Using AI for customer service and support isn’t a new phenomenon. Since the 80s, some form of IVR technology has been a part of the customer experience.
However, the recent boom in AI technology has made AI chatbots almost universal in scope. Most businesses today use some form of AI to perform customer support. But this creates a different problem.
Let’s say you’ve created an AI chatbot and want to see how customer service ROI has improved using the tool. How do you measure the difference? In this article, I would like to take you through some methodologies that have worked for our clients.
Customer Service and Support ROI
Measuring Performance
There’s a fairly simple way to calculate ROI. Every time you measure ROI, you’re calculating:
((Revenue – Expenses)/ Expenses) * 100
However, It’s always difficult to single out a specific department and find its ROI. For example, your sales team is responsible for a significant part of your revenue. Does that mean that none of your other departments are essential?
That’s why we use different performance indicators for every department. But, even this is difficult in customer service and support because you measure:
- CSAT (Customer Satisfaction Score) – The amount of customer satisfaction
- NPS (Net Promoter Score) – The number of people who would recommend your service to their friends or family.
- Average Resolution Time (ART) – The time you take to resolve a service ticket.
- First Response Time (FRT) – The time you take to answer a query you receive.
Here, you will get an objective number on your ticketing tool for ART and FRT. You will get CSAT and NPS through surveys and post-conversation ratings.
Measuring Impact on Revenue
Now, we have to connect this to revenue, and here’s what we know:
- Increasing CSAT by 1% increases revenue by 3-5%.
- Increasing NPS scores by 7% contributes to 1% growth in overall revenue.
- Reduction in ART and FRT is strongly correlated with increased CSAT and NPS.
This is how we get to a formula. We know that we can directly improve our resolution and response times through different methods and that it increases CSAT and NPS.
There’s a more subjective number where we provide better resolutions to customer questions. Still, given how customer support can vary from agent to agent, it is pretty difficult to optimize (enterprises invest millions of dollars every year trying to maximize this quality of the service).
Now, you have a basic framework that lets you calculate the individual impact of your customer support and service department. What happens when you add an AI chatbot into the mix?
The Impact of AI Chatbots
Customer service professionals are already familiar with the impact of AI in general. To recap, I would just like to present a very brief overview of the benefits in general in the context of the ROI calculations we did above:
- Faster FRT (First Response Time) – AI can respond almost instantaneously to any request received 24/7.
- Increased CSAT (Customer Satisfaction Score) – One of our clients, Lula, improved CSAT scores by 40% using an AI-powered chatbot.
- Decreased ART (Average Resolution Time) – AI Chatbots can resolve repetitive queries much faster than human agents. Automating up to 65% of the repetitive messages.
Now, we know that these statistics correlate directly with revenue. We can use these numbers to get to a possible ROI calculation. But, before that, we need to take measurements.
Calculating AI Chatbot ROI
Measuring Impact
It’s easy to deliver platitudes about the use cases and effectiveness of AI, but measuring the impact of these tools without a proper analytics tool is difficult.
You will have to measure a few things:
- Bot CSAT Rating – You need to measure the individual CSAT rating for your chatbot.
- Chatbot Containment Rate– The number of chats contained by the chatbot that don’t require a human agent.
- Bot Resolution Time – How long a bot takes to resolve a query.
Given that your First Response Time (FRT) from a chatbot will almost always be nearly instantaneous, treating it as a separate metric won’t make sense.
However, there are some auxiliary metrics that we track:
- Time Saved – Your average query takes 5 minutes to resolve, and you get 1000 messages/day. If a bot solves 600 of these queries, you’re saving:
600 * 5 = 3000 minutes/day
- Cost Saved – By automating the above 600 queries, your chatbot has freed up 3000 minutes/day of human time. Considering a minimum wage of $15/hour for an average agent, you’ve saved.
(3000 minutes/ 60 minutes/hour)* $15 dollars/hour = $750 dollars/day
So, we get two components that we have. The money that a chatbot saves you through automation is the money the AI chatbot makes by improving CSAT and resolution rates. Let’s calculate the final ROI, then.
Calculating ROI
So, we have two facets of ROI here. One is the simpler calculation of the cost saved. This is measured by:
Cost Savings from Chatbot = (Amount of Time Saved in Hours * Average Wage of an Agent)
The other facet is revenue increase. Since the dynamics of how CSAT, NPS, and Resolution Rates connect with revenue are dynamic, this is also difficult to measure.
Usually, you can do this over a long period. Still, if you need to justify your tech purchases to executives, you can focus on the delta you experienced in CSAT instead.
For this, you need to:
- Calculate the average CSAT for your AI chatbot.
- Calculate the average CSAT for your customer service agents.
- Calculate the weighted average for the actual CSAT and see how your chatbot affects overall performance.
- Repeat for the resolution rates.
With these two calculations, you should clearly understand the ROI an AI chatbot can deliver for your business.
Some Caveats
I have talked with many CX professionals, and these statistics do not capture the full picture talk about them:
- Customer Service Agents Still Handle Difficult Problems – While AI chatbots have the easier job of automating responses, most customer service professionals will deal with more critical issues. So, the cost savings will differ given human agents’ time at repetitive tasks.
- CSAT Might be Decreased due to Conflicting Policies – CSAT must also be measured objectively. Cases where the customer is dissatisfied with company policies (where the query will almost certainly go to a human agent), will naturally reduce CSAT.
As someone who has been around data for a long time, there is no quick fix to these caveats. The hope is that the outliers are averaged over a long duration, or we start looking at things with a magnifying glass.
However, on a surface level, the ROI calculation would give you the tools you need to advocate for more tools at the workplace,
Some Thoughts
As customer experience and support equations change, focusing on the business fundamentals is crucial. An AI chatbot is an added expense, and in my experience, while it saves money for businesses, finding the exact amount and performance has been consistently difficult.
With this article, I want to clarify how we calculate the ROI of a chatbot. While it doesn’t wholly bypass all caveats, I think CX professionals will find it helpful to adopt this model when practicing tech or AI advocacy in their business.
If you want to incorporate an AI chatbot into your workflow, you can try Kommunicate without a credit card for 30 days!