Too often, it goes like this: a small but mighty support team starts out as three employees dedicated to helping early customers. They carefully read and respond to each request with grace and enthusiasm. All is well.
Then business grows. Ticket volume increases, expectations rise – yet there's still just those three team members managing the new flood of requests.
So what happens? Systems clog, employee stress levels skyrocket, and customer wait times eventually dribble out to infinity.
Or perhaps not quite infinity, but you get the point. We all know the drill: teams can't always scale to keep up with business.
In this situation, teams often begin to spend significant time triaging – manually going through each ticket simply to determine what it concerns and who should handle it. We generally triage based on things like issue area, urgency, complexity, and what kinds of resources and processes we'll need to solve the problem.
Which is great. But again, only until we scale.
When we have ten tickets to deal within a day, going through them manually is great. But when that customer request volume shoots to 1,000? Chances are we don’t have 100 employees ready and waiting to triage those tickets.
And when we care about metrics like first response times, first contact resolution rates, and average time to resolution, then we suddenly care a lot more about time we squander simply triaging instead of actually resolving customer issues.
Happily, there’s a better way.
By finding a way to simply improve routing time, we can elicit faster responses, solve customers problems more quickly, and ultimately increase customer satisfaction.
Key to improving those metrics is just getting the right ticket to the right person as fast as possible.
So how do we do that? Enter the beauty of machine learning. Using machine learning, you can teach and train an algorithm to automatically sort large volumes of customer support tickets into categories. You can rank them by topic or urgency. You can make sure they get routed to the right team member. You can take time spent on the back end triaging tickets and reallocate it to the front end doing what you do best: supporting and satisfying your customers.
And you can do all of this in mere minutes. Faster responses. Faster resolutions. Less bombarded employees. Happier people all around.
At MonkeyLearn, we've paired up with Zendesk to offer an integration that will help you do just that: automatically route customer support tickets to the most appropriate person.
Since this technology is integrated through and operating within Zendesk, the algorithm is able to see all of the already existing tickets, what content they contain, and how they've been categorized by your team in the past. It takes all that information and learns to develop similar associations based on how your team has already been working.
For example, if it sees that a subject line of “lost password” has tended to be categorized as “login issues,” then it will replicate that pattern in the future. From now on, all tickets containing text about lost passwords or similar can be automatically routed to the person who handles login issues.
It's a simple step, really – and that's partly why we're baffled that machine learning in this area is not already more widespread. Sorting customer support tickets by content is one of those mundane, fossilized tasks that no person should still be needing to do in the 21st century. Let's let a machine automate that in minutes, and let's leave the stimulating, thoughtful work to humans – which in this case is helping and developing relationships with other humans.
The goal here is to automate processes, not relationships. By using MonkeyLearn’s integration with Zendesk behind the scenes to better enable your employees, you can free your team to focus on what people do best: empathize, troubleshoot, create, reason, and build positive relationships. Not to mention saving time, energy, and headaches for all.
Here are four steps to start using machine learning within Zendesk to make your customer support team’s lives a little easier:
Install the MonkeyLearn integration in the app marketplace from your Zendesk dashboard.
Locate “Choose a ticket field to classify” from the side menu: This is where you tell the algorithm how you want it to categorize each customer ticket. Normally, new messages come in as uncategorized tickets – you see the subject, the message, and then an unassigned “about” field. That's exactly what this integration will change. Select your “About” field (the custom field you use to categorize your tickets according to their topic):
Set the automation level: This allows you to determine how confident you want the machine to be in order to automatically classify a ticket: low, mid, or high. The higher the value, the more tickets will be automatically classified.
Add your API token: API (application programming interface) tokens are used as a form of ID that helps identify who is requesting information from the system:
Set up your Zendesk trigger: To complete the process, set a trigger on Zendesk that will take the information provided by the MonkeyLearn integration – the automatic analysis of and classification of each tickets’ content – and use it to route each ticket to the specific person best equipped to address it.
At this point, the algorithm downloads all the existing tickets already in a machine learning model on MonkeyLearn. It looks at the messages, subjects, and any categories you may have already applied via Zendesk. It learns from those existing associations and uses them to train the model. Depending on the volume of your data, it will take several minutes to download those texts, learn from their associations, and train your new model.
As the system is doing this, you can go to MonkeyLearn and see the model being trained. After the training is done, you can see new tickets on Zendesk now being categorized automatically with machine learning and being routed to the most appropriate agent.
From there? Your team can dive straight into the good stuff: responding to and resolving issues, ensuring customer success, and keeping wait times well above that slow dribble out to infinity.
If that sounds like a win-win, send me an email to email@example.com. I’ll be more than happy to field any questions, try out the integration for yourselves, and help your team get started using machine learning to be one step closer to faster, better customer support.
April 25th, 2018