HubSpot and Motion AI: Chatbot-Enabled CRM

9 – 518 – 067

R E V : O C T O B E R 4, 2018

 

 

 

 

J I LL A V ER Y

T H O M A S S T EENBU R G H

 

HubSpot and Motion AI: Chatbot-Enabled CRM

On September 20, 2017, HubSpot, an inbound marketing, sales, and customer relationship management (CRM) software provider, announced that it had acquired Motion AI, a software platform that enabled companies to easily build and deploy chatbots to interact with their customers. Chatbots were pieces of conversational software powered by artificial intelligence that had the capability to engage in one-to-one chat with customers on their preferred chat platform, such as Facebook Messenger or WeChat. Fueled by pre-programmed algorithms, natural language processing, and/or machine learning, chatbots conversed in ways that mimicked actual human communication.

Since its founding in November 2015, Motion AI had facilitated the building of 80,000 bots for brands including T-Mobile, Kia, Sony, and Wix, which were busy conversing with customers via 40 million total chat messages sent to date. The software was simple to use and enabled anyone, regardless of their level of technical knowledge, to build and manage a chatbot. The entire Motion AI team, including founder and CEO David Nelson joined HubSpot following the acquisition.

HubSpot saw great potential for chatbots for its business-to-business (B2B) customers, who could use them to automate many of their customer interactions that today were staffed by humans. Unlike other automated customer service solutions, such as interactive voice telephone response (IVR) systems that were almost universally disliked for their robotic nature, chatbots were getting closer to passing the Turing Test, convincingly simulating a human conversational partner so well that it was difficult to sense when one was chatting with a machine. Thus, chatbots had the potential to enable a company to nurture and manage one-to-one customized relationships with prospects and customers efficiently at scale by making artificial intelligence the new frontline face of their brands.

Chief strategy officer Brad Coffey and chief marketing officer Kipp Bodnar were responsible for working with Nelson to bring Motion AI’s technology into the HubSpot family of products. Before unleashing bot-building technology to its customers, HubSpot first needed to develop some best practices for the use of chatbots for CRM. Without proper instruction, Coffey worried that companies, in their rush to incorporate the newest marketing technology, would build bots that would do more harm to their brands than good. He prognosticated,

In the not-so-distant future, there’s a bleak, forsaken landscape. Civilization, absent. Communication channels, silent. All of the people have fled, terrorized by never-ending notifications and antagonizing messages. What could cause such a desolate scene? Bad

 

 

HBS Senior Lecturer Jill Avery and Professor Thomas Steenburgh (University of Virginia) prepared this case. It was reviewed and approved before publication by a company designate. Funding for the development of this case was provided by Harvard Business School and not by the company. Jill Avery has served as a paid consultant to HubSpot. HBS cases are developed solely as the basis for class discussion. Cases are not intended to serve as endorsements, sources of primary data, or illustrations of effective or ineffective management.

 

Copyright © 2018 President and Fellows of Harvard College. To order copies or request permission to reproduce materials, call 1-800-545-7685, write Harvard Business School Publishing, Boston, MA 02163, or go to www.hbsp.harvard.edu. This publication may not be digitized, photocopied, or otherwise reproduced, posted, or transmitted, without the permission of Harvard Business School.

[Type here]

 

 

[Type here]

 

518-067 HubSpot and Motion AI: Chatbot-Enabled CRM

 

 

 

bots. Okay, maybe that sounds a bit too much like the next superhero blockbuster. But it wouldn’t be the first time that brands abused a new technology until people were buried in spam up to their eyeballs.

He continued, “5% of companies worldwide say they are using chatbots regularly in 2016, 20% are piloting them, and 32% are planning to use or test them in 2017. As more and more brands join the race, we’re in desperate need of a framework around doing bots the right way — one that reflects the way consumers have changed.” The Motion AI technology would be incorporated into HubSpot’s product over the next few months, so the team had little time to make important decisions. First, they had to clearly assess the implications associated with the use of bots versus humans to create, nurture, and manage customer relationships and to determine whether and where in the marketing and sales funnel bots were appropriate for use during marketing and selling processes.

