as the initial point of contact. In this example, no voice calls are redirected, but 15% of emails and 10% of portal engagements are front-ended by chatbots, with 2 out of every 3 automated interactions being escalated at some point to a live chat (and cobrowse) agent. Using the new interaction volume by channel and re-running Erlang, the model estimates the number of live support agents needed, down from 5,000 to 4,390. 


Step 3: Compare Total Cost of Ownership (TCO) of Chatbots to Cost Savings over Time

To translate this agent optimization into realistic labor savings, we estimate the total cost per full-time agent each year and takes into account the fact that it may take months, even as much as a year, for the labor reduction to occur completely. So, benefits may not start accruing until the 4th month, reaching 100% only at the end of Year 1, for example. 
To offset the benefits further, the model incorporates all related costs for implementation of Live Assist for Dynamics 365, including subscription licenses, professional services, a lab testing environment, and administrative staff to maintain the solution. Even with these conservative assumptions and all-inclusive costs, the benefits clearly outweigh the costs - with payback on investment being achieved well within a year. In the example below, the support organization achieves a 3X return on investment during a 3-year period, with payback occurring after only 9 months. 

One could argue that the benefits can be even higher, given that the addition of co-browsing can help increase revenue upsell and cross-sell opportunities as agents visually guide visitors to relevant products at the point of purchase. However, with or without the additional functionality that Live Assist provides, we agree with Forrester that pragmatic AI is very compelling commercially and potentially transformative for customer service organizations.

If you have an interest in developing your own business case for AI-assisted customer service, please contact us to create an ROI model based on your organization’s unique requirements. Our customer service and contact center experts would be happy to show you the real benefits of intelligent chatbot integration.