Embrace the Chatbot Revolution
Chatbots can add a dash of conversational artificial intelligence (AI) to your business across customer support, accounting and expense management, e-commerce, and beyond.
By Rob Marvin
Chatbots are everywhere in 2017. You can find thousands of them in Facebook Messenger and WhatsApp, in Slack and Skype, and embedded as automated conversational experiences within countless other apps and services we use every day.
Chatbots are exactly what they sound like: They’re a form of conversational artificial intelligence (AI) in which a user interacts with a virtual agent through natural language messaging. These automated interactions can be used to surface contextually relevant information, help a user complete an online transaction, or serve as a helpdesk agent to resolve a customer’s issue without ever involving a human.
One important distinction to make is that chatbots are not full-blown virtual assistants, like Alexa, Cortana, and Siri. The distinction between a text or voice-based chat interface is less important than the scope of where digital assistants live and what they do for a user as opposed to the more narrow focus of a chatbot. Virtual assistants are omnipresent AI helpers embedded within our smartphones, Bluetooth speakers, operating systems, and other computing environments, offering predictive recommendations and performing a wide range of evolving functions. Chatbots, for the most part, live within a single application or messaging interface and can be programmed more simply with a selection of automated actions for specific business tasks.
For the average small to midsize business (SMB) or enterprise organization, the specialized nature of chatbots actually makes them more useful. Companies and software providers are already implementing chatbot experiences to revolutionize business processes including e-commerce, customer support, expense tracking, and more. In this article, we’ll lay out what chatbots can do in real-world business settings, how the technology is evolving, and explain why investing in chatbots isn’t always the right decision in place of a good ‘ol fashioned human employee.
Customer Support and Conversational Analytics
We can’t talk about every one of the thousands upon thousands of chatbots out there, so we focused on a few examples that show defined business use cases. Usually, one of the first chatbot applications people think of is customer support and virtual helpdesk agents.
There are dozens and dozens of customer-support bots available within Facebook Messenger, Slack, Skype, and other chat apps, including Kik and Telegram. Even popular helpdesk platforms such as Zendesk Support have launched customer support chatbots.
These types of chatbots help customers submit and resolve tickets without leaving their messaging app of choice and also serve as a natural language interface to a company’s knowledge base. Rather than going through a painfully antiquated automated call system just to talk to a human, a chatbot surfaces the information a customer would’ve had to search for otherwise without involving a human customer-support representative.
For Robert C. Johnson, CEO of business-to-business (B2B) customer support software company TeamSupport, chatbots are one layer in the push toward integrated omni-channel customer support. Johnson is skeptical about the potential of chatbots in more complex B2B interactions, but even for customer-facing support, he said automation is very tricky to get right.
“Five or seven years ago, the support space was an email ticket system, a knowledge base system, a chat system, and a phone system, and they didn’t really talk to one another,” said Johnson. “We’ve also had issues with automated technology before, whether it’s automated agents or phone trees. We know when we call and get an automated answer that we’ll have to go through seven layers of automation to talk to a person. It’s frustrating. Omni-channel customer support has the benefit of integrated search and workflows, things like a keyword in a ticket triggering a knowledge base article. Chatbot systems are layered on that as another channel.”
Beyond the basics of setting up a customer support bot triggering and answering tickets, you can configure these types of virtual agents to do just about anything. IBM offers a Watson-based platform leveraging the AI’s natural language processing (NLP) and cognitive computing to build customized business chatbots for customer interactions. Watson Virtual Agent gives businesses not only a customizable chatbots builder for customer-facing experiences, but deep analytics and an engagement metrics dashboard to measure the chatbot’s effectiveness.
“I think about cognitive computing as a set of intelligence capabilities that provide strength and leverage to our human mind,” said Rob High, IBM Fellow, Vice President and Chief Technology Officer, IBM Watson. High spoke at Mobile World Congress earlier this year about cognitive computing in chatbots and virtual agents, explaining why these types of AI interactions make sense in customer support and online merchant interactions.
“We focus on how to enable other people to create conversations in the context of these kinds of assistants and interfaces that offer multiple modes of interaction: voice, gestures, chat — as we go forward, what we’re seeing today is just the beginning of experiences that will be multi-modal and far-reaching,” said High.
“Much of what has driven the internet has been about merchants attempting to create relationships with consumers. Mobile and web applications give merchants the opportunity to control the experience they want to deliver to users, but consumers are already occupying these social spaces,” he said. “That’s where they are most of the day. Why shouldn’t the merchant go to those apps with [a chatbot] and give up that control for the sake of accessing consumers where they are?”
Watson works with Botkit to integrate virtual agents into Facebook Messenger, Slack, and other messaging platforms. But more than that cross-channel availability, the analytics you get from chatbots can tell you a lot about customer behavior and help refine the chatbots experience to make the investment in that automated interaction worthwhile. Arte Merritt, CEO of bot analytics platform Dashbot, spoke on the same MWC panel as High about how actionable bot analytics can increase user engagement and drive monetization.
