In today's episode of the Support Insights Podcast we talk to Clemens Behrend, ex-Global Director of Customer Support at Bitpanda, about how he harnessed AI in his support department to share insights across teams and identify customer churn risk, and what he recommends for other businesses looking to do the same.
Having explored and implemented many different AI tools, Clemens talks us through how he has harnessed automation to identify and solve the most frustrating problems for both agents and customers, resulting in higher employee and customer satisfaction and improved customer effort scores.
Building a central dashboard allowed Clemens to share insights on customer sentiment, predict customer behaviours and monitor changes to Bitcoin price affecting customer happiness.
Find out how you can utilise AI to share reports across departments, how to quantify AI insights into business goals, and how to make a business case to other stakeholders to make the investment in tools.
Watch the full episode or read the highlights below 👇🏻
To explain it quickly to the audience - AI is usually a software that executes a process which would otherwise require human intelligence to perform.
It'a not the same as automation where you pre-defined a rule and say "click this button, then click this button". AI is a little bit more complex.
My approach to AI is that if you implement a proper tool to automate your biggest pain points, that will enable you to get more free time and more resources, meaning the upside and rework potential is very high.
AI should assist your customer support agent but not replace them.
One great example of what you can do with AI is, you start by making it super easy for your customers to contact you, encourage complaints, encourage those contact volumes, that way you have more conversations to analyse.
You can then deal with the ticket in whatever form it comes in while allowing AI to automatically classify it into categories. You can then gain a true representation of quantified qualitative data because the AI has read your tickets and tagged them automatically, so you can go and fix those issues.
The good thing is, this isn't limited to tickets or emails, it can track your social media conversations too to get a realistic picture of how your community is feeling and what you should focus on next.
AI can also stop you from having to rely on CSAT surveys, as it just makes a sentiment analysis of all comments that you have in your inbox to tell you how your customers are feeling.
AI can also ensure that you're assigning tickets to the right agents based on the historic backlog of the agents' solved tickets. You then don't rely on any hard coded skill set, and this also removes a lot of frustration in your customer support team because they don't have to bother with tickets that they're not experienced in.
We had hard-coded some of our solutions because we still relied heavily on a contact form, but we realised very quickly from the feedback we were getting that we should focus on improving the customer effort score.
Since we committed on this score, we realised more and more that we needed solutions that would work with AI, otherwise we would have to continue with this very difficult solution of contact forms, hard coded routings and relying too much on CSAT surveys.
All of these tools I mention in this podcast were either implemented or on our roadmap, except for auto-routing as we weren't considering that yet.
The impact on the team was that employee satisfaction went up.
Insights are an exciting topic, people liked it, people were curious about it, and it just motivated them.
We always focused on the most frustrating processes and tried to automate them with the help of AI. We found that usually the processes which were the most painful for the customer support agents were also the most painful for the customers.
We built a dashboard with an overview of certain topics, the customer sentiment, the volume for a certain timeframe, the user group, and churn risk.
We identified that before users are stopping our service, they usually complain and they use specific wording.
For instance if a lot of churning customers are complaining about the payment method not working and so on in a certain way, we can identify which users are likely to quit our service from their complaints.
The most valuable insight for me being in the crypto industry, which we also had on the dashboard, was we had the sentiment in one chart compared to bit Bitcoin price in another. This is because we found that the sentiment of our users was following Bitcoin price with two weeks delay.
Whenever Bitcoin crashed, people ran red, more complaints came in.
When prices were going up, CSAT and everything would follow as well.
Because of the influence of the Bitcoin price, we knew that CSAT was not really a reliable metric. So therefore, sentiment was better for us to monitor.
We also made sure that product management could access the data on their own so they could come up with ideas based on customer feedback and not rely on customer support telling them what customers are saying.
This also helped the customers support agents because they didn't need to run to product management and tell them what's not working anymore.
With this style of reporting, we now focus on retaining users together - customer support and product management.
Music: Savour The Moment by Shane Ivers - https://www.silvermansound.com