What is conversational AI and how does it work?
Conversational AI is a set of technologies that can recognize and respond to speech and text inputs. Learn why conversational AI is the future of customer service.
Published March 2, 2021
Last updated March 22, 2022
If you think consumers are bot-resistant, think again. According to Markets and Markets research, customer interactions with automated chatbots are steadily increasing, with the global conversational AI market expected to grow from $6.8 billion in 2021 to $18.4 billion by 2026. And as it turns out, a majority of customers actually prefer using bots for simple tasks like changing an address.
A majority of customers actually prefer using bots for simple tasks like changing an address.
As messaging becomes increasingly popular, businesses should learn how to best leverage conversational AI for customer service. That means understanding how conversational AI works, how it benefits customers and agents, when (and when not) to use it, and how to best optimize it for CX.
Luckily, it’s not that complicated. Conversational AI can be surprisingly easy to adopt and implement when supported by a clear strategy and careful planning.
How does conversational AI work?
To understand how conversational AI works, there are two key concepts that must be defined: natural language processing and machine learning.
Natural language processing (NLP), sometimes referred to as natural language understanding (NLU), allows computers to comprehend speech and text so they can communicate with humans. NLP analyzes speech and writing patterns and tries to determine what is actually being said in order to interpret the customer’s intent. It learns to account for incorrect grammar, typos, differences in intonation and syllable emphasis, accents, and so on.
Once customer intent is clear, conversational AI technology uses machine learning to form a response. Machine learning is exactly what it sounds like—it’s the process of machines (computers) using algorithms to parse data, learn from that data, and then apply what they’ve learned to deliver relevant answers.
Every time these workflows occur, conversational AI technology becomes more sophisticated by identifying which responses provide customers with the best results.
Conversational AI use cases
When most people hear “conversational AI,” they think of chatbots communicating with customers. But chatbots and conversational AI aren’t synonymous. While a bot is a type of conversational AI technology, it’s not the only type. Voice tools fall under this category, too.
Chatbots are computer programs designed to simulate conversations with humans. They empower users to get the answers or help they need quickly and at any time of day, usually through social media messengers or chat applications built into a website or mobile app.
Rule-based chatbots work like a flowchart with humans mapping out conversations based on predefined rules. They’re dependable, they’re easy to program, and they integrate into your preferred customer support channel. However, rule-based chatbots’ lack of AI doesn’t allow for much personalization or flexibility. They are typically used to automate the answering of simple FAQs, for example.
AI chatbots, on the other hand, can independently lead a conversation. NLP enables an AI bot to generate relevant answers by analyzing the conversation and trying to understand customer intent. Based on the intent, machine learning then formulates a response.
AI chatbots are more difficult to set up than rule-based chatbots but are much more versatile and able to answer more complex queries. Ecommerce websites often use AI chatbots to better understand the shopper’s intent when providing recommendations.
Hybrid chatbots that are both rule-based and AI-equipped also exist. A chatbot on a healthcare site, for example, might use AI technology to understand a patient’s issue and rule-based technology to offer medicine instructions.
Zendesk’s Answer Bot is a good example of an advanced hybrid bot. Customer service providers can use it to quickly answer common questions and also to identify when to bring in an agent to help resolve an issue.
Voice assistants are AI applications that are programmed to understand voice commands and complete tasks for the user based on those commands. Starting with speech recognition, human speech is then converted into machine-readable text that can be processed in the same way chatbots process data. While chatbots are typically used on social media platforms and websites, voice assistants are often found in search engines, smart speakers, and operating systems. Alexa and Siri are common examples of this technology.
The main advantage of voice assistants is that customers can use them hands-free, which makes them popular options for the physically disabled. And just like chatbots, voice assistants can be programmed to recognize a wide variety of languages.
Interactive voice assistants (IVAs)
Interactive voice assistants (IVAs) are automated phone systems where a user can express their intentions through both voice and keypad input and receive verbal answers.
They function as a hybrid of chatbots and standard voice assistants, combining mapped-out conversations with a verbal interface. If you’ve ever wanted to request information from your bank via phone or wanted to make an inquiry regarding a utility bill, you’ve probably used an IVA.
IVAs make it very easy for customers to connect to the right department when making an inquiry. They also allow customers to leave voice messages or access a pre-created company FAQ over the phone if they call at a time when human agents are unavailable.