“People think about bots for customer service, but they’re so much more,” said Merritt. “We see a lot of interest from brands across media, entertainment, travel, retail, to have an app-like bot experience inside something like Facebook Messenger,” said Merritt. “These interfaces are also asynchronous; they allow for unstructured data. You’re not just clicking buttons and links. In the case of Facebook Messenger, people can send images, video, or their own voice. We see all sorts of fascinating data come through that you can leverage. Users treat the bots as people.”
Accounting and Expense Assistants
Another business scenario where chatbots make a lot of sense is accounting and finance. Accounting and business management software provider Sage built a chatbot named Pegg last year, giving users and small business owners in particular a natural conservational chat interface for staying on top of invoices and expenses, tax deadlines, cash flow issues, and more. Pegg is built into the Sage platform, but also available as a bot for Facebook Messenger and Skype.
Kriti Sharma is the Vice President of Bots and Artificial Intelligence for Sage. She said the company has seen adoption in both the enterprise community and with small businesses as a way to automate boring tasks that entrepreneurs often put off.
“Some of these tasks businesses have to carry out are really boring, like filing tax returns and expense reports, and staying on top of purchase ledgers,” said Sharma. “Particularly with the rise of the gig economy and freelancers, business owners also may not have a dedicated accountant there to give them advice. A natural language interface is something that simplifies these connected scenarios at the right point in time, and that can be very useful.”
Pegg puts accounting jargon in natural language. Ask “who owes me money?” and the bot will quickly pull the data from Sage and tell you how much the invoice is, when it’s due, and the quickest way to contact the person. For expense tracking, the bot uses computer vision and optical character recognition (OCR) algorithms to analyze a photo of a receipt uploaded in the chat, and automatically catalog it in Sage.
Sharma is also the founder of Messaging Bots London, the largest network of bot developers in the city. She said developing good AI is less of a pure tech problem, and more of a combined technology and human problem. If you approach a chatbot purely as a technology solution, Sharma said you’re not going to get it right.
“We want to pay a lot of attention to the design and user experience [UX] of bots. For the small business world, accounting can be overwhelming,” said Sharma. “Chatbots in accounting is a use case that really makes sense. You need to do your accounts and capture your expenses, and doing it today is painful. If you have to create and send an invoice for a thousand dollars, a bot can do that. It’s an easy experience where the bot goes into the back-end system and you don’t have to worry about the process.”
Expensify is another example of a smartly integrated chatbot experience. Expensify’s Concierge assistant is built in throughout the company’s expense reporting and management platform. Expensify CEO and founder David Barrett views AI as a tool for solving specific problems. In Concierge’s case, this means a level of customer support and engagement for small businesses that wouldn’t be possible without automation.
“Think about why you would get a bookkeeper or accountant. You want someone to help with time-sensitive financial requirements unique to your business. The hard part is trying to configure a platform like Expensify to tackle this particular thing, or to answer a question like ‘I got this letter from the IRS, what do I do?’” said Barrett. “Normally you’d need a highly paid and trained accountant to do this, but a lot of small businesses don’t have the resources to afford that. Concierge brings that high value expertise into their hands.”
Expensify began by focusing on a narrow set of bookkeeping and accounting questions, and built out Concierge’s body of knowledge and automated responses from there. The bot reviews and categorizes uploaded receipts, manage expense reports automatically, and also serves as a natural language consultant for questions about specific expenses or price comparisons.
Barrett said Concierge is also evolving in a number of different ways. Beyond the ability to process a myriad of data types and systems — credit card import, email parsing, mileage tracking, mobile OCR, a company’s internal reporting structure, and more — the chatbot is also learning how to interact more naturally with users, and to leverage the data it collects to surface recommendations…in a way that’s not creepy.
“One area we’re working on a lot is getting Concierge to help you understand what’s happening on a daily basis in a way that’s not spam,” said Barrett. “We want to summarize data in a way that’s actionable, and for Expensify that’s about making the process of business travel more fun and social. We have data on how much you like Thai food over sandwiches, where you like to get it, all this info and preferences expressed through expense data. We also have calendar functionality, so think about automatically reconciling those preferences against meetings and attendees, and then suggesting a restaurant based on availability, location, and each user’s preferences.”
Automated E-Commerce Agents
For online businesses, the third big bucket for current chatbot technology is e-commerce automation. You can order an Uber or a Starbucks coffee without leaving Facebook Messenger, the application leading the e-commerce bot charge in the U.S.
Since Facebook opened up its platform to developers we’ve seen major brands launch bots for everything from travel booking (Hipmunk, Kayak, Travelocity) and food (Domino’s, Whole Foods) to major retailers including American Eagle, Kohl’s, Sephora, and Victoria’s Secret. Plenty of bots still employ live customer support and sales reps, but any bot in the directory carrying an “automated messaging” tag is a fully autonomous chatbot.
“Chatbots are everywhere,” said Beerud Sheth, founder & CEO of messaging bot creation platform Gupshup. “Inside a messaging app, everything is just a thread. If you’re chatting with an entity, it could be a human or just as easily be a program. Businesses can now develop a whole range of services that to the user seems like just another user you’re messaging.”
