After completing step 1, you now understand your customers a lot better. So, what’s next? It’s time to help them solve their problems! Artificial intelligence for customer care is an ongoing journey and it has to start somewhere. Like everything in a business, efforts have to match the reward. Thankfully, there are plenty of low-effort yet high-reward initiatives that you can take. When we talk about an AI solution, it is important to remember that it does not mean that a computer system will take over all decisions. In some cases, the AI part helps to identify the fix, which may end up being a non-tech solution! In this blog, we’ve listed a few common quick wins you can make while keeping your investment low.
- Making information available
One of the best ways to reduce customer care costs is to allow customers to help themselves. This often means that your customers should be able to find information related to their issues, typically on your website.
Available means two things. The obvious one is that the information should exist. There should be an FAQ, a page on the website dedicated to an issue, an info button, etc. The second one is less obvious, but even more important. The information should be findable with reasonable efforts. If it takes two clicks to find the phone number to your call center, but five clicks to get to the page solving your customer’s problem, they will likely opt for the first option.
A good exercise is to go through the list of common problems, attempt to find a solution on the website, and record how obvious it is. You might realize that some solutions simply don’t exist or are very difficult to find. Both issues can easily be solved by restructuring the pages, or even renaming some of the content so that it becomes easier to search.
2. E-mail handling
For many companies, e-mails represent a huge workload, and often, the root of the problem is to get the e-mail to the right person. A great advantage of e-mail routing is that it is one of the rare machine learning problems whose training data generates itself. Indeed, for each e-mail, the goal is to assign it to a target mailbox, and if this mailbox is the wrong one, the e-mail will be forwarded to the correct one in most cases. Since you can track an e-mail’s final destination, it is also possible to retrain the routing algorithm without creating a separate annotation tool.
Another quick win related to e-mails is to set up automated replies. Especially during off-hours, sending automated responses, with at least a list of relevant FAQ items, may solve problems without requiring human involvement. It also makes your customers aware of the resources at their disposal.
3. Creating contact forms
The main big challenge with customer care and machine learning is to acquire quality training data. If you want to build a chatbot, you need to create lots of data to make sure the chatbot understands human input. If you’re going to develop a system that extracts topics out of audio conversations, someone typically has to listen to hours of calls to label them. All of these efforts are time-consuming and expensive. Thankfully, there are clever tricks to reduce this workload tremendously.
Instead of having customers e-mailing you directly, you can create a contact form that asks them to pick which topic fits their issue best. For some topics, you may be able to ask them relevant information right away, which may avoid additional back and forth communcation. In some cases, it reduces the need for a customer care agent’s involvement altogether. Knowing the topic and key pieces of information right away helps direct the conversation to the right person. This solution is also directly applicable to call centers. Customers can fill in a form before they call, and then, when the agent picks up, a lot of the work might have been done already.
Big or small, tech-savvy or not, your business can do a lot already to improve their customer care. Clever UI/UX tricks are valid solutions and may reduce the need for more expensive solutions down the line. It is also important to think ahead, especially when it comes to gathering and annotating data. You may not have plans to build a digital assistant yet, but if you do in the future, you will be extremely happy that you did all this work upfront. Also, you may want to check out this blog where we highlight other AI quick wins along with some real-life examples that might be relevant to your business!
Happen to miss the other relevant NLP blogs? Don’t worry! You can check them out here:
This blog is written by Valentin Calomme – AI Engineer & Business Line Manager at Mediaan