Second, they had to decide to what extent to anthropomorphize chatbots. How human-like should they be? Was a conversational user interface (UI) the desired solution or would a more functional UI produce more efficiency for customers? How much should the bot embody the brand’s personality or mimic the conversational style of an individual user? Should users know when they were interacting with a bot or could human-like bots create stronger relationships?

Historically, HubSpot had “practiced what it preached,” using its own products to build its business. Coffey and his team had to consider whether to use chatbots to nurture and service its own customer relationships. Currently, a team of chat representatives worked to engage, nurture, and prime prospects for HubSpot’s sales team. Could they and should they be replaced with chatbots? Was HubSpot ready for bots to become the face of its brand to prospective customers?

 

HubSpot’s Acquisition of Motion AI

HubSpot was founded in 2006 as an inbound marketing software-as-a-service (SaaS)a solutions provider, which helped primarily business-to-business (B2B) companies develop online content, attract visitors to it, convert them into sales leads, and acquire them as customers. HubSpot’s software helped companies develop, host, disseminate, and analyze digital content to execute inbound marketing programs, a collection of marketing strategies and techniques focused on pulling relevant prospects towards a business and its products during a time when they were actively searching for solutions.

In 2016, HubSpot’s revenues were up 49% to $271 million and were derived from 23,226 small and medium sized business (SMB) customers (see Exhibit 1 for the company’s financials). The company was excited to expand its value proposition and reposition itself as a robust, multi-product growth stack platform that helped SMBs combine all of their marketing, sales, and customer success software solutions into one convenient and easy to use platform. The growth stack platform was premised on delivering the following customer promise: “to fuel your growth and build deeper relationships, from first hello to happy customer and beyond” and included three product solutions:

· Marketing Hub: Grow your traffic, and convert more visitors into customers. Prices ranged from $50/month for a starter package to $2,400/month for an enterprise solution.

 

 

 

 

 

a HubSpot’s software was sold via a software-as-a-service (SaaS) model, where users paid a recurring monthly fee to access the software.

 

2

 

HubSpot and Motion AI: Chatbot-Enabled CRM 518-067

 

 

· Sales Hub: Drive productivity and close more deals with less work. Prices ranged from

$50/month for a starter package to $400/month for a higher end, professional solution.

· Customer Hub: Connect with your customers on their terms and help them succeed. At this point in the time, HubSpot was offering this product free with its other products.

 

At the heart of the new platform was the free CRM system that allowed companies to collect and analyze deeper insights on every contact, lead, and customer. A feature called “Conversations” empowered the CRM tool to collect customer conversations from Facebook Messenger, web chat, social media, email, and other messaging outlets into one cross-team inbox, to help marketing and sales teams manage, scale, and leverage one-to-one communications with their customers across all conversation channels. With its acquisition of Motion AI, HubSpot was hoping to further power efficient and effective customer conversations for its clients by bringing chatbots to the masses to better engage, convert, close and delight their customers at scale. Said Bodnar,

Today’s buyers expect that conversations with a business happen where they are. That might be the website, but it could also be social media, Skype, Slack, or any messaging app. They expect that conversations are portable. Regardless of where a conversation gets started, it should be able to be transferred to any other channel seamlessly. A thread kicked off on live chat should be able to be passed to Facebook Messenger or email without data loss or crossed wires. And, they expect that conversations have context. Context shouldn’t leave with the person who fielded the initial inquiry. All of a customer’s historical interactions and information should be attached to a common record which populates instantaneously. We need new technology paired with automation to live up to our buyers’ expectations and make these types of conversations a reality.

