What makes a great chatbot? Laser focus on customers

Fewer technologies have garnered more attention over the past year than chatbots, those virtual assistants that mimic human speech while facilitating tasks on behalf of humans, typically via a conversational messaging interface. At a time when software is driving unprecedented levels of automation, companies are using chatbots to help customers order anything from food to office supplies to additional computing capacity.

Chatbots are a big reason why corporate adoption of cognitive systems and AI will drive worldwide revenues from nearly $8 billion in 2016 to more than $47 billion in 2020, according to IDC. But what exactly makes a great chatbot? Perhaps more importantly given enterprises’ investments in such tools, what makes a bad one? What precautions should CIOs take in building them?

Perhaps no one is better equipped to answer these questions than Conversable CEO Ben Lamm, whose company has built chatbots for the likes of TGI Fridays, Whole Foods, Budweiser, and, most recently, Shake Shack.

Chatbots are all about the customers

A great chatbot is designed, implemented and deployed based on a deep understanding of a company’s customers, Lamm says. That means your chatbot initiative must have clear customer experience mandates, goals and key performance indicators that enable it to add demonstrable and measurable value for the user and the brands. “Whether that’s more convenient purchases, streamlined conversational support, or elevating your experience of a live event, the chatbot needs to take the customer experience to a new level of engagement,” Lamm says.

Lamm says the best chatbots can be continuously improved by line-of-business employees. Conversable, for example, has built an adaptive response system called Aqua that uses machine learning to identify broken queries, and allows employees to write and modify chatbot responses. “If your chatbot isn’t getting better, it’s definitely getting worse,” Lamm says.

Conversable has documented its chatbot-building process:

• Design the conversation: Identify the best use cases in customers’ existing operations and write out the conversation flows. Generally there are multiple stakeholders involved, varying depending on the customer, but it spans business and IT.
• Build the conversation: Think of this as the 1.0 of the final product. Test drive to make sure the experience as a whole is up to par. Revisions to the conversation flows and other facets of the experience happen at this stage.
• System integration: This is classic integration work. Conversable enables webhooks to make sure the data needed for each conversation flow is available. For example, if someone wants to know the price of a product, or how many calories are in a menu item, we need to pull on that data on-demand during the conversation.
• Learning: This is a human and machine collaboration to improve algorithms. Too many people assume you can just set an AI loose and it will figure it out.
• Expansion: Think of this as enabling more sophisticated conversations, expanding to other important areas where customers have questions or needs. People often leap from one topic to the next during a conversation, and there’s a relationship somewhere in that leap. When you identify a relationship between one conversation and another, easily linking the flows saves customers a lot of time, and ensures the chatbot can stick with the user as an organic conversation unfolds.
• Advanced AI: This is all about continuous improvement over time. You’re not done once you push the chatbot live, and our technology makes it easy to analyze what’s happening in bot-user interactions and identify areas to improve. With that data in hand, you can increase the sophistication of your chatbot using our AI.

“While every customer is a little different, we stick very closely to this process because it consistently gets us in production in weeks rather than the many months of development you can see elsewhere,” Lamm says.

Pitfalls that lead to bad chatbots

What makes a bad chatbot? One that tries to boil the ocean, Lamm says. “I’m amazed when companies launch their first chatbot and it claims to have functionality ranging from customer support across a giant product portfolio to e-commerce capabilities,” Lamm says. “Then it’s pushed live on six channels the company has little experience with, and the problems pile up at an exponential rate.”

Moreover, Lamm says the proliferation of general purpose bots littering industries does a major disservice to businesses and their customers. “Narrower applications of conversational AI ensure the experience is accurate, consistent and scalable,” he adds.

Rob Harles, Accenture Interactive’s head of social media and collaboration, says the consultancy does as much as he can to help clients pump the brakes and avoid chatbot pitfalls. Some clients want to rush out a chatbot because it’s cool and because a client wants to automate something to avoid dealing with customers. Others, echoing Lamm’s comment about “boiling the ocean,” try to apply automation technologies to as many things as possible right out of the gate. Both approaches are recipes for poor user experiences.

Harles asks clients to take a step back and understand the fundamental pain points they have and the discrete tasks they are trying to accomplish. Accenture will dive into a client’s customer journey to understand whether a task can in fact be better performed by algorithms, machine learning software or a good old-fashioned human.

A laser focus on the customer experience may offer enterprises the best recipe for a successful chatbot, the technology must be expertly built. That means adhering to design thinking principles for building the software with the human-centric approach. But the reality is few IT departments are equipped with enough design thinking specialists, let alone software engineers who have built chatbots before.

That’s why CIOs are turning to startups such as Conversable and consulting giants for help. What does a successful chatbot look like? Accenture’s interactive and mobility units in late 2016 rolled out a chatbot on behalf of Avianca, Colombia’s national airline. Built in six weeks on Facebook Messenger, “Carla” enables passengers to access information about check-ins and flight status and request seat changes from their smartphones, eliminating the need to make a separate phone call, download a new app or visit the company’s website. “[It] was designed to help passengers answer some of the more basic questions versus trying to do everything for them,” Harles says.

Harles says Accenture elected to build Carla on Amazon Web Services’ cloud computing software to enable Carla to scale easily during peak periods without affecting customer experience. Since launching in December 2016, Carla has racked up more than 31,200 unique users and fielded 1 million unique interactions.

Ultimately, a great chatbot requires thinking through the business use case and customer journey and a willingness to course correct when and where necessary. And while the machine learning and AI technologies must be functional, they are not the be all, end all of the user experience.

Hard work is, says Lamm. “I don’t care how ‘smart’ your AI is,” Lamm says. “AI is not magic and it doesn’t mean you can skip the usual rigor applied to core business functions and IT projects.”