Article

What is a chatbot & how do they work? The ultimate guide

The number of tickets handled by AI-powered chatbots increased by over 50 percent in 2021. But what, exactly, is a chatbot and how do chatbots work?

By Hannah Wren, Content Marketing Associate

Published October 22, 2020
Last updated May 19, 2022

Messaging support has become a go-to for customers, with tickets jumping 370 percent over WhatsApp alone, according to our CX Trends Report. As messaging rates have risen, so has the use of AI bots. The number of tickets handled by automations and AI-powered chatbots year over year increased by over 50 percent from 2020 to 2021, with customers using chatbots to do everything from checking if they have COVID-19 symptoms to finding the perfect lipstick color.

But what, exactly, is a chatbot? How do chatbots work? And will they steal customer service representatives’ jobs? We'll answer all your chatbot questions.



What is a chatbot?

A chatbot is a type of conversational AI that enables businesses to put a layer of automation or self-service in front of customers in a friendly and familiar way. And with companies increasingly adding messaging channels to provide faster resolutions and always-on support, bots have quickly become a key component of any messaging strategy. They can be deployed over any messaging app or channel and ensure customers get instant responses when an agent is busy helping other customers—or watching Bridgerton.

How do chatbots work?

Bots use predefined conversation flows, natural language processing, and/or machine learning to answer questions and guide customers through different scenarios, such as login issues, payment problems, or booking instructions–to name a few. AI bots can also learn from each interaction and adjust their actions to provide better support.

Chatbots work best with straightforward, frequently-asked questions. Unless their underlying technology is especially sophisticated, bots typically can’t handle difficult, multi-part questions like a support agent can.

Types of chatbots

Read on to learn the three main types of chatbots.

Chatbots that recommend knowledge base articles

Service has moved from calling at the first sign of trouble to adopting a “Google it” mantra. Research tells us that customers want to resolve as many issues as possible with a company’s online resources and prefer to resolve issues independently via self-service.
Chatbots can learn based on content in your knowledge base and recommend the right help articles to answer FAQs on any page of your website, inside your mobile app, over email, or via messaging channels. Since bots provide one-to-one support, they make the effort of searching for answers easier and more personalized than using a generic search engine. The business benefit of a bot that works alongside your knowledge base is that it uses your existing resources to deflect questions you already have answers to for the fastest time to value, which makes it easy to get started.

Decision tree chatbots

Decision tree bots enable you to design customized conversation flows that direct customers to quick answers, suggest knowledge base articles, and include points for handoffs to a live agent. Companies that are just getting started with a decision tree bot should look for a builder tool, like Zendesk’s own, that doesn’t require coding and provides a visual interface to help them create decision trees.

Task-specific bots

These chatbots are designed to help customers with a specific task and are typically highly specialized. They require more resources and a bigger budget because they need more comprehensive training and deeper natural language processing. Task-specific bots understand many different types of questions, and with access to the right customer data, can deliver personalized responses. For instance, a recruiting company might use this kind of bot for automating the first step of the recruiting process by helping potential employees submit their applications. Zendesk makes it easy to integrate third-party task-specific chatbots into your support system for seamless bot-human handoffs and more personalized conversations.

Why are chatbots important?

The benefits of chatbots go beyond “increasing efficiency” and “cutting costs”—those are table stakes. Bots are at their most powerful when humans can work in tandem with them to solve key business challenges.

Convenient one-on-one service, 24/7

Being constantly connected has increased customers’ desire for instant support. Customers today expect help as soon as they need it on channels convenient for them. Over 40 percent of customers think 24/7 real-time support is a top factor of good service, according to our Trends Report.

In deploying a chatbot across customers’ preferred channels, businesses ensure customers get seamless, always-on support. If Sally’s sushi delivery is late, she can text a chatbot and get an update on her California roll in real-time. If Rachel lost her credit card, a virtual assistant can help her freeze it, so she doesn’t have to worry about mysterious charges.

Deliver better customer service with a chatbot

Try our chatbot software for free.