 

The Market for Chatbots

Chatbots were part of a wave of new artificial intelligence tools that were changing the way people interacted with technology. Digital virtual assistants housed in a smartphone, desktop, or laptop computer, such as Apple’s Siri and Microsoft’s Cortana, had paved the way for person-bot communication. More recently, Amazon’s Alexa, which could be awakened at any time by a voice prompt that spoke her name, provided ambient virtual assistance to consumers in their home.

Unlike these virtual assistants, chatbots were less sophisticated and tended to specialize in executing simple tasks rather than providing omnipresent and wide-ranging functionality (see Exhibit 2). While the most advanced virtual assistants were powered by artificial intelligence, which enabled them to understand complex requests, personalize responses, and improve interactions over time, most of today’s bots followed a simple set of rules programmed by a human coder who simulated a typical conversation and then programmed the bot to prompt a conversation by delivering a series of queries to a customer and to respond to them with canned responses triggered by simple if-then statements. Explained Derek Fridman, global experience director at Huge, a digital agency that helped its clients build chatbots, “The illusion that HAL [the computer from the movie 2001: A Space Odyssey] is out there, and the machine is alive is just that: an illusion. There’s machine learning taking place and algorithms making decisions, but in most cases, we’re scripting sequences.”1

According to McKinsey & Company2, technology companies spent between $20-30 billion on artificial intelligence in 2016. The market for chatbots was estimated to be $1 billion and was expected to nearly double by 2020 and triple within a decade. A recent Forrester study3 claimed that worldwide, 57% of firms were already using chatbots or planned to begin doing so shortly and 80% of businesses

 

3

 

[Type here]

 

518-067 HubSpot and Motion AI: Chatbot-Enabled CRM

 

 

 

wanted chatbots in place by 2020. In the U.S., 31% of marketers already used chatbots to communicate with consumers, with most (88%) of them deployed on Facebook Messenger. Following Facebook’s opening of its Messenger platform to chatbots in 2016, 100,000 were created within the first year.4

A recent study5 found that among companies using AI, the most common use cases were customer service (39%), marketing and sales (35%) and managing noncustomer external relations (28%). (See Exhibit 3 for examples). It was estimated that today 60% of customer service support issues could be resolved by chatbots and that number was expected to be 90% by 2020. Companies were finding that chatbots completed customer interactions at twice the speed and a fraction of the cost as human- provided telephone support. Oracle estimated that the cost of building a chatbot ran from $30,000 to

$250,000 depending upon its sophistication. While chatbots were reportedly saving businesses $20 million per year today, they were expected to help cut costs by more than $8 billion per year by 2022.

 

Chatbots and CRM

HubSpot’s CEO Brian Halligan was excited by the potential, “It’s impossible to ignore the impact of chat and messaging, not just on the way B2B companies operate, but on society as a whole. We’re in the midst of a massive shift as businesses embrace this new platform and consumers come to expect more immediate, always-on communication from brands.” Coffey echoed his enthusiasm,

There’s no downplaying what bots could do. For brands and consumers alike, we have a chance to facilitate a new type of communication and commerce. Research would be convenient, purchases streamlined, and service personalized. A conversational interface, powered by bots, can facilitate a response that’s as fast as talking to a human, with the depth of a full website, and a simple texting-like interface that everyone is already accustomed to using.

Bots provided instant responses to customers’ needs without the stress of waiting in a call queue or having to call in during business houses. Calling or emailing a company was quickly falling out of favor with consumers; Tech Crunch reported that 9 out of 10 consumers wanted to use messaging to interact with companies. Because chatbots were deployed within messaging app platforms, such as Facebook Messenger, WhatsApp, and WeChat, customers could speak with a company and accomplish their task without having to leave their preferred chat interface and without the hassle of downloading yet another app to their smartphones or visiting a company’s website. Five billion active users accessed messaging apps each month and their usage had surpassed social networks. According to Facebook, “convenience creates closeness…messaging makes commerce personal.”6 Research showed that 63% of people said chatting with a business made them feel more positive about the relationship, 55% were more likely to trust the business as a result of their chat conversations, and 53% were more likely to shop with a business they could contact via a messaging app.

