Artificial Intelligence (AI) and its application (bots, for example) have grown significantly in recently years. Such technology has been adopted by companies to improve performance, particularly in customer service. Chatbots appear to be a commonly used and growing form of AI-powered technology, especially in the financial sector. Below is a summary of key findings on the use and acceptance of AI, customer sentiment and a number of examples of successes and failures, as well as case studies on specific forms of AI technology.
Current Use and Acceptance
Percentage of Companies Using AI/Bots
- Only 24% of service organizations use AI in costumer service in 2019, according to Salesforce’s “State of Service,” a survey of more than 3,500 customer service agents and decision-makers worldwide.
- However, 34% are planning to use AI within 18 months (from March 2019, when the study was published). This is a 143% projected growth.
- A survey of over 550 senior executives conducted by MIT Technology Review (2017) finds that AI is used to improve costumer satisfaction in 91% of companies with world-leading brand recognition and high customer satisfaction. Only 49% of other companies in their fields employ such solutions.
- The report on the survey refers to these top companies as “iconic.” They include Alibaba, BT Global Services, Lexus, Nubank, Uber and Zurich Insurance.
- Of these iconic companies, 60% say they have the right mix of human and automated customer communication channels. Only 26% of poor performers and 40% of overall companies believe so.
- The report finds that leadership in technology adoption is crucial in maintaining excellence in customer experience for these iconic companies. Only half of low-performing companies use enabling technologies and 10% of them say they are not intended to adopt them.
- According to Salesforce’s “State of Service,” about 23% of service organizations use chatbots in 2019 and 53% is projected to do so by 2020, a 136% growth rate.
- The chatbot market size is estimated to be $2.6 billion in 2019 and projected to grow to $9.4 billion by 2024 at an impressive CAGR of 29.7%, according to Markets Insider.
- The chatbot market in North America is growing at a CAGR of 31.2%.
- Although there are multiple segments in the chatbot market, chatbot service has a majority of the market share. Customer service chatbot is expected to be the fastest growing segment between 2019 and 2026 at a CAGR of 31.6%.
- Industry segments for the chatbot market are banking, financial services and insurance (BFSI), media and entertainment, retail and e-commerce, travel and hospitality, telecom, healthcare and life sciences, and others.
- The retail and e-commerce segment is projected to be the fastest growing between 2019 and 2024.
- However, the banking segment is currently dominating the chatbot market, making up 24% of the global market. North America is the biggest source of revenue for this segment, followed by Europe and Asia Pacific.
- The travel and hospitality (tourism) segment is projected to reach $1.6 billion in 2026.
- These portals help customers find answers to basic questions fast and on their own terms.
- Nearly two-thirds of service organizations use these portals in 2019 and 84% is projected for 2020, according to Salesforce’s “State of Service.”
Social Monitoring Tools
- In 2019, 72% of service organizations use social monitoring tools to detect action posts by customers on social media to track their online sentiment, according to Salesforce’s “State of Service.”
- These tools are integrated with customer relationship management system and allow agents to have a 360-degree view of the customer before responding.
How AI Is Used by Customer Service Organizations
- It is used to gather basic information (81%), automate the handling of routine customer issues (75%), classify and route cases (74%), provide management with operations insights (71%) and pre-fill fields in the agent console (71%).
- According to Salesforce’s “State of the Connected Customer” (2019), a survey of more than 8,000 consumers and business buyers, 54% of respondents think companies need to transform how they engage. The survey reveals that customers expect companies to know what they want before they ask.
- About 75% of customers say they expect companies to use new technologies to create better experiences for them.
- According to Salesforce’s “State of the Connected Customer” (2018), a survey of over 6,700 consumers and business buyers, 59% say tailoring engagement based on their past interactions is very important to win their business.
- Respondents are 2.1 times more likely to say personalized offers as important than unimportant.
- They are 9.7 times more likely to view AI as revolutionary than insignificant.
- The survey also finds that customers expect immediate and responsive services with 64% of individual customers and 80% of business buyers saying they expect real-time responses and interactions. This expectation is even higher among younger population, the survey finds. This calls for companies to make a choice between increasing costumer service headcounts or deploying chatbots and some other forms of AI-driven technology.
- In addition, 75% of business buyers expect companies to anticipate their needs and make product suggestions before they initiate contact.
