What is CX? What is AI? Insights and learnings on why they matter, how AI can improve CX, and what companies are—and should be—doing to stay ahead.
Key Takeaways
Key Takeaways
Nearly half of US marketers polled in September 2021 have increased CX-related AI investments in the preceding year.
AI will play an increasingly important role in CX.
Investment in AI for CX purposes is increasing, and AI will be used in a growing number of ways. Companies that want to stay ahead of the curve will need to adopt the technology for CX purposes over the next five years.
Firms should experiment with different applications of AI to find what is most valuable for them. Companies like Spotify and Tripadvisor are trying out different AI applications for CX, like predictive recommendations, and intend to scale these uses right from the start.
Companies deploying AI should be transparent about its use with their customers. Consumers are more likely to share data for a better experience if they know how their data is being handled. Businesses that prioritize data privacy and responsible AI practices will have a competitive edge.
I. AI Spending in CX
Nearly half of marketers have increased their AI-related CX spending. That’s according to a survey from the Harris Poll performed for RedPoint, which surveyed US marketers in July 2021 about their outlays over the preceding year.
CX technology spending is set to balloon. Increases in AI-related CX spending will mirror an overall increase in CX technology spending, which is set to reach $641 billion in 2022, according to International Data Corporation (IDC) data.
IDC also reported that two customer service applications—augmented customer service agents and sales process recommendation and augmentation—are expected to account for more than 20% of all US AI spending in 2025.
What Is CX?
What Is CX?
Customer Experience (CX): the perception that customers have of an organization that is formed based on interactions across all touchpoints, people, and technology over time.
Source: Customer Experience Professionals Association (CXPA)
What Is AI?
The term AI can mean many different things, but there are a few sub-sections relevant to this report:
Machine learning (ML): A branch of AI and method by which computer systems learn, analyze, and interpret data to take actions on their own without programming.
Natural language processing (NLP): A branch of AI that enables computers to understand, interpret, and respond to human language. It includes subsets like natural language generation and natural language understanding and has applications in both text-based and voice-based interpretation.
Predictive Analytics: Use of ML techniques and statistical modeling to identify the likelihood of future outcomes based on historical data. Advanced analytics uses ML in combination with other sophisticated techniques and tools (like visualization techniques) to identify deep insights.
Why CX Matters
CX is important to consumers and can drive revenue and loyalty. A July 2021 survey of US internet users conducted by Alida in select countries found the following:
- Consumers prioritize companies who care about the CX: 92% of US respondents said that CX was an important factor in their purchasing decisions. The survey also found 87% of US respondents were loyal to brands focused on CX, while 85% agreed positive CX is more important than convenience.
- Good CX adds value in the eyes of the market. The 20 companies with the highest customer satisfaction scores generated double the shareholder value than the companies in the S&P 500 did during the period lasting from 2009 to 2019, according to Boston Consulting Group (BCG). The American Customer Satisfaction Index (ACSI) similarly reported that companies with higher ACSI scores—a measure of customer satisfaction with a firm—typically offered better returns than the S&P 500.
Why AI Matters
AI drives innovation, which is important for CX success. Nearly 90% of decision-makers in North America strongly agree or agree that continuous innovation in CX is needed to prevent losing customers to competitors, according to Q2 2021 research from Zendesk and Enterprise Strategy Group.
CX applications are one of the most promising uses for AI tech. AI—and especially machine learning (ML)—is changing how CX is built and delivered to customers. According to [24]7.ai, 72% of CX professionals globally said their companies used software-based phone assistance like a voicebot in 2021, up from 69% in 2020.
There is a general consensus among CX professionals that the role of AI in CX will only become more prominent over the coming years. A March 2021 survey by Talkdesk, a CX services company, found 64% of CX professionals expected their companies’ contact centers to prioritize deeper investments in AI tech.
What Types of AI Can Help CX?
ML, natural language understanding (NLU), and natural language processing (NLP) are all types of AI that can help analyze customer sentiment and customer feedback on a large scale. Analytics tools used with ML can provide powerful insights about customers.
Organizations now use different types of AI to solve for different CX problems. According to Coresight Research, 44.2% of US executives are experimenting with facial and voice recognition, 43% with personalized channel experience, and 41.2% with NLP for content generation.
Key Uses of AI in CX
Businesses are already well versed in the use of certain AI tools to improve CX, like voicebots and chatbots. But adoption of newer applications of the technology, like content generation, remains low. Here are the main uses of AI that companies will use to enhance their CX.
Customer Service
AI can improve customer service and support through contactless, personalized customer care.
AI-enabled support. More than half of CX professionals worldwide planned to implement software-based phone and digital support channels delivered from the same AI-enabled software solution platform, according to July 2021 data from [24]7.ai. A majority also planned to implement voice agents and software-based phone assistance using digital capabilities like visual information shared with customers, the research revealed.
AI is changing customer service in three significant ways:
Automated response: Conversational AI technology, including NLU and processing in virtual customer assistants, can reduce response times. This can eliminate the hassle of returns or provide immediate answers to customer questions.