Conversational AI vs. traditional chatbots
Chatbots are a type of conversational AI. But what does it mean to be conversational? Conversational chatbots are one part of a company’s larger conversational customer service strategy—the ability to offer fast, personalized, uninterrupted service across web, mobile, and social apps. Instead of siloed chats that start and stop each time a customer reaches out (or switches channels), each interaction becomes part of a larger conversation that carries over a lifetime of interactions with the company. The result is a seamless experience for customers and agents alike. Chatbots that are truly conversational work across a variety of channels, including messaging apps like Facebook Messenger and WhatsApp. They also offer seamless human-to-agent handoffs so customers don’t have to repeat themselves when the conversation is passed to a live agent.
Components of conversational AI
AI is becoming incredibly popular, but businesses don’t always implement the technology successfully. According to a Pactera Technologies report, 85 percent of AI projects end up failing. To increase your chances of success, make sure you’re laying the proper groundwork for implementing conversational AI at your company.
You won’t know if your conversational AI initiative is paying off unless you know what you’re hoping to gain by using the technology. Be specific about your objectives and the problems you want to solve, so you can gauge which conversational AI technology is the best fit for your company.
For example, say your main pain point is that your support agents are wasting time answering basic questions and don’t have enough bandwidth to handle complex customer inquiries. What types of conversational AI would best solve this issue? Perhaps it’s a combination of IVAs that deliver automated answers to common questions and rule-based chatbots that can address FAQs.
Specify what customer service goals and key performance indicators (KPIs) you want to achieve before moving forward with implementation. That way, you can measure the success of your conversational AI strategy once it’s in place.
You might have a good idea of the type of conversational AI your customers will appreciate, but how will you know for sure? Thankfully, there are ways to test conversational AI technologies without having to immediately invest in their implementation.
For example, a tool like Botmock develops and delivers chatbot prototypes to your customers, enabling you to A/B test different AI solutions and how customers interact with them. Launching prototypes helps you confirm how effective they are before you put time and money into developing them.
You can also use the “Wizard of Oz experiment” method: Have a real agent roleplay as a virtual agent with customers via chat or social media. A simple messenger app and a list of responses within a flow chart are enough to test your assumptions.
This initial testing will give you a sense of your infrastructure needs, whether that’s back-end integrations or revisions to rule-based chatbot scripts.
Your existing infrastructure
You know what conversational AI technology you would like to use based on your goals and test results. Now, it’s time to investigate your current communication channels to determine what tools you’ll require and whether you’ll need to make budget adjustments.
What conversational AI platform investments have you already made (if any)? Leverage any existing architecture to deliver value and cut costs. For example, if you already have a messenger app on your site, build a chatbot that can integrate with it instead of developing a similar tool from scratch.
Remember to think ahead and consider the scalability of your infrastructure as you develop your strategy.
The next step is securing support for the initiative. When pitching your idea to stakeholders, make sure your arguments are closely aligned to top business objectives. Focus on the importance of:
- Understanding customer needs: Demonstrate how conversational AI tools learn about customer needs, behaviors, and preferences—and explain how that will improve CX.
- Improving agent satisfaction: Emphasize the positive impact AI can have on your agents. Spending less time on repetitive, manual tasks increases both productivity and employee satisfaction, for example.
- Getting a good return on investment: Decision-makers will want clear ROI projections. Use resources like Dataiku and Nexocode to learn how to calculate, frame, and pitch the ROI metrics of AI projects.
The success of your conversational AI initiative hinges on the support it receives across your organization. According to Deloitte’s State of AI report, AI projects cannot succeed if company leaders aren’t setting core, overarching business strategies to achieve the vision.
Depending on your conversational AI objectives and tools, you might need to build out a dedicated team. Whether you decide to hire internally or externally, there are a few roles worth considering:
- User experience (UX) designer: This person will map out conversation paths and design necessary UX components like quick-reply buttons. Ideally, they have a strong background in copywriting as well.
- Conversational analyst: This is a cross-functional role that works with developers, data scientists, and UX designers to create testable customer conversations. The analyst will also continually monitor and optimize conversational AI interactions in order to improve the customer experience.
While you certainly won’t need a complete overhaul of your IT team, you may want to create new roles as you implement new technologies.
What are the main challenges in conversational AI?
Conversational AI isn’t meant to fool someone into thinking they’re talking to a real person. Your company should be upfront with customers about when they’re interacting with a bot versus a human. And if the customer wants to speak to a human agent at any point, your business should allow them to make that decision for themselves.
“Remember, the goal is always to reduce effort and increase satisfaction,” Chethuan says. “So, the ‘talk to an agent’ option shouldn’t be hidden.”
But conversational AI should still use language that customers are comfortable with. “Even if a person knows they’re talking to a bot, we should make it the most natural and friendly experience possible,” Chethuan adds. “The AI should be configured to speak in the same terminology as our customers.”