Costumer Experience and Engagement
- Majority of millennials and Gen Zers (67%) use and prefer voice-activated personal assistants (e.g., Siri and Alexa) to connect with companies, compared to 39% of baby boomers and traditionalists and 50% of Gen Xers, according to Salesforce’s “State of the Connected Customer” (2018).
- The majority of the three groups use and prefer online portals (e.g., self-service account information)–88% millennials and G Zers, 83% Gen Xers and 77% baby boomers and traditionalists.
- Overall, digital channels are widely used and preferred (see details here).
- AI is impacting customer sentiment–with 14% say it has already transformed their expectations of companies, 37% say it is actively transforming and 36% say it will within five years.
- Both chatbots and voice-activated personal assistants are gaining grounds–with 77% and 84% of customers view chatbots and voice-activated personal assistants, respectively, as the technology that either has already transformed, is actively transforming or will transform their expectations in five years (see details here).
- AI is winning customer approval, with 67% recognizing the good it can do and 61% believing that it provides positive opportunities to society in 2018. In 2019, 59% agree that AI will revolutionize how they interact with business.
- In 2019, 62% of customers are open to the use of AI to improve their experiences, an increase from 59% in 2018.
- Customers’ most favorite AI-driven experience is credit card fraud protection, with 86% say they love or like it in 2018. In comparison, 76% say they love or like automatic reminders, 56% personalized recommendations and 53% voice-activated personal assistants.
- While customers approve of AI, they do not yet fully trust companies with their data, with 57% say they are uncomfortable with how companies use their information and 45% say they are confused about how their data is used.
- Nonetheless, 79% are willing to share relevant personal information in exchange for contextualized interactions in which they are immediately known and understood. Note the condition of “known and understood.”
Success Stories of AI/Bot Customer Service
Allstate Business Insurance
- Allstate Business Insurance uses ABIE, a customer service bot previously designed to help salespeople. Insurance agents use the bot to look up information that used to require manually searching through hundreds of documents.
- ABIE handles more than 25,000 inquiries every month.
- The technology won Allstate the KM Reality Award in 2015.
- ABIE has been extended to customers as well. It provides answers to initial questions asked by small business owners. This allows their later conversation with agents more productive. It works to supplement the agent/customer relationship, according to Allstate.
- According to the designer of ABIE, “[m]any chatbot implementations struggle to scale because they’re built on single purpose content that is expensive to build and maintain.” He attributed the success of ABIE to its structured content.
- With its implementation, ABIE has reduced 25,000 calls per month.
InterContinental Hotels Group (IHG)
- IHG uses virtual agents to solve technical problems for employees who call the IT help desk. In this case, the employees are customers to the IT help desk.
- Machine learning is employed to read chat transcripts in order to answer common questions. By handling the common questions, the virtual agent frees up the help desk staff for tasks requiring human attention. It works with the staff instead of replacing them.
- It is reported to be handling 80% to 85% of the volumes of the tasks it is trained to do.
- Even in cases where it cannot answer the question, the virtual agent collects basic information before handing it off to a staffer.
- Among those engaged with the system, satisfaction rates remain high.
- Marriott deploys multiple chatbots available through Facebook Messenger, Slack, WeChat and Google Assistant.
- The chatbots are offered in multiple languages, allowing many guests to communicate in their native language.
- Through natural language processing (NLP), the chatbots are able to detect customer behaviors and preferences to create “hyper-personalized recommendations” for customers.
- Aloft, a part of Marriott International, incorporates its bot butler Chatbotlr, which allows guests to request services from their smartphones. With each request, the bot becomes smarter (more accurate) as it learns more from each interaction.
Examples of Failures
Mr. Cooper, Formerly Known as Nationstar
- The U.S. largest non-bank mortgage provider launched an intelligent recommendation system for its customer service agents, aiming to suggest solutions for customer problems.
- After nine months, the company realized that its agents were not using the system. It took six months to figure out why.
- The system were offering recommendations that were not relevant.
- While the machine learning algorithms were not the problem, the system used data training based on technical descriptions of customer problems instead of customers’ own words.
- In 2018, T-Mobile discontinued its interactive voice response (IVR) system.
- While stating that the IVR performance was not measured by customer happiness, T-Mobile made a public announcement to scrap the system without giving details of what went wrong with its IVR. The company quite enthusiastically says “your call goes straight to your team, no bouncing, no bots, no BS.”
- It cited a survey saying 39% of customers would rather clean a toilet than deal with an IVR.
- McKinsey (referring to T-Mobile anonymously as a major U.S. mobile operator) reports that the reason for the discontinuity is a high level of customer frustrations.