HubSpot’s own research showed that consumers were showing greater interest in using messaging apps (see Exhibit 4). Explained PR manager Ellie Botelho, “Consumers want to be able to engage with a company when and where it’s personally convenient for them, meaning that businesses that are unable to respond quickly are leaving money on the table.” Added Coffey, “The way folks communicate externally is shifting towards messaging. Large companies manage these via live chats with an army of employees responding in real time. Few smaller companies can pull that off.”

 

 

 

 

 

4

 

HubSpot and Motion AI: Chatbot-Enabled CRM 518-067

 

 

Delivering a Human Touch via Artificial Intelligence

A Preference for Humans?

Today, consumers could order a Domino’s pizza, hail an Uber, book a flight via Travelocity, and reorder their favorite lipstick from Sephora via chatbots, all without leaving Facebook Messenger. The B2C world was rapidly adopting chatbots as an efficient way to execute simple transactions with customers without devoting human resources to them and without forcing consumers to visit their website or mobile app. Chatbots could be deployed to help with many different types of customer interactions that were common in B2B customer relationships, such as booking meetings, qualifying leads, diagnosing problems, and providing customer service to solve them, but it was unclear whether B2B customers would be open to robotic rather than human support, as B2B customers were often more demanding than B2C customers: “It’s no secret that today’s consumers expect personalized, relevant, contextual, and empathetic brand interactions throughout the entire buying process,” proclaimed digital analyst PJ Jakovljevic.7 B2B customer relationships were often more complex, more relational, and less transactional, so they often required the deft touch of a highly trained consultative salesperson.

“Chat is good when powered by humans. Chat is awesome when powered by AI,” claimed Christopher O’Donnell, HubSpot’s vice president of product. Bodnar, however, wasn’t so sure, claiming, “Automation is a funny thing. Too little is the enemy of efficiency. Too much kills engagement.” He continued,

Think about email. Automated email nurturing campaigns were the answer to individually following up with every single person who downloaded a piece of content from your website. In the name of efficiency, marketers queued up a series of emails via workflows to automatically deliver ever-more-helpful content and insights, gradually increasing the person’s trust in the company and stoking the flames of their buying intent. If at any time they had a question, they could respond to the email and get routed to a person who could help. But as the number of inbound leads skyrocketed, this system became untenable. The dreaded [email protected] address was the solution for scalability. Over time, this set the expectation with buyers that marketers didn’t want to have a conversation with them via email. Automation made us more efficient, but at the cost of relationships — ultimately defeating the purpose.

Then came live chat. Buyers were empowered to get answers to their questions in real time from a real person. Better yet, this interaction took place directly on the company’s website — where they were already doing their research. We started using website chat at HubSpot in 2013. Over the past four years, live chat has facilitated countless conversations between curious prospects and our business. But, just like what happened with email nurturing, at a certain point the system started to strain. According to our usage data, one in every 30 website visits results in a chat. For companies that receive thousands of website visits a day, trying to keep up is daunting. And, customers are again the ones suffering when companies can’t manage the demands of live chat.

Recent research found that 21% of live chat support requests go completely unanswered. Even if the buyer gets a response, they can expect to wait an average of 2 minutes and 40 seconds for it. I wouldn’t call this ‘live’ — would you? Responding slowly (or failing to respond at all) on a channel advertised as ‘live’ is a contradiction in terms. Forcing customers to wait after we’ve set the expectation of immediacy is unacceptable. We can do better. Today, we’re at the same inflection point we came to with email. What

 

 

5

 

518-067 HubSpot and Motion AI: Chatbot-Enabled CRM

 

 

 

should companies do to accommodate the tidal wave of live chat conversations? Hiring an increasing number of chat coordinators clearly isn’t a scalable answer. If marketers are going to advertise ‘live’ channels — we need to step up and deliver.