Customer service training: AI that analyzes qualitative feedback from customers can help companies create customized staff training programs that better meet customers’ needs and desires.
Identifying pain points: Businesses can use AI to unearth the causes of CX problems, allowing them to prioritize improvements.
Chatbots are popular NLP tools to address all of those issues. Chatbots have been one of the earliest types of AI adopted by companies: Coresight’s November 2021 survey of US executives found 64% use the technology to improve customer satisfaction.
Consumers have an overall positive view of AI and automated chat technology with about 60% having a positive perception, according to November 2021 data from Botco.ai.
Botco.ai also found chatbot uses vary widely among consumers: 18% rely on them to find out business hours, 17% for product information, and 16% to find nearby locations. Companies will need to experiment to figure out which uses offer them the best return on investment.
Insights
AI-enabled analytics like predictive analytics and deep analytics can help companies understand customer behavior better, leading to improved CX.
There are some shortcomings in two measures widely used to understand what customers think and feel, according to research from the University of Cambridge reported by the Harvard Business Review: customer satisfaction and Net Promoter Score pertaining to the efficiency and accuracy of those metrics.
The adoption of AI models and tools is still low today and often limited to indicating positive or negative sentiment. But this is changing.
- Predictive analytics can anticipate customer purchasing preferences and remove roadblocks ahead of time. It can also be used for customer segmentation, allowing companies to make CX enhancements in real-time.
- NLP can extract insights and analyze sentiment from a wide range of customer feedback touchpoints like reviews, complaint emails, and chatbots. NLP can determine user satisfaction levels and identify important customer support and critics.
- Touchpoint analysis can determine which media or interactions—like mobile app usage or social media—improve CX and drive repurchases. The touchpoints that customers actually care about may not be the ones companies expect. This helps businesses learn about their customers and anticipate valuable touchpoints.
- Enhanced facial recognition, sometimes referred to as emotional AI, can improve CX measurement by capturing emotional and cognitive responses in real-time. It can also contextualize responses through customer evaluations, like complaints or suggestions. This captures real-time feedback that can be lost in post-purchase surveys.
Customized Content
According to the November 2021 Coresight Research survey, 72.2% of US executives were either currently using or planning to use AI for content generation and customizing what content customers see.
- Content personalization: AI can help marketers create bespoke content for each consumer across marketing channels. These tools can help target consumers by selecting the best words and phrases or identifying consumers’ individual interests. This is also sometimes called dynamic content.
- Content generation: Digital marketers are already using AI to distribute, manage, and analyze content. Its use in creative content generation—identifying topic ideas or creating written or visual content—is next. The Coresight Research survey found more than 4 in 10 respondents used NLP to generate content.
Personalized Experiences
Understanding the consumer through real-time contextual data—like location, browsing data and purchase history—and then delivering a personalized experience is another key area where AI can help.
More personalized and flexible CX is a priority for companies. RedPoint found 41% of US marketers named personalized experiences and deep relevance as leading drivers in CX strategy innovation. More than a third said creating flexible delivery systems for services was also a key driver.
For example, in 2021 American Express released a “contextual and predictive search capability” in its app designed to predict what customers want before they even begin typing. If the person, say, opens the app at the airport, it will assume they are trying to find the lounge. The tool is trained on an NLP model initially designed for their customer service chatbots.
Levi Strauss & Co. uses AI to help set pricing, as well as to personalize marketing, predict demand, and optimize fulfillment. It credits a 20% to 21% year-over-year increase in revenue seen in Q3 2021 at least partially to the use of AI.
Identifying Challenges
IT decision-makers worldwide found lack of data quality and lack of expertise within the organization were the primary factors that led to AI and ML research and development failures, according to Rackspace Technology.
The future opportunity lies in getting more—and the right types of—data. While the infrastructure needed to support AI initiatives may be in place, companies are limited by the types of data they’re gathering. Often, the necessary context is lacking—such as when someone places an item in a cart but doesn’t take another step.
Companies should strive to augment simplistic data with more context and emotional intelligence.
AI Technology in Use: Case Studies on AI Implementation in CX
Tripadvisor
Tripadvisor used conversational AI technology for a campaign executed on voice assistants like Alexa and Google Assistant with the goal of maintaining engagement during the pandemic, according to Adam Ochman, global head of marketing at Tripadvisor.
The issue: The company saw engagement levels drop as the pandemic stifled travel plans. As a result, it wanted to improve the ease of CX and increase touchpoints in a time of low engagement.
What Tripadvisor did: Tripadvisor launched two voice assistant campaigns where users could interact with itineraries on Alexa or Google Assistant, one for Abu Dhabi and the other for Visit Orlando. The campaigns attempted to fit into the audience’s daily life and met consumers where they spent the most time—in their homes.
The user was able to have a conversation with a voice assistant and engage with the itineraries, eventually heading to Tripadvisor’s website if they decided to continue planning their trip. The experience weaved in content from the website like reviews, which allowed users to access this information in a new way. Tripadvisor also tested, learned, and amended its plans during the campaigns.