Businesses can scale quickly

In our trends report, we found that 42 percent of customer service leaders expect customer requests to grow, yet only 36 percent can expand headcount. This gap represents a sweet spot where a chatbot can help. Chatbots can act like extra support reps, triaging simple questions and basic requests.

Consider Spartan Race, which deployed Answer Bot to help its small team of agents tackle spikes in customer requests during races. Spartan Race has seen a 9.5 percent decrease in chat volume, extending its team’s live chat availability by three hours every day.

France’s national rail carrier, SNCF, needed to provide quick support to on-the-go passengers using its mobile app. But it couldn’t hire another team of agents to deal with the influx of requests. Mindsay's Zendesk integration helped SNCF take the pressure off its overwhelmed agents. They deployed a chatbot to help find travel itineraries, provide departure information, and send alerts. And the integration led to a 50 percent reduction in incoming support tickets.

Chatbots give support teams the ability to scale without having to hire more staff.

More opportunities for conversion

Customer service bots can boost conversions with smarter self-service.

A chatbot can enable customers to self-serve outside of a help center, like on a checkout page, with knowledge tailored to their context.

  • Dollar Shave Club uses Answer Bot to connect visitors to help center articles and answer questions—before a customer abandons their cart.
  • Freshly deflects around 2,200 tickets each week. Its chatbot collects website visitors’ email addresses before they ask a question, which captures both context for agents and leads for marketers.
  • Wavy uses a chatbot to help with prospecting and lead generation. It's bot is the reason why Wavy’s conversion rate is at +25 percent.

Chatbots can also convert customers within the message chat by providing ecommerce opportunities for immediate action. Messaging types like carousels, forms, and picklists let customers book a hotel reservation or purchase a pair of shoes—all within the chat.

Chatbots can also automate cross-sell and upsell activities. With the right context, a bot can check if a consumer is eligible for a discount on a hotel room with a view or ask if he wants a pair of socks to match his new Nikes. In fact, 50 percent of companies say they use AI to make recommendations based on purchase or search history.

what is a chatbot

Best practices for getting started with chatbots

Ready to incorporate a chatbot into your customer service strategy? Here are some best practices to guide the way.

1. Make it clear to users when they’re interacting with a bot

We already know that customers are open to, and even prefer, to resolve their issues with a chatbot for simple queries. A study by Goldsmiths University and Mindshare, even found that one in four people are more willing to trust customer service chatbots with their sensitive information than a human.

But that same study also revealed that 48 percent of people feel it’s "creepy" if a chatbot pretends to be human. It’s always important to disclose to users when they’re interacting with a bot. Customers are open to communicating with bots, but they don’t want businesses to deceive them.

2. Always include an option to reach a live agent

Zendesk’s CX Trends Report found that the topmost frustrating chatbot experiences are when customers:

  • Can’t get an accurate answer
  • Have to start over when they get a human agent
  • Can’t get through to a human

In fact, 46 percent of customers get frustrated that they don’t have a choice in human vs. bot at the start of service. Chatbots aren’t meant to resolve every issue, so it’s crucial to include an easy option to reach a live agent. It’s also important to ensure that your bot can pass on context and conversation history, so agents have all the details they need, and customers don’t have to repeat themselves. (Nearly half of customers believe AI should keep people from having to repeat things.)

3. Let bots handle simple issues

Here’s where your chatbots should play a starring role, according to customers:

  • Providing links and information that can help resolve their problems
  • Offering quicker, more convenient replies for simple issues
  • Delivering answers and responses outside of normal business hours

But when it comes to filing a complaint or asking for technical support, 40 percent of customers prefer to interact with a human agent. Customers prefer bots for simple issues but still want the option to speak to a human for more sensitive and complex queries. Chatbots work best when their task is clear and simple. A good starting point is to look at the one-touch tickets that your agents see frequently. These are good issues to hand over to a chatbot.