Even if a person knows they’re talking to a bot, we should make it the most natural and friendly experience possible.Giovanna Chethuan, customer success executive at Zendesk
Beyond crafting a natural voice, use customer data to make the messaging experience feel more personalized for the user. If your AI is empowered with the right customer data, a chatbot can:
- Address a customer by name
- Refer to the customer’s status
- Know what products and services the customer has purchased
- See the path of the customer’s journey
With insight into a customer’s previous interactions, bots can provide relevant and helpful support. “If a customer just read an article from our help center, the bot should take that into account and suggest different content to resolve their issue,” Chethuan says.
As long as a bot is genuinely helpful and provides great service, customers won’t begrudge it. Sam Chandler, Zendesk’s director of startup success, encourages businesses to worry less about how “human” their conversational AI sounds and to instead “embrace their inner bot” to create a more authentic experience.
“The best chatbot strategies are the ones that are transparent and authentic,” Chandler explains. “They incorporate their branding into the bot and create an opportunity to bring their customers further into the fold.”
She cites the example of BarkBox, an online pet-products retailer that set up a “dog bot” on Twitter for the holidays. The canine bot was able to “lend a paw” and help eliminate one-touch questions with simple answers during a busy season. As a bonus, it also doubled as an adorable marketing campaign that delighted BarkBox’s dog-loving customers.
Benefits of conversational AI
Why is conversational AI important? Two words: Customer self-service. With the use of chatbots, your messaging channels can provide quick, convenient 24/7 customer support. A tool like Zendesk’s Answer Bot can respond to customers’ simple, low-priority questions and lead them to a speedy resolution. Each support ticket that’s resolved by a “human-like” conversational AI interaction is one more ticket your agents don’t need to worry about.
“Conversational AI is primarily used to drive customer satisfaction, but it ends up improving agent satisfaction as well,” explains Paul Lalonde, a Zendesk product expert. “One of the major reasons for call center turnover is that agents find tasks boring and repetitive. But when you have bots working towards solving customer problems, agents don’t have to worry about doing the more mind-numbing, repetitive tasks. They can just get involved when more critical thinking is needed.”
Conversational AI is primarily used to drive customer satisfaction, but it ends up improving agent satisfaction as well.Paul Lalonde, Zendesk product expert
When an automated messaging conversation does require a human touch, a chatbot can transfer the customer to a live agent. The bot will also pass along the information that the customer has already provided, such as their name and issue type.
“Agents get all that context right away, so they never have to ask customers to repeat themselves,” Lalonde adds. “Everybody hates having to repeat themselves, but that goes away in a world where you have conversational AI embedded into the customer experience.”
Conversational AI can increase customer service productivity while cutting costs. According to Accenture, AI technologies are projected to increase labor productivity by up to 40 percent by 2035. Companies can deliver personalized experiences more easily, too.
- Boost productivity: With conversational AI, companies can offer automated support 24/7. Making it easy and convenient for buyers to solve their own problems lowers customer effort, decreases the number of support tickets, and increases satisfaction. When conversational AI and human agents work in unison, customers get help faster, and agents gain the bandwidth to tackle more complex issues.
- Reduce operating costs: Conversational AI can also handle requests at a higher volume than human agents while still surfacing accurate information to customers—at no extra cost. A Juniper Research study predicts that chatbots will be responsible for savings of over $8 billion per year by 2022.
- Provide personalized customer service: Consumers are more likely to purchase from companies that deliver personalized experiences. With conversational AI, you can tailor interactions based on each customer’s account information, actions, behavior, and more. The technology can also relay relevant information when there’s a bot-to-human hand-off, giving agents the context they need to provide better support.
Is conversational AI the future?
Seven out of 10 consumers now strongly agree that AI is good for society while 66 percent of customers give AI a thumbs up for making their lives easier. And 69 of customers say they’re willing to interact with a bot on simple issues, a 23 percent increase from the previous year.
Companies often view bots as a cost-saving measure—and they certainly can be. But ideally, conversational AI will enhance the capabilities of your support staff, not replace them.
“Don’t think of chatbots as a substitute for humans,” Lalonde cautions. “One of the great things about using conversational AI for customer service is the beauty of having both bots and humans working together towards solving customer problems.”
Conversational AI makes it easier and faster for customers to get answers to simple questions. At the same time, support agents have fewer tickets to resolve, freeing them up to address the complex questions that chatbots and virtual assistants can’t handle. When companies combine the strengths of AI tools and humans, it leads to a better customer experience—and a better bottom line.