Our new feature, Root Cause AI, analyzes the reasons behind long AHT and finds ways to save operational costs
How can you reduce operational costs at a call center? Historically, companies focused on reducing average handle time (AHT), since the longer agents spend on each call, the more it costs a company. But here’s the problem: There’s a gap between knowing that longer calls cost more money and reducing those call times effectively without sacrificing customer satisfaction or call resolution.
After all, AHT alone doesn’t show companies if they’re meeting customer needs. Sometimes, a longer call time can be a good thing, like when agents work with a customer to solve a complex problem.
We’ve developed a new feature, Root Cause AI, that implements machine learning to identify how specific call events or agent behaviors influence how long a call lasts (AHT). This information can show you preventable, undesirable behaviors. Then you can make changes to your coaching programs and identify what issues can be self-serviced on other channels.
How we analyze AHT
As we were beta testing our new feature, we worked with a few of our customers to look into their contact center metrics and see what drove their call times – and if anything could be done about it. For one customer, we analyzed a sample of more than 13,000 calls over a two-week period. Their average handle time was just over 9 minutes, with an estimated cost per minute of $1.
Our machine learning analyzed these variables and showed us, and our users, what they needed to focus on if they wanted to reduce AHT and save operational costs. To do this, we taught our AI to ask three questions.
Question 1: How much time do specific events add?
We first looked at different types of incidents and how much time, on average, each added per call. For example, every time an agent verified a customer’s account, it added about 30 seconds to a call.
Question 2: How often do these incidents occur?
Once we had a breakdown of the different incidents contributed to AHT, we looked at how common they were. In one analysis, we found that when an agent misdiagnoses an issue, it can add more than a minute to a call. But in this case, those incidents were so infrequent that it didn’t make a noticeable impact on overall costs.
In contrast, agent confusion added nearly a minute to every call, and occurred more than 5,000 times in the 13,000 calls.
Question 3: What’s the cost impact of these incidents?
When we combine the seconds per incident with the volume of the incidents, we get a picture of the total impact different behaviors have. This allows us to quantify potential cost savings for contact centers. Once we factored in the cost per minute, we could see what different behaviors end up costing our users.
In the case of agent confusion, we found it cost our customer more than $100,000 a year. This valuable piece of information can help them know how to coach their agents to reduce call center costs.
Good behaviors cost money, too. For example, when agents ask probing questions it can add 40 seconds to a call. That amounts to nearly $200,000 a year – but it also leads to agents resolving customers’ issues. In that case, actions that lead to call resolutions should be seen as money well spent.
We also learned that scripted events in calls resulted in significantly shorter calls. Agents who know what to ask, and when to follow prescribed scripts, tend to be better at handling customer issues.
Companies often say they’re more interested in resolving issues than reducing handle time. The good news is, they are not mutually exclusive.
The full potential of AI analysis for other call center performance issues
Although the first iteration of this technology focused on AHT, we know its potential use is much bigger.
Using the same type of AI analysis, we can pinpoint the root causes of a myriad of contact center issues, such as contact resolutions, hold time, abandoned calls, and customer satisfaction. We can also use the technology to analyze customer journey issues, such as repeat contacts and customer friction, or operational issues, like dropped calls or long holds.
We could also use Root Cause AI to look at what behaviors cause difficult calls, instead of merely measuring how many difficult calls a company has. This can guide companies not just toward reducing costs, but improving their service level.
It can also examine different layers of the same problem. For example, perhaps Root Cause AI revealed that putting customers on hold created long calls. That’s no big surprise.To fix it, you know you need to to reduce those average hold times to capture cost savings.
This solution allows you to dig deeper to see what issues caused holds, perhaps revealing a specific account issue that customer service agents had to complete while customers wait on hold. You could then modify the internal process to make that easier for agents, reducing the time customers wait on hold.
In the earlier example, you could analyze what specific issues caused agent confusion. Armed with that information, team leaders can coach agents on common areas of confusion so they know how to handle the issue in the future.
Using AI analysis to predict sales
Root Cause AI could help more than just customer service.
In another customer beta test, we’re looking at what behaviors led to closing a sale. With this information, you could pinpoint specific actions that directly correlated with a successful close. That has the potential to turn your average-performing sales representatives into sales leaders.
Interested in seeing what AI tools can do for your business?
Imagine all that you could uncover from the wealth of information inside your customer interactions. With Tethr, you can analyze data from your customer phone calls, chats, or cases and extract valuable insights – which end up adding money back to your business.
Interested in seeing all the information we extract from conversations? Sign up for a trial and get a self-guided tour of our platform, which allows you to see sample data and upload your own calls.