- McKinsey provides five reasons that IVR could go wrong: one-size-fits-all mentality, confusing navigation and terminology, poor integration with other channels, lack of timely updates and lack of satisfaction measurement.
Cases of AI/Bot Customer Service and Engagement
Chatbot Case Study: Lemonade Insurance
- Chatbots appear to be a more common form of AI technology in customer service. The systems can typically handle 80% of questions without human assistance.
- Experts suggest that an efficient system is one that is designed for service flows in which the bot and humans work together. See the success stories above.
- As mentioned in the “Use and Acceptance” section above, the BFSI industry is currently dominating the chatbot market. Business Insider Intelligence estimates that the global annual cost saving by the deployment of chatbots will increase from $0.5 billion in 2020 to $5.8 billion in 2025 for the insurance sector alone.
- It also predicts that the use of such AI interactives will be the standard for customer service across all industries.
- Lemonade is a provider of homeowners and renters insurance. It was looking to cut cost and improve time efficiency by replacing human process with bot technology in its customer service.
- The insurance company developed a scalable bot framework which consists of different chatbots.
- The first chatbot is a policy chatbot, Maya. It helps users to get a new policy within minutes, make changes to their policy, and cancel their existing policy and create a new one. Maya can onboard users in as little as 90 seconds whereas online traditional insurers would take approximately 10 minutes.
- Jim is another chatbot that settles claims. According to Lemonade, Jim settles claims in as little as seconds, compared to anywhere between 48 hours and over a year it could take incumbents.
Emotion Analystics Case Study: USA Today
- According to Markets Insider, global market size of emotion analytics is projected to reach $5.1 billion in 2025 with a CAGR of 15.6%.
- The AI technology extracts information from all customer touchpoints and channels, including calls, texts, video, facial, emails, chats, and social media platforms.
- Tracking historical and real-time data, it helps call agents to tailor word choices based on the patterns and trends of customers.
- It also helps companies to create offers that suit the customers, improving customer retention.
- The types of analytics used in this market are text analytics, speech analytics and facial/video analytics.
- Emotion analytics are used in customer experience management, sales and marketing management, competitive intelligence, workforce management and others. Of these segments, customer experience is projected to be the largest share of the market.
- International news company USA Today experienced customer frustration in account-related activities. The frustration was detected by analytics of surveys, ratings/reviews and live chat transcripts as part of its efforts to retain print subscribers and improve relationship with digital audiences.
- With the information it gathered, the media company responded by improving its online transaction performance and adjusting search functionality to make finding account-related resources easier.
- Its analytics also suggested that many digital subscribers were unaware of its wide range of products. In response, USA Today created an online member guide and an ad-free subscription app to improve experience for those who were frustrated by autoplay ads.
- USA Today has since seen an increase in subscriptions and engagement satisfaction.
Product Recognition Case Study: Amazon Go and Further Development
- Amazon Go, a convenience store by Amazon, uses AI technology to track the products that customers pick up and charge them accordingly without checkout counters. Customers only need its smartphone app to get in the store and walk out without checking out items. Amazon Go bills the customer through its app.
- Amazon Go stores are smaller, mostly from 1,200 to 2,300 square feet, but they can be as small as 450 square feet. Thus, product variety is limited.
- Almost all Amazon Go stores reportedly have user ratings between four and five stars on Yelp. Amazon Go app has a 4.5-star rating on Apple App Store.
- Neilson and Oxford University have partnered to develop the largest pool of product data references for product identification and discovery.
- The team is working to identify and classify in-store products quickly and accurately based on images captured by Neilson’s eCollection.
- Oxford researchers are tasked with building and enhancing the eCollection algorithms. Utilizing advanced deep learning capabilities, the task will enable a more automatic detection of store products, promotions and prices without manual intervention.
- While AI has been successful in using big data, the next step is to achieve the use of very small details such as recognizing specific features on different products with similar visual characteristics. This ongoing AI development project is aimed to achieve that.
To carry out this research, we separated our search into two categories: companies and customers. We began our research for relevant data on news sites that typically cover topics related to business and technology in order to identify potential data sources. From there, we gathered information from a number of business technology data sources such as Salesforce and MIT Technology Review–both offer a number of extensive surveys. Much of the data presented in this research rely on these surveys. The case studies were selected from top news related to AI topics. Please note that a bot is an application of AI and the two terms are often used interchangeably.