Build-a-Bot: How to Create More User-Friendly Chatbots
A competent chatbot is one of the most versatile tools a CMO can utilize in today’s martech landscape. With the ability to handle customer service experiences (both good and bad) and collect data from those experiences, chatbots can be a saving grace for marketers. Unfortunately, consumer backlash is almost inevitable when a new technology sees such a quick rise to ubiquity.
There has been a steady increase in disdain for chatbots as of late. A study shows that Facebook’s chatbots fail to understand nearly 70 percent of all user interactions. Forrester Research predicted in late 2018 that consumers will begin a full-scale rejection of the automated helpers beginning this year. In spite of these bleak outlooks, nearly 80 percent of CMOs are looking to either keep working with or adopt chatbots by the end of this year.
Here’s how you can ensure that your chatbots aren’t irritating your customers:
One of the most common complaints about chatbots is that they are too robotic. The Facebook chatbots we cited earlier are a prime example. How do marketers go about making their bots smarter?
Unless you’re extremely adept at programming, completely rewriting the AI of the bot on a whim is likely to be well outside of the realm of feasibility. So if you cannot drop the money for a sufficiently advanced bot, your best bet is to include more humans in your bot strategy. If users are reporting frustrations with your chatbot, you should be setting the logic to redirect them to a human member of your team. This will minimize the impact on customers and alert your team to the weaknesses of the bot until such a time it can be revised or replaced by an AI specialist. Escalation paths such as these are fairly simple and there’s no reason to eschew them.
On the notion of improving the logic of your bots, you need to ensure that there is more than one response for a given input. The idea of a chatbot implies a conversation. Conversation by its very nature can go in any of dozens of directions. When setting up the logic of your bot you need to account for any of several different responses the user might give. Only allowing for one answer is extremely restrictive and makes your “chat” bot feel more like a form bot. The short answer is that you should structure your bot to expect multiple types of responses when dealing with customers.
In that same vein, when your bot encounters unknown questions, and a team member is not available, have it link to something that may be generally helpful. Instead of “I’m sorry, I didn’t understand the question. Can you please rephrase?” your chatbot could say, “I can’t help with that, but here’s a link that may be able to.”
How else can you improve the chatbot experience? The same way you enhance most aspects of marketing; better data application.
Strong data usage is the key to success in business overall. By utilizing machine learning and neural networks your bot can leverage any and all consumer data you’ve collected over your various marketing efforts, including past chatbot interactions. Taking advantage of data indirectly makes your bot smarter over time by basing its interactions on past information.
If predictions and studies are to be believed, the public is having second thoughts about chatbots due to negative experiences they’ve had with the technology. Instead of packing up and going home, a forward-thinking marketing team will buckle down and improve their strategy the best they can to make the most of a still emergent tool in their kit.