4. Incorporate bots into your self-service strategy

Eighty-one percent of customers attempt to take care of matters themselves before reaching out to an agent, according to Harvard Business Review. What’s more, mid-to-large-sized businesses saw nearly a 40 percent increase in knowledge base views from customers during the pandemic. Customers prefer to resolve issues via a knowledge base or self-service portal because it’s simple and convenient.

A chatbot can help your customers self-serve more efficiently by highlighting your FAQs outside your knowledge base, such as on your checkout page or website homepage. You can also incorporate a chatbot into your knowledge base to personalize the experience. A bot can help customers find the right articles for their issue on a one-to-one level.

5. Use chatbots to collect key customer information upfront

Chatbots can collect key customer information upfront, such as order number, email address, or location, to ensure agents have all the details they need to personalize conversations and resolve the issue more efficiently. Chatbots can use this information to route customers to the right agent for the problem based on things like language, skill, or account type. This also makes chatbots a helpful lead generation tool, as they can collect prospect contact information and route them to the right salesperson for their inquiry.

6. Use data to improve your bot

Getting your bot up and running isn’t the end of your chatbot strategy. You also need to track metrics to improve and get the greatest value out of the tool. Here are a few metrics to watch:

  • Use rate
  • This is the percentage of customers who choose to engage with a bot when prompted or given the option. It will help you determine if customers like using bots, if they respond to nudges to use bots, and what channels they look to bots for answers.

  • Volume
  • This refers to the number of questions answered by your bot, which will allow you to understand how cost-efficient and time-efficient your bots are.

  • Bot response failure rate
  • This is the rate of questions your chatbots can’t respond to because they don’t understand them. This will help you get an idea of what people are looking for when using self-service and see where you can improve your bot flows.
    Most answered questions: Tracking this metric can highlight common areas of frustration and opportunities for improving customer experience.

  • Referral data
  • This metric indicates how often a bot transfers an interaction to an agent. The referral data can help you determine how bots integrate with agents’ workflows and give you insight into how customers approach complex support questions.
    Customer-reported bot success metrics: Ask your customers to rate their service after using your bots so you can find out whether or not your bots provide good service.

What are examples of chatbots?

Read on for some chatbot examples to give you inspiration for the kind of bot your company might want to implement.

1. Fintiba

Fintiba offers online solutions for people who want to work or study in Germany. Human agents are critical for resolving high-empathy issues, like when a customer’s visa gets declined. Chatbots help streamline this process. Every conversation goes through Fintiba's virtual agent first before going to a human agent. That takes the pressure off the support team, so they have the time they need to solve problems chatbots can't handle. For example, when customers want to change their phone number, they complete a form and send a selfie inside the chat. An agent can then jump in with the process already started.

Fintiba’s chatbot solution, Solvemate, integrates with its customer service software. The bot is able to route chats with context and conversation history to live agents. When chatbots enhance the agent experience—instead of replacing it—it’s a better experience for everyone.

2. Discord

Discord is a free service that makes it easy for groups to communicate over voice, video, and text. During the pandemic, the number of unique visitors to its customer service site doubled, and pageviews increased by 70 percent as a flood of new users began using the service. The company was easily able to shoulder this unprecedented surge in users and support queries thanks to an early decision to make big investments in automation, deflection, and AI. These tools enabled Discord to help customers quickly and easily get the answers they needed, while freeing agents to focus on more complex, high-priority conversations.

“Automation and scaling is always at the forefront of our minds, and we’re always looking for ways to do more with less,” said Danny Duong, Discord’s Director of Customer Experience and Community Management. “We’re focused on getting customers the help they need as quickly as possible, while never sacrificing the human touch.”

To date, Discord has made more than one million suggestions to customers through its chatbot, and resolved more than 100,000 tickets. In 2020 alone, the team has also used AI to resolve approximately 14 percent of all email inquiries. In total, the team has an overall deflection rate of 10 percent, meaning that one in 10 of its customers have their problems solved without ever engaging with an agent.

