Skip to Content


Artificial intelligence

From support function to growth engine: The future of AI and customer service

Sophisticated algorithms drive and inform a new level of fully automated and human-assisted customer service.

In association withSalesforce

When it comes to imagining the future, customer service often gets painted in a dystopian light. Take the 2002 sci-fi film Minority Report. Tom Cruise’s John Anderton walks into the Gap, an identity recognition system scans him, and a hologram asks about a recent purchase.

There’s something unsettling in this vignette—an unsolicited non-human seems to know everything about you (or, as in the movie, mistakes you for someone else). But the truth is, customers today expect this kind of sleek, personalized service. Their relationships with retailers, banks, health-care facilities—and virtually every organization they have business with—are changing. In an always-on, digital economy, they want to connect when they want, how they want. Customers want their product questions answered, account issues addressed, and health appointments rescheduled quickly and without hassle.

They’re starting to get it. Today, when customers call a company for details on its products, the conversation is guided by a chatbot. They answer a few simple questions, and the chatbot steers them in the right direction. If it can’t answer a query, a human agent steps in to help. The customer experience is fast and personalized, and customers are happier. On the flip side, agents are more effective and productive. Behold the real future of customer service.

Artificial intelligence (AI) and customer relationship management (CRM) software are paving the path to that future. Together, the technologies can automate routine tasks, freeing up human agents and providing them with data-driven insights to help swiftly resolve customer problems. They’re helping retailers, banks, government agencies, and more rethink the goals of their customer service centers, allowing their teams to evolve from a support function to a growth engine.

Today, advancements in AI and machine learning are enabling deeper levels of customer engagement and service than ever before.

But stiff challenges remain. The goal for organizations is to offer the same customer service across all channels—phone, chat, email, social media—but at most organizations today, the technology isn’t quite there yet. AI technologies must be able to understand human speech and emotional nuances at a deeper level to solve complex customer problems. And in the absence of universal standards governing the ethical use of AI, organizations need to build a set of guiding principles that puts the needs of customers first—and establishes the kind of trust between humans and machines that makes it all tick.

Automate or stagnate

In a February post, Gartner predicts, “by 2022, 70% of customer interactions will involve emerging technologies such as machine learning (ML) applications, chatbots and mobile messaging, up from 15% in 2018.”

Today, advancements in AI and machine learning are enabling deeper levels of customer engagement and service than ever before. Powerful and trainable algorithms can parse through massive amounts of data and learn patterns to automate and assist customer service processes. This technology is changing the face of customer service and helping organizations understand customers’ needs—often before they even do—providing the service they need at the right moment, says Jayesh Govindarajan, vice president of AI and machine learning at Salesforce.

“AI being used in nearly all aspects of customer service, starting with auto-triaging customer cases to agents with the right skill sets, and followed by assistive AI that steps in to surface information and responses that help agents resolve cases faster and with precision,” says Govindarajan. There’s even AI that can use context in a conversation to predict a response. “If I say ‘I’m hungry—it’s time to grab some …,’” Govindarajan says, “it knows I'm probably going to say ‘lunch’ because it's mid-afternoon.”

The 2020 coronavirus pandemic is accelerating the transition to digital-first service. Human interactions are becoming increasingly virtual: people are doing more of their daily tasks over the internet, shopping online, and meeting and collaborating through virtual platforms. Organizations are recognizing the rapid shift and answering the challenge by adopting chatbots and other AI tools to gather information, classify and route customer cases, and solve routine issues.

The trend is playing out across all industries, with the greatest adoption in retail, financial services, health care, and government, according to Govindarajan. When people need help returning a product or renewing a driver’s license, the process is increasingly automated. The retail automation market alone was valued at $12.45 billion in 2019 and is expected to reach $24.6 billion by 2025, according to research by Mordor Intelligence.

Such wide-reaching adoption is possible because language models, the engines behind natural language processing, can be trained to learn a specific vernacular. In retail, for example, a conversational AI system could learn the structure and contents of a product catalog, Govindarajan says. “The vocabulary of the conversation is domain-specific, in this case retail. And with more usage, the language models will learn the vocabulary employed in each industry.”

The human-machine alliance

As this new level of customer service evolves, it’s heading in two general directions. On one side, there’s a fully automated experience: a customer interacts with an organization—guided by chatbots or other automated voice prompts—without the help of a human agent. For example, Einstein, Salesforce’s AI-powered CRM system, can automate repetitive functions and tasks such as asking a customer questions to determine the nature of a call and routing the call to the right department.

“We know exactly what the structure of a conversation looks like,” says Govindarajan. “You're going to see a greeting, collect some information, and go solve a problem. It’s practical to automate these types of conversations.” The more the model is used, the more the algorithms can learn and improve. A study conducted by Salesforce found that 82% of customer service organizations using AI saw an increase in “first contact resolution,” meaning the issue is resolved before the customer ends the interaction.

But AI-assisted responses have limitations. When a question is more complex or less predictable, human involvement is required—think of a tourist explaining a problem in a second language, or someone struggling to follow assembly instructions for a ceiling fan. In these scenarios, empathy is critical. A human has to be in the loop to work with the customer directly. So an agent steps in, refers to the CRM system for up-to-date customer data to get the needed context, and helps the customer resolve the issue.

“You can think of the role of the agent as training the system—agents correct machine-generated responses and take follow-up action,” says Govindarajan. “While the the system assists the agent towards the right answer using machine-learning models trained on prior similar, successfully resolved cases and on the customer’s previous interactions with the company.”

The agent is also able to cultivate a better relationship with the customer by supercharging the conversation with data-based insights, making it more personal.

Overcoming technology, ethics challenges

All this paints an exciting picture of the future of customer service—but there are hurdles to jump. Customers are increasingly engaging with companies via online and offline channels. Salesforce research found that 64% of customers use different devices to start and end transactions. This means organizations must adopt and deploy technologies that can provide the vaunted “single view of the customer”—an aggregated collection of customer data. Having this view will help enable multimodal communication—meaning customers get the same experience whether they’re on a mobile phone, texting, or emailing. Further, machine-learning algorithms need to become more efficient; conversational AI needs to evolve to more accurately detect voice patterns, sentiment, and intent; and organizations need to ensure that the data in their algorithms is accurate and relevant.

The challenges go beyond just technology. As contact centers adopt AI, they must focus on developing trust between the technology and their employees and customers. For example, a chatbot needs to let customers know it is a machine and not a human; customers need to know what the bot’s limitations are, especially in cases in which sensitive information is being exchanged, as in health care or finance. Organizations using AI also need to be upfront about who owns customers’ data and how they handle data privacy.

Organizations must take this responsibility seriously and commit to providing the tools customers and staff need to develop and use AI safely, accurately, and ethically. In a 2019 research note, Gartner advises data and analytics leaders: “Reach agreement with stakeholders about relevant AI ethics guidelines. Start by looking at the five most common guidelines that others have used: being human-centric, being fair, offering explainability, being secure and being accountable.”

In a world where it’s increasingly important to build strong relationships between organizations and the public, service presents the biggest opportunity to elevate customer experiences and go for growth. The value in doing so is becoming increasingly clear, says Govindarajan. “When you implement AI systems and do it well, the cost of handling cases goes down and the speed of resolving them goes up. And that generates value for everyone.”

This content was produced by Insights, the custom content arm of MIT Technology Review. It was not written by MIT Technology Review’s editorial staff.

Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

What’s next for generative video

OpenAI's Sora has raised the bar for AI moviemaking. Here are four things to bear in mind as we wrap our heads around what's coming.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at with a list of newsletters you’d like to receive.