The result: Tripadvisor considered the campaigns successful. Users spent more than 4 minutes on average engaging with the campaigns, answering approximately seven prompts. In all, users engaged with the Visit Orlando campaign for a combined 450 hours and spent more than 1,300 hours with the Abu Dhabi campaign. Both campaigns increased performance. Tripadvisor also won a Drum Award for “Most Effective Use of AI/Machine Learning” as a result of this campaign.
Where else is it considering AI: Tripadvisor is considering the use of dynamic content tailored to their individual users as well as predictive analytics for better creative output and delivery.
It is also contemplating the metaverse—with ideas like itineraries accessed in virtual reality before buying—and travel retail applications.
Spotify
Spotify uses AI for many purposes, including its pioneering personal recommendation model, which relies on semantic analysis. Magnus Friberg, data scientist at Spotify, said customer support is another area where the company is experimenting with AI.
The issue: Spotify wanted to make the customer support journey as autonomous and efficient as possible.
What they did: Spotify implemented a customer support chatbot around two years ago, before which all interactions went through a customer service representative. The chatbot offers several basic support functions and is connected to Spotify’s API. The company has focused the chatbot’s functions on instances where it can solve tasks based on automation provided by the bot-API connection. For instance, it can sense when users are repeating certain behaviors, like changing account settings, and provide automated help with those tasks.
Bots may handle most support tasks within the next five years due to the expected increase in size of AI language models, which help predict the occurrence of a word or phrase. But customer service representatives will still be necessary for more complex issues and to interact with users who prefer talking to people. Spotify also directs users to informational articles.
In addition to the bot, Spotify also built an AI model to help classify issues, a task once considered infeasible due to the sheer quantity of them. The model classifies interactions from all sources—including the website, app, Twitter, Reddit, Facebook, direct emails, and the bot—to identify larger trends and determine whether an issue is actionable. Spotify also uses various models for outlier detection, which can help them determine the behaviors of customer segments using forecasting models.
Spotify is also using AI to try and predict a user’s next steps and if the user will contact them. The eventual goal is to nudge them in the right direction—for example, by alerting them about an expiring credit card.
AI is also used to support customer service representatives. Models are used to predict issues and present employees with relevant information cards ahead of time to improve their experience.
Most of the AI models used by Spotify are open source but a lot of internal algorithms are built internally. Spotify relies on the open-source software catalog and developer platform Backstage, which uses templates. If someone wants to start a new AI workflow, they can use these templates as a first step.
The result: Spotify’s AI-driven efforts have improved customer perception and removed friction. The efforts decreased the number of responses and handle time. Spotify uses a few key performance indicators for its bots, including resolution score, closed cases per online hour, and deflection rates.
Where else is it considering AI: Language models are rapidly increasing in size and capability and will enable dramatic improvements in all areas Spotify is investing in. They’ll also open entirely new applications.
In March, Spotify filed a patent for an AI-based system that can predict breaking artists. The new system can predict which artists are likely to become popular based on the listening patterns of early adopters. If all goes as planned, this could be a game changer for how artists get discovered and recognized by both Spotify and its users.
Best Practices
- Understand CX’s role across the company before considering an AI strategy. This means mapping and analyzing the customer journey to understand touchpoints and experiences. The organization’s governance model should allow accountability to be assigned throughout this process.
- Focus on capturing the right high-quality data. AI is only useful if companies first understand what their customers’ pain points are. Capturing good quality data is at the center of this. Adding context to data wherever possible is also incredibly important.
- Strive for digital intimacy through “good” friction. AI is currently very good at removing friction points. But those thinking ahead of the curve will use AI to introduce good friction—like amplifying nostalgia or gamification through technology such as prediction engines and emotional AI, according to Jennifer Dalipi, global consulting partner at Ogilvy. The ultimate goal is for companies to anticipate customers’ needs and understand their values using an empathetic set of data.
- Ensure the responsible use of AI. Consumer concerns about the ethical and responsible use of AI by organizations—and wariness of bias in the technology—will only grow. These worries now also apply to companies using third-party AI tools. Businesses need to have responsible AI policies and assess their risk regularly. Companies should consider using third-party evaluations or certifications to detect and correct bias in their AI.
- Focus on transparency around privacy issues. Companies need to be proactive about transparency with customers by educating them, building relationships, and making a distinction between data used for CX and data sold to advertisers or used for other purposes. Consumer trust around these issues will lead to less friction and more innovation.
- Make sure CX initiatives are aligned across the business. According to BCG, a retail bank had 220 different uncoordinated initiatives across 30 functions to improve CX that failed to come to fruition due to lack of collaboration and planning. While this may be an extreme example, the failure to ensure alignment across functions will lead to an unsuccessful AI rollout.
- Test first, scale second. Experiment with different applications but have systems in place that determine which are succeeding. A plan to scale each option is also critical.