Consumer research offered conflicting opinions. While 40% of people claimed that they didn’t care if they were serviced by a person or an AI tool as long as they were helped quickly and easily, 42% of people wanted a human agent to help answer complex questions and requests. 75% of people didn’t think that chatbots would be sufficient for complicated troubleshooting and 90% felt that they should always have the option to transfer to a live agent. Direct experience with existing IVR phone systems and online chat demonstrated that many consumers still preferred speaking with a live customer service representative in an instantaneously synchronous manner, pressing “0” for an operator in IVR phone systems and bailing out of online chat conversations to dial in to a call center for help.

Botched Bots

Although bots were currently chatting with customers at astonishingly high rates, their record of success was less high flying. Facebook reported that chatbots failed to serve customer needs 70% of the time. As an example, only 12% of bot interactions in the health care sector were completed without the need to pass off the customer to a human operator. Lamented Coffey,

Bots provide a scalable way to interact one-on-one with buyers. Yet, they fail when they don’t deliver an experience as efficient and delightful as the complex, multi-layered conversations people are accustomed to having with other humans. Too often, bots today don’t understand conversational context, or forget what you’ve said two bubbles later…Consider why someone would turn to a bot in the first place. Of the 71% of people willing to use messaging apps to get customer assistance, many do it because they want their problem solved, quickly and correctly. And, if you’ve ever struggled to have Siri or Alexa understand what you’re asking, you know there’s a much lower tolerance for machines to make mistakes.

Despite rapid advances in artificial intelligence, most chatbots were still quite reactive and dumb. Programmed to only recognize a very limited set of commands, they had difficulty with back and forth conversation with humans. Tim Tuttle of MindMeld says, “The opportunity is clear, but today most companies still have huge challenges building chat applications that actually work. The industry is in a state of shock at how hard this is.”8 Explained Sarah Guo of Greylock Partners, “Language is hard to model (and program) because it is so ambiguous. Similar sentences can have very different meanings, seemingly different sentences can have the same meaning. Humans are strange, unruly, unconscious, and inconsistent in their communication, but make up for that by being so flexible in their ability to understand imperfect, ambiguous communications from others — based on context.”9 While humans effortlessly dealt with this complexity of language, bots stumbled.

While advancements in machine learning were helping, AI required big data to be effective, said Robert C. Johnson, CEO of TeamSupport, “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 a lower volume of interactions but higher complexity, which can lead to higher error rates. Chatbots are good for B2C interactions where there’s a high volume and the value of each customer is not very high.”10 Bots also struggled to handle complex problem solving. Explained Daniel Polani of the University of Hertfordshire,

There is an art to handling the exception, and good customer service is often about the unusual or unanticipated cases involving potentially angry customers. While chatbots can

 

6

 

HubSpot and Motion AI: Chatbot-Enabled CRM 518-067

 

 

 

convincingly source answers to basic questions, AI isn’t yet smart enough to deal with the rare and exceptional examples…Automated systems might be able to handle regular cases. But they can’t yet adapt themselves to exceptional circumstances or even recognize that the flexibility of human intervention is needed. And…some situations require not just human understanding and problem solving, but a level of compassion and empathy. A chatbot can be programmed to adopt a certain style of interaction, but that will still sound oddly out-of-place in unexpected or difficult contexts.11

However, much of the challenge of creating an effective chatbot derived not from the limitations of the technology, but rather from the difficulties associated with designing a Conversational UI, anticipating the conversational flow that a bot would need to have with diverse customers. “The difficult in building a chatbot is less a technical one and more an issue of user experience,” said Matt Harman, director of seed investments at Betaworks.12 Proclaimed Bodnar, “We need conversational strategy + the automation of bots. This is what will make us more efficient, but more importantly, more effective for our customers. This is automation that creates relationships instead of frustration.”