3. Carousell

Carousell is one of the world’s largest and fastest growing marketplaces across Southeast Asia, Taiwan, and Hong Kong. A majority of the questions Carousell’s customer-facing team receives are straightforward and transactional, like how to create an account or how to accept payments. This is where Carousell’s chatbot comes into play, which serves customers in English, Chinese, and Indonesian. It guides customers to relevant help center articles, which in turn drives ticket deflection.

This investment in self-service tools has been a boon for Carousell’s support team. The help center receives more than a million views a month, and its chatbot responses have a 22 percent click-through rate. As a result, the company is able to successfully deflect 24 percent of inbound general enquiries, and 7 percent of all incoming requests. And since the team spends less time on low-level requests, it has been able to dramatically improve its response time—77 percent of the 30,000 monthly tickets are answered in less than 24 hours.

4. LendingClub

LendingClub connects U.S. borrowers and investors through an online marketplace that offers ethical and easy ways to access credit. To help surface help center articles and anticipate customer questions, its support team adds some of the most popular articles to automated replies with LendingClub’s chatbot.

The team has seen valuable cost savings when its chatbot handles a ticket, with the potential for further savings month over month, and a resolution rate of 12 percent. Its chatbot has been helpful in addressing quick questions for customers. “We’re happy to see customers find the answers they need, when they need it. [Our chatbot] points them in the right direction to find solutions independently,” said Alina Doyle, a LendingClub Specialist.

Limitations and drawbacks of full chatbot automation

Chatbots weren’t built to take over customer support. Rather, they were designed to serve as intermediaries, to keep customer support accessible when agents are off the clock and free them to handle tricky requests.

Chatbots are most successful when they work together with human agents. Bots aren’t meant to solve every issue: 56 percent of customers say bots are helpful for simple requests. For example, bots do a good job offering basic troubleshooting, but the number of variables and the complexity of deeply technical troubleshooting makes it better suited to a well-trained agent.

Bots can help your brand scale 1:1 communication, but agents bring empathy to the table, and efficient handoffs between bots and their human counterparts mean customers don’t ever have to repeat themselves.

How have chatbots evolved?

Chatbots might seem like a relatively new concept since companies are still figuring out how they fit into their overall business strategies. But bots have been around for decades. The first chatbot was developed by Joseph Weizenbaum, a scientist at MIT's AI lab, in 1966—that’s even before the first personal computer. It was named ELIZA. The next chatbot was PARRY, a natural language program built in 1972. Another early chatbot is Dr. Sbaitso. Dr. Sbaitso was constructed in 1992 and is considered one of the first AI chatbots.

But if chatbots have been around for decades, why are we starting to see a boom now?

Chatbots have come a long way since the 1990s, partly due to advances in AI and machine learning. But most recently, the rise of messaging has made bots an essential part of any customer service and engagement strategy. Today, nearly all the top messaging platforms offer APIs so businesses can offer seamless messaging experiences with a bot.

What is the future of chatbots?

Here are a few key points to consider when it comes to the future of chatbots.

  • The rise of messaging
  • There are already 300,000 active Facebook Messenger chatbots, and messaging will only become a more critical customer engagement channel.
    Businesses need to engage over the channels that matter most to their customers. With support that’s fast, personal, convenient, and secure, it’s no surprise that messaging has seen an upswell of adoption by both customers and businesses.

  • Chatbots will see an adoption boost
  • AI isn’t new technology at the workplace anymore; however, many companies are still learning how it fits in their overall business strategies. But as AI continues to improve and chatbots become more sophisticated, more businesses will adopt it. In fact, the Chatbots Market was worth USD 1274.428 million in 2018 and is projected to reach USD 7591.82 million by 2024, registering a CAGR of 34.75 percent over 2019–2024, according to Research And Markets.

  • Human agents will remain important
  • Even as chatbot technology gets smarter, bots shouldn’t replace human agents but help them perform their jobs even better. Businesses get the greatest value from AI when using augmented intelligence—human intelligence and machine intelligence combined.

What chatbot platform is right for you?

Today there are many chatbot examples. Chatbots can be deployed across any messaging channel. But chatbots are relatively new to customer service, and companies are still figuring out how they fit within their support strategy. That can make it tough to know how to find the right chatbot solution for your business. Answering these questions will help you find a solution that best fits your support team’s needs.

