Questions about AI often lead to answers that sound like they’re coming from a bot. Case in point: I reached out to an AI expert when researching this article who, quite astutely, has a Messenger bot that answers PR requests. The idea seems novel in practice (why shouldn’t an AI aficionado use a chatbot to answer questions?) but the execution left a lot to be desired.
Suffice it to say, all I got were a bunch of links to chatbot articles that this expert had written, none of which were actual answers to my detailed questions.
This frustrating encounter represents one of the biggest hurdles with using bots for customer service: artificial intelligence with too much emphasis on the ‘artificial’ and not enough focus on the ‘intelligence’. Research has shown that customers prefer talking to a human over any other form of customer service help, yet chatbots are far from replicating human compassion or understanding (or even basic problem solving, in some cases). In fact, a recent study from Narvar shows that only 2 percent of consumers surveyed actually prefer speaking to a bot.
Despite all the hype about chatbots and their potential to revolutionize nearly every aspect of business, it still feels like they’re a ways away from being autonomous in the ever-so-demanding world of customer service. Or are they?
Investigating the status of bots for customer service and whether or not they can be useful for small businesses and customers, I spoke to a few AI experts to see what the present (and future) of bots for customer service holds.
Command-based or conversational?
Before deciding on whether or not chatbots are useful for customer service, we need to outline the different types of chatbots that are available. Roberto Gonzalez, CTO at digital product agency Aerolab, explained the reason why command-based bots might be better for customer service than conversational bots.
“There are two main ways to go about building a chatbot from a product design point of view,” he said.
“Command-based bots are the simpler products and the ones we usually recommend for building your first chatbot. They are centered on a few simple buttons with common questions, which means they are very good for general support, since users can just click on a few common questions and get fast, clear responses,” said Gonzalez.
Conversational bots, on the other hand, are a bit more complex. They rely on more advanced AI development and resources, making them not only out of reach for small businesses, but, according to Gonzalez, less useful for customer service in their current state.
“These are the more ambitious versions of chatbots, and the most failure prone of them all. This encompasses AI assistants like Siri, Alexa or Google Now but, since they are based heavily on Machine Learning and Natural Language Processing, the training and development processes are insanely expensive,” said Gonzalez.
Clear distinctions in functionality mean that the pros and cons of each need to be weighed before implementing one or the other.
Command-based bots are easier to set-up, but are limited in terms of the questions they can answer, since anything asked outside of their realm of pre-defined questions will respond with the equivalent of a “does not compute” answer.
Conversational bots, while more expensive and complicated to implement, have the potential to answer more complex questions. They are, however, only as good as the AI engines behind them. They can just as easily answer with gibberish if a question veers off the beaten path.
This balance needs to be weighed in order to make bots that work well for both the company and the customer.
Using bots for customer service
Customer service covers a broad range of channels and technologies, and bots don’t necessarily work for all of them. Olivier Jouve, Executive VP of customer experience software Genesys PureCloud, noted which customer service channels have been most successful for adopting bots: self-service and chat.
“Chatbots on websites are a really interesting strategy. While searching, browsing and using navigation menus have long been the primary methods for consumers to find content and/or accomplish a task on a brand website, chatbots are becoming an extension of the self-service interface for consumers to engage with businesses in a more natural and conversational manner.”
This makes a ton of sense for customer service; the templated question and answer format of self-service sections on websites is exactly where command-based bots shine.
Another useful implementation of bots for customer service is with popular messaging applications used to chat with customers.
According to Jouve, “Facebook Messenger, WeChat, and LINE, for example, allow consumers to engage with businesses in a conversational-by-nature touchpoint. For consumers, chatbots can offer self-service capabilities they would typically find on the business website and mobile app, but right within their preferred messaging application they extensively use already with their friends, colleagues and family. For businesses, serving consumers through messaging applications without the help of chatbots (and therefore exclusively with human advisors) is a frequent barrier to offering customer service onto these globally popular touchpoint. Chatbots are therefore a win-win technology for both consumers and businesses.”
If, however, conversational bots don’t stack up, they can cause frustration for consumers. A study from Narvar shows that 35 percent of customers surveyed didn’t like communicating with bots because they needed to repeat things, wait for a response, and didn’t get answers to their questions.
“Most companies try to create AI-based bots without understanding the development costs and limitations of the technology, so they end up building products that essentially repeat ‘Sorry, I didn’t get that’ ad-eternum or start responding nonsense,” said Gonzalez.