Sheth spoke on an MWC panel about bots and e-commerce along with representatives from PayPal, Google, and others. Within e-commerce, chatbots now have the ability complete an entire online shopping transaction within a conversational message experience.
Brick-and-mortar retailers have even turned to the technology to not only drive online and mobile conversions, but to help increase foot traffic as well. Macy’s used the Watson Virtual Agent platform to build and launch a bot called “Macy’s On Call,” which gives shoppers a customized chatbot to answer questions as you browse a particular store. In this instance, the chatbots are learning over time to provide better assistance as they analyze purchase pattern data.
What these automated experiences can do is fast-evolving, but the killer app for e-commerce chatbots is payments. In China, where the messaging app ecosystem isn’t as diverse and the majority of users are active on WeChat, the app’s built-in Weixin Pay feature has reached 600 million monthly users, according to parent company Tencent’s Q4 2016 results.
The payment functionality is integrated into the chatbots running on WeChat, so if a customer needs to make a payment the chatbots handles that transaction in a few clicks. Messaging app Kik has taken a cue from WeChat and has begun developing chat-based payment methods, and Facebook Messenger has native chatbot payments and a buy button feature currently in beta.
Harper Reed is the founder of mobile commerce startup Modest (acquired by PayPal) and now an entrepreneur-in-residence focused on Next Generation Commerce at PayPal. During the MWC panel Reed said he sees payments as one of the keys to chat-enabled commerce.
“What I’m excited about is that all of this conversation around chat and assistants is laying the groundwork for the future of commerce,” said Reed. “On the payments side, we want to create the open APIs for everyone to enable payments. I think there are a lot of hurdles between a normal consumer brand figuring out their mobile strategy — let alone their chat app strategy — and programming a Facebook Messenger chatbot. But we have a future ahead of us where chat is a big part of it, and that context of having an assistant in your pocket isn’t necessarily the only place it will be.”
The Human Touch
For all the ways an automated conversational interface can enhance your online business, chatbots also have limitations. Look no further than Microsoft’s Tay chatbot fiasco for the growing pains with natural language AI. In scenarios where human intuition and contextual understanding are still paramount, for instance in social media management, situational awareness and some cheeky sarcasm go a long way.
Social media is often a brand’s first line of defense when a customer has an issue or complaint with a product or service. Rather than an automated message that may miss context, a thoughtful response from a human showing understanding and empathy with the customer’s issue has far less chance of being dragged on Twitter and leading to a bigger public relations incident.
Scheduling is another example. Using a chatbot to schedule a meeting is a simple use case, but the nuances of that type of exchange highlight how much further development is still needed in natural language processing and deeper cognition.
When coordinating a call for this story with Expensify CEO David Barrett, we used a scheduling chatbot called x.ai. The bot ultimately scheduled the meeting, but it took multiple emails to get through to the bot that I was on Eastern Time and Barrett was on Pacific Time. This is the kind of NLP hiccup that’ll be worked out as the AI software learns and evolves, but it’s a good example of why chatbots aren’t always an effective substitute for a human.
“So much of this development is incredibly nuanced, especially around language which is a more fluid and flexible sort of medium,” said Barrett. “Imagine your friend texted you with two questions at once: where do you want to go to dinner and what time are you available? Which one of those do you choose to answer first? Or do you answer both at once? For a chatbot, maintaining that same conversational flow can be a really challenging thing.”
This all comes back to the human touch that a chatbot can’t quite replicate (yet). As useful as chatbots can be in all these business scenarios and more, we’re still dealing with an AI technology in its infancy. Let’s go back to customer support, one of the most prevalent business applications for chatbots. TeamSupport CEO Robert C. Johnson said that for higher complexity B2B interactions, chatbots may not result in the best customer experience.
“We’re a long way from the HAL 9000 and Jetsons-type stuff. AI that’s as intuitive as humans in being able to accurately connect the dots; a robot that can really think,” said Johnson.
“Accurate machine learning requires a huge number of data points and experiences to pull on. Without that volume, you really can’t do machine learning. In B2B interactions, you’re dealing with lower volume of interactions but higher complexity, which can lead to higher error rates,” Johnson continued. “Chatbots are good for B2C interactions where there’s a high volume and the value of each customer is not very high. If Nike loses a single customer and they’re not going to buy Nike shoes again. But if one of our B2B customers loses a million dollar contract, that error hurts the bottom line.”
It’s important to understand the current limits of conversational AI technology so your business can focus on investing in the chatbot experiences that are truly worthwhile. Building a chatbot for accounting and finance or as a productivity helper can improve organizational efficiency and streamline workflows. Rolling out a chatbot for e-commerce or customer-facing support can open up your business to native messaging channels and give users a fast, intuitive way to interact with your brand.
Chatbots are not a catch-all solution to every challenge facing online businesses. But deployed wisely, they can be a valuable tool to arm employees with contextual data and keep customers engaged within native app experiences. Artificial intelligence is advancing by leaps and bounds every year, changing the way we interact with technology. If a chatbot makes sense for your business, don’t let the opportunity pass you by.
Read more: “Amazon Offers ‘Lex’ Chatbot Platform to Devs”
Originally published at www.pcmag.com.