1. What problem are you looking to solve—and what resources do you need to solve it?

For starters, you need to decide what use cases to automate. Decide based on the problem you need to solve and what resources you have to solve it.

Some companies may need a bot to deliver help center articles across a variety of channels and capture basic customer context. Other companies may need bots for personalized requests, like telling a customer how much data her iPhone used this month or recommending a new plan based on usage.

Offering personalized service with a chatbot requires more resources and a bigger budget. You’ll need a chatbot solution that integrates with customer service software and other relevant systems.

2. Is your chatbot flexible enough to work across different channels?

Customers expect to get support over their preferred touchpoints—whether they’re interacting with a human or a bot. And 40 percent of companies are already using AI to engage with customers via their preferred contact methods.

Research tells us customers want to interact with brands on channels they use with friends and family. Messaging and social media channels, like Facebook Messenger, WhatsApp, Twitter Direct Message, LINE, Apple Business Chat (which integrates with iMessage), and SMS lend themselves to more convenient conversational experiences. For instance, Samsung Australia created a Twitter chatbot to give customers personalized TV recommendations.

It's important for your chatbot to work across all these different channels. With a chatbot platform that's flexible, you can connect your bot to any channel, without any heavy lifting. For example, Zendesk lets a business build once and deploy anywhere. In other words, you can use the best version of a rich bot experience across all your channels, even those with no native bot support.

3. What level of context will your chatbot need?

44 percent of customers say it is most frustrating when they have to start all over with a human agent after interacting with a bot. More context leads to better chatbots—and more personalized conversations.

Freelance platform Upwork’s chatbot displays proactive CTAs tailored to what a user is trying to accomplish, like what help center article they’re viewing.

Upwork’s bot also uses contextual metadata, like a user’s name for a personalized greeting. It knows if a user is a client or a freelancer, tailoring quick replies accordingly. It also integrates with our Support Suite, so agents have the context they need to handle every escalated chat.

Beyond passing on relevant information to agents, bots can also pass on context to a CRM or other software. This enables things like:

  • Understanding that Rose has a necklace in her cart and sending a message to a marketing automation tool, so she receives better-targeted email offers
  • Knowing that IT buyer Bob signed up for a demo and qualifying him as a lead in a sales CRM

Bots can read context coming from a conversation, too. With sentiment analysis, a virtual agent can understand when a customer is frustrated and react accordingly.

4. How will you manage conversations between chatbots and agents?

Businesses need tools to both deploy chatbot conversations on the front end and manage them on the back end. This ensures agents can understand the intent behind every conversation and streamline hand-offs between agents and chatbots.

With the help of triggers, automation, and workflows, support teams can centrally define engagement rules and track, manage, and prioritize chatbot interactions at scale. In fact, 39 percent of companies already use AI to prioritize customers based on their status or account type. This opens up possibilities, like automatically assigning:

  • A high priority to VIP customers, so a bot can route them to a live salesperson for help—with conversation history
  • A repeat dissatisfied customer to a specialized customer support team by looking at context, sentiment, and intent

To effectively control bot interactions, a business will need to integrate its chatbot solution with its customer service software.

The agent workspace in Zendesk provides agents with a real-time, conversation-focused interface to seamlessly manage conversations between agents and bots.

The best bot experiences are built with Zendesk

Zendesk makes it easy to enhance your customer support experience with a chatbot. Answer Bot can leverage your existing help center resources to guide customers to a resolution via self-service and collect customer context. And if you want a little more control, our click-to-build flow creator enables you to create rich, customized bot conversations without writing code. Most importantly, our award-winning support platform provides teams with a real-time, conversation-focused interface to seamlessly track and manage conversations between agents and bots. It also integrates with all the systems your team depends on, including third-party bots.

Deliver better customer service with a chatbot

Try our chatbot software for free.

Deliver better customer service with a chatbot

Try our chatbot software for free.

Start free trial