Coffey believed that chatting with a bot should be like talking to a human that knows everything. But, Altimeter suggested that emotional intelligence was as important as IQ, “Detecting emotion, expressed in word choice or tone, [is] also critical to ensure that conversational experiences are satisfying for users.”13 A strong Conversational UI could capture users’ attention through an engaging and evolving narrative that combined automation with intimacy. However, this required significant relational intelligence and the ability to perceive differential relational styles and trajectories. Clara de Soto of Reply.ai, agreed, “You’re never just ‘building a bot’ so much as launching a ‘conversational strategy’ — one that’s constantly evolving and being optimized based on how users are actually interacting with it.”14 And, this was difficult, explained David Shingy of AOL, “The challenge [with chatbots] will be thinking about creative from a whole different view: Can we have creative that scales? That customizes itself? We find ourselves hurtling toward another handoff from man to machine — what larger system of creative or complex storytelling structure can I design such that a machine can use it appropriately and effectively?”15 According to Advertising Age’s Annie Fanning,

Fully owning your conversational relationship with your customers requires building a brand-specific chatbot personality…you’ll need word nerds on both the front and back end to feed and teach your new baby chatbot. Not only does someone need to craft chatbot responses with personality (brand-guided voice and tone) but a writer/strategist/UX expert will need to think through the customer journey and provide sample customer input. To build an effective bot, every use case needs to be considered and a chatbot response written for every type of interaction you can think of…This means knowing what your customers are asking, and how they [will] phrase their questions, is just as important as knowing how the bot will respond.16

Consumers were getting frustrated with many of the bots with which they interacted. Said one after interacting with travel-related bots, “Every experience I’ve had has been a total waste of time. I would love to hear at least one positive anecdote about using artificial intelligence.”17 Fanning cautioned marketers about the downside of bots, “When a chatbot guesses wrong and serves up content we didn’t ask for, it is at best hilarious, but at worst offensive and embarrassing.”18 Echoed USA Today, “These early days of…bots…are a cautionary tale. Technology may be good and getting better but nothing replaces a person. That’s unlikely to change for a while, and maybe ever.”19

 

 

 

 

 

7

 

518-067 HubSpot and Motion AI: Chatbot-Enabled CRM

 

 

 

How Human is Too Human?

As HubSpot looked ahead to a world of chatbots, one thing they needed to address was how human-like bots should behave. Some were suggesting that companies should not disclose that customers were interacting with AI, but rather, allow them to assume that they were chatting with a live human in order to reap the benefits of human-built relationships. Said Beerud Sheth of Gupshup, a bot creation platform, “Chatbots are everywhere. 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 seem like just another user you’re messaging.” 20 “People don’t even always know they’re interacting with bots. The whole thing only works when it’s just so easy that you don’t even think about the fact that it’s a bot,” said Matthew Hartman of Betaworks.21 Left to their own devices, humans had a tendency to interpret computer-generated conversation as coming from a person anyway, so customers often anthropomorphized chatbots, observed Arte Merritt, CEO of bot analytics platform Dashbot, “People think about bots for customer service, but they’re so much more…Users treat the bots as people.”22 In a humorous example, the company x.ai humanized its meeting scheduling bot so well that customers were asking “Amy” out on dates, not realizing that “she” was an AI-driven personal assistant.23

This often led to an uncomfortable situation labelled “the uncanny valley”. While people generally preferred to engage with computer programs that were more, rather than less human-like, their response to an anthropomorphized robot would abruptly shift from empathy to revulsion if the robot suddenly failed to act human enough. Explained Justine Cassell of Carnegie Mellon, “When a bot is clearly a bot, the person interacting with it generally knows how limited its functions are…The bot’s narrowly defined purpose guides the human that’s interacting with it. By contrast, a smooth-talking virtual assistant that tries to mimic human speech…can create different assumptions. The more human- like a system acts, the broader the expectations that people may have for it.”24

 

"Looking for a Similar Assignment? Get Expert Help at an Amazing Discount!"