facebook-pixel
  • Artificial Intelligence
  • Daniel Newman
  • AUG 14, 2019

Does Your Company Need That Chatbot?

That’s the million-dollar question in digital transformation today, as companies work to one-up each other on customer experience and 24/7 availability. While it might feel necessary to implement virtual agents for your business, however, it may not be right for your company. Does your company need a chatbot? Even more importantly, can it safely handle implementing one? Consider the following before you decide. 

uncaptioned

GETTY

How busy are your customer service agents?

We’d all like to use chatbots to save time and money. Avis has been able to automate nearly 70% of customer inquiries just by using virtual agents, for instance. Talk about time—and I’m sure headache—savings! However, the way AI chatbots and virtual agents work is through learning. That means they need to process not just hundreds but thousands of inquiries to optimize their ability to effectively handle customer questions. Dish Network, for instance, uses its chatbot DiVA to answer about 4 million calls per year—about 40% of its volume. If you aren’t receiving that type of flow, it might not be worth the effort of “training” a bot to help your business. Your customer service people will likely be intervening on the bots’ behalf just as much as if they’d handled the calls themselves.

You might not need it now, but if you’re continuing to grow and scale your digital transformation it could be useful in the future. But as I always say, don’t buy tech for tech’s sake and make sure the experience is better because of it.

Can you wire your bot into your customer systems?

Chatbots work the best when they can scan a request for intent by looking back on a customer’s history over time. That means they can’t operate independently; they need to be wired into their order history, past inquiries, payment history, etc. If your systems aren’t set up to work interdependently, it may not be worth implementing chatbots after all; the most your bot will be able to do is hand off the call to someone who has higher levels of access.

Now if you’re still using a legacy customer service system then be mindful of the complexities of integration (but move fast as possible), but if your system is only a few years old, is integrated seamlessly in your CRM and works well for your employees, don’t fix what isn’t broken.

As we’ve discussed before, it’s always a risk to network and share customer information. In using chatbots, you’re relinquishing control of a wide range of customer data to AI. How will you keep it safe? What outer boundaries have you established to ensure it won’t be shared beyond your chat or phone call? What pieces of the chat will you keep, and which will you ditch in the name of data security? Companies like Bank of America have chosen to service customers on their own site or app, where they have complete control of the chatbot experience. Others, like Butterball, have decided to work with established platforms like Alexa, to field customer questions through Alexa’s “skill” platform. Obviously, questions about one’s personal finances need higher security than those about one’s Thanksgiving turkey recipe. Only you know the security requirements that will best fit your customers and industry.

Do you have time to wait?

Just by the nature of AI taking time to learn what it needs to know, you need to understand that your chatbot won’t work perfectly straight out of the gate. As noted above, it may need to field thousands of inquiries to understand a customer’s intent or desire perfectly. Only you can answer the question: Is it really worth the effort? If you’re a small company fielding a few inquiries a day, you’re likely going to offer a more efficient customer service experience fielding inquiries the good old-fashioned way. If you’re a larger corporation, you may find the awkward interim period of AI training to be worth it.

Do you have a specific project to pilot?

Companies that have successfully used chatbots have found it useful to start with piloting a specific product or issue. For instance, if you’re upgrading systems and know most customers will experience a brief outage—and you have clear instructions for how to guide them to fixing it—you may begin implementing chat bot during the release of that upgrade. That way you’ll have greater “manpower” in dealing with the deluge of calls and a relatively easy answer in place that your bots will already know. From there, you can begin to inform your bot about other elements of the company and build its bandwidth. Trying to launch a bot with full company knowledge from the get-go would be setting your company up for failure—and your customers up for massive irritation.

Does your company need a chatbot? Really?

Do you have money to invest in good AI? There is nothing that can annoy customers more than dealing with a bot who can do literally nothing besides direct them to a person who can. We’ve all had experiences of dealing with chatbots that were efficient and smart—and those that made us want to smash our computer screen. Does your company have money to invest in quality AI? If not, it might be worth skipping the chatbot option.

As chatbots continue to develop more advanced sensitivity skills, they’ll likely be able even more complex inquiries from your customers. Still, every chatbot needs to start somewhere. And only you can decide whether training a chatbot will truly enhance not just your bottom line, but the overall experience you offer to your customers.

This article has appeared in Forbes.

The Harvard Innovation Lab

Made in Boston @

The Harvard Innovation Lab

350

Matching Providers

Matching providers 2
comments powered by Disqus.