Whether it’s E.L.F., Satisfi, or the Staples Easy Button, more companies are leveraging the power of Artificial Intelligence (AI) than ever before. Domino’s chatbot, Dom, even gives hungry customers the ability to order a pizza straight through Facebook Messenger.
These easy-to-use customer services come in many shapes and sizes, but they all come as a result of the same massive breakthroughs in deep learning (or what many scientists call “deep neural networks”). In the last five years, deep learning has been successful in many applications where it wasn’t before, and we can expect to see significant advancements moving forward.
The implications for the customer support industry are vast. Companies like PullString now offer bot-authoring tools and analytics for a variety of industries, including customer service, while Zendesk just rolled out their Automatic Answers platform.
These tools are becoming available for everyone from budding startups to worldwide enterprises. Now let’s talk about what this actually means for your businesses:
How Does This Benefit Your Customers?
There’s no doubt that machine learning will pave the way to quicker response times for common customer inquiries. In a world where 50.6% of customers expect 24/7 support, AI could be the answer we’ve all been looking for.
Bots can make a great first line of defense for your support staff. By automating answers for your simplest and most routine questions, trained personnel can focus their efforts on customers with more complex or difficult issues. Bots can also help efficiently gather information about a customer’s inquiry before escalating them to an expert on your team, providing an overall smoother troubleshooting experience.
Another exciting feature of services like Zendesk’s Automatic Answers is that they’re not just reading and responding to one individual customer’s question – every interaction is teaching the program how it can respond more effectively in the future.
Laurel Hart, an Artificial Intelligence Engineer at PullString, stated in a panel discussion last month that their bots keep track of every previous interaction they’ve had with any given customer: “It will remember your conversation state and the whole pileup trail that you’ve been through, and you can store specific pieces of information that you can then recall later.” While your human support team can look back at customer histories, it’s important to consider the time it takes to read through lengthy back and forths. AI can do this instantly.
Finally, in our ever-expanding global economy, it becomes increasingly important to find more effective strategies for supporting customers in a variety of languages. Wit.ai provides developers with the tools to create text or voice-based bots that can interact in 50 different languages, from Danish to Swahili. At a time when 86% of contact centers report having non-English speaking customers, the ability to automate multi-language responses could make a big impact in improving customer experience.
All that said, it’s important to keep in mind that transparency is also vitally important to your customer’s satisfaction. Always make sure customers know when they are talking to a bot — nobody likes to feel like they are being deceived.
So Can AI Replace Human Employees?
The short answer: no. Implementing AI technology may lower your customer-service costs and allow your agents to focus on more sophisticated, higher-level areas of support, but machine learning is still a long way away from replacing humans entirely.
While autonomous robots and self-driving cars are increasingly outperforming their man-powered equivalents, language tasks are significantly more difficult.
This is particularly true for companies operating within niche markets, or agents supporting a product with numerous complex features. There are several reasons for this, both technological and with regards to public opinion.
First, let’s talk about the technology.
The most important consideration in customer support AI systems is how general they’re allowed to be with regard to their area of expertise and how they come up with answers. Researchers classify the manner in which AIs generate answers into two camps: retrieval-based AIs and generative-based AIs.
Retrieval-based AIs can identify the subject of a customer’s question and learns how to retrieve canned responses from a database.
Generative-based AIs are capable of composing their own answers, learning how to compose better answers extemporaneously based on data it collects from each interaction.
Retrieval-based models are perfect for fielding a lot of the most repetitive questions your team encounters. However, as soon as a customer asks an unexpected question, or responds in an unexpected way, you’ll be happy to have a human support team member back in the driver’s seat.
Generative-based models are more adaptive, but since you’re allowing the computer to “think for itself” in many ways, they are more likely to make grammatical errors or produce frustrating generic responses (like “I don’t know what you mean”). These may be perfectly logical statements, but they aren’t actually helpful to your customer. This is especially likely to happen in instances where a customer is less articulate or not their question is not quite clear. Again, these are times when a human support agent will be crucial.
Now, on to Public Opinion.
It won’t come as a surprise to anyone working in the customer support world that customers aren’t always happy to be receiving an automated response. One poll suggests customers in higher income brackets are more likely than others to find bots “invasive” in support interactions.
It’s important to consider the types of inquiries that may be too sensitive for an automated system to handle – not because the question is complex, but because the implications are more delicate. For example: How will customers feel if it’s a computer and not a human telling them they can’t have a refund?
Situations like this will be important ones for us to create policy around. Ultimately, as long as a positive customer experience is still firmly in our minds as a primary goal, the gains will be tremendous. By combining the speed and efficiency of AI with the thoughtfulness and adaptability of human support, your customers will have an improved experience when they need support.
When is Your Help Desk Ready for AI?
Let’s face it: AI stands to save your company serious time and money — some level of AI support is bound to become the norm in the future.
Many companies that start out staffing an entirely human help desk early on expect to automate it once they’ve collected enough data. However, for more complex products, there’s an advantage to considering AI as early as possible.
Your AI support improves when it has access to as much data as possible. Even if most of your questions can’t be answered by bots right away, a bot can learn from every interaction it sees your human support staff complete. Over time, this will organically help expand the bot’s knowledge base.
Not only that, but even if you only have a few employees, AI can help you gain useful insights about trends and customer pain points that might otherwise be too complicated or time-consuming to unearth.
The good news is the leaders in AI tech have provided plentiful open source resources you can use to start training your team now (see: Google’s TensorFlow, Microsoft’s Cognitive Toolkit and Amazon AI). Looking forward, it will be useful to consider specific use cases in which AI could be of the most value to your business.
Be prepared to be patient. Today’s machine learning software requires a huge amount of data. But with the help of a diligent, flexible, innovative support staff, it could make a huge difference to the future of your business.
Also published on Medium.