“This is especially terrible in customer support, since most developers pretend the automated responses come from a human. Once the bot sends a nonsensical response, your customers will become offended, as bad responses are interpreted as if the support team doesn’t care about them or their problem. This is horrifyingly bad for customer loyalty,” Gonzalez said.
Like my experience trying to contact the aforementioned AI expert via a Messenger bot, unless you’re able to add in human intervention when the bot has reached its threshold of answerable questions, you’ll end up with frustrated customers.
This is where a blended strategy becomes useful.
From “robot” to “Robert”
Because AI à la Ava in the movie Ex Machina aren’t quite at such an advanced level of intelligence, a combination of bots and human intervention is one of the best ways that a small business can get started with chatbots.
Jouve said that a blended strategy using a combination of chatbots and human advisors can be the most useful for consumers and for businesses. In this scenario, chatbots start the conversation, and a human agent picks it up.
Jouve outlined the different ways in which a chatbot can be trained in order to allow this type of handoff, including:
- Giving automated answers when it can answer the question, and automatically passing it on to the appropriate agent when it can not.
- Qualifying customer status before passing it on to a human advisor.
- Determining which channel would be most useful to hand-off to a human agent, whether that be live chat, phone, video, or even in-person appointments.
- Collecting customer info that a human agent would be able to use to help the customer (ie. get some background info on the customer or issue).
- Being the first touchpoint for a customer when accessing a website.
These types of handoffs are useful because they provide the immediacy that customers want when reaching out, while being able to direct the questions to an appropriate agent in waiting. This can save time answering qualifying questions and ensure that once contact with a human is made, the root of the request is being dealt with immediately.
Small business bots
Small businesses may be weary about implementing chatbots in their customer service strategy right away. I get it– it can be scary taking such a bold technological leap. But it doesn’t have to be as complicated as that.
Jouve said, “Today’s reality is that the majority of chatbots are Q&A interfaces that recognize the consumer’s intent, but are merely providing a general content, FAQ-style answer. Consumers would also demand personal information (e.g. “what’s my credit card balance”) and/or completing a transaction (e.g. “please activate international data roaming”). These self-service capabilities are already available through business websites, mobile apps and self-service IVRs, but are often not yet orchestrated by chatbots.”
Jouve said that this isn’t a limitation of chatbots, but rather, a lack of implementation.
“The great news is that businesses can have a phased approach to deploying bots. Artificial intelligence doesn’t have to be complex or costly. Businesses can often simply leverage the existing customer engagement platform they already have to connect with AI and bot services. For those just getting started, the focus should be on specific, narrow goals with plans to optimize,” said Jouve.
Dariusz Zabrzenski, Head of Research and Development at Live Chat, also believes that small businesses are ready to capitalize on chatbots. “The time for small and mid-sized companies to jump into the world of chatbots is now,” Zabrzenski said. “Currently, bots are completely ready to handle the most common and repeated customers inquiries which don’t require any human reaction. In this way, it is possible to run complex, 24/7 customer service even in companies with limited resources.”
Gonzalez agrees: “You can start trying these out with tools like Chatfuel, which is a great alternative for small business owners, and can even do things like facilitate bookings or answer common questions, which are the main use cases for chatbots. The Facebook Messenger platform is used by more than a billion users and provides excellent support for products in this category.”
The important thing to remember with any of these strategies, is that despite a bot being on one end, there’s still a person on the other side of the conversation.
“You just need to understand the limitations and strengths of the medium and always think about how the user will interact with the product,” Gonzalez said.
The bot[tom] line
Merging existing technologies and incorporating bots, as opposed to creating them from scratch, is a small business’ gateway into using bots for customer service. Despite some of their apparent shortcomings, they can be useful, and they’re only set to become more useful as the technology continues to develop.
“In terms of maturity, the advances in natural language understanding and machine learning are astonishing, compared to chatbots a decade ago which were merely a keyword search interface. However, we are at the beginning of the journey for chatbots implementations to become a real extension of a businesses’ self-service interface,” Jouve said.
Far from being robotic, customer service bots will only continue to develop. In the meantime, using them where you can will help your small business get a technological leg-up on the competition and prepare for fully-fledged, automatic, bot-friendly customer service departments in the not too distant future.
Read more about bots in business:
- Are customer service chatbots full of chit?
- Artificial intelligence, chatbots, and the future of marketing
- Bot or not? How to use chatbots in your small business
Check out some tools for building your own chatbot: