If you’ve contacted a company looking for customer support recently, chances are you were prompted with a digital interaction first. Companies are increasingly relying on digital-first interactions like chatbots and AI agents to help them scale their customer support function.
In my time working at a leading chatbot company, I’ve seen companies build great digital interactions that felt personalized and human-like. But I’ve also seen plenty of not-so-great chatbots that were stale and frustrating to engage with.
Using AI in your digital customer interactions is the best (and dare I say “easiest”) way to make those interactions feel more human, which in turn makes engaging with them a better experience for your customers, offering up responses that are tailored to your customer’s unique needs.
In this article, I’ll discuss:
AI Agents Versus Chatbots
What is a chatbot?
A chatbot is a software application that uses artificial intelligence to simulate human-like conversation.
The goal of a chatbot is to help customers by processing requests, generating answers, and providing general support. You may hear the term “conversational AI” floating around. This term essentially summarizes the experience of using AI to create communication that feels smooth and conversational in tools like chatbots, voice assistants, email writing tools, and more.
Chatbots typically use natural language processing models (NLP) to create that human-like communication.
Here are some of our favorites.
What is an AI agent?
An AI agent is a software program that uses artificial intelligence to perform tasks and make decisions on behalf of a user.
AI agents use advanced large language models (LLMs) to understand and respond to user inputs, and they’re able to handle multi-step processes.
AI agents go beyond just gathering information needed to make a decision or solve a problem — they can interact with external sources to obtain the required data and then autonomously complete the task for you. (Example: Rebooking your hotel room reservation for another day, or generating a report).
The Differences Between AI Chatbots and AI Agents
When you hear AI agents vs. chatbots, you may be tempted to use them interchangeably. While both utilize AI to help businesses or users complete a task, they differ in their level of knowledge and their ability to execute.
- AI chatbots are built for conversations. They help users find information that they need to complete a task by using two way communication. This makes them great at handling commonly asked questions or helping with simple requests.
- AI Agents are built to handle complex, multi-step requests that are not predefined. Agents have the capacity to actually execute actions in order to complete the request.
I spoke to Ben Gardner, VP of customer care at AvidXchange, about how he explains the differences between AI chatbots and AI agents. Here’s what he said.
“Chatbots are simply conversational and are providing answers or instructions on how to do things yourself,” he said.
In contrast, Gardner said that, “an agent can do more than just have a conversation and provide information … [it] can look up data in other systems, perform actions, and do things that typically a person could do.”
“Both options are useful but they have different focuses,” he concluded.
Here are a few key ways I’ve learned that they differ.
Interacting With Customers and Helping With Requests
What AI Chatbots Can Do
At their core, AI chatbots are designed to offer support within the context of a conversation by providing information or offering next steps for a user to complete their request.
An AI chatbot is more advanced than a traditional chatbot in that it can understand language outside of pre-programmed commands and offer AI-generated responses based on data it has been trained on. However, an AI chatbot needs to ingest large quantities of information (like conversational data) for it to analyze and use to formulate responses.
- Since AI Chatbots are trained on things like your past conversations, your knowledge base articles, and your content, they are really good at performing basic tasks like solving frequently asked questions and troubleshooting.
- AI chatbots are great for helping customers self-serve. When connected to your knowledge base, AI chatbots can surface help articles and content based on keywords provided within the conversation.
- Chatbots can obtain information (like asking the visitor for their name, order number, etc.) and pass that along to other connected systems like your CRM or ticketing software. They can sometimes even use APIs to pull information back into the chat.
However, when it comes to multi-step interactions, AI chatbots have limited capabilities.
What AI Agents Can Do
AI agents function more like a real customer service rep would operate.
Since AI Agents are built on LLMs, they can learn and adapt in real time while also understanding context and sentiment. So instead of solely relying on keywords or pattern matching to provide a resolution, the agent can understand more nuanced questions and even help solve multi-step interactions.
This not only makes the conversation feel more natural, but it also allows for the agent to solve more complex problems.
- AI agents shine at managing complex interactions and multi-step requests. An AI agent can do things like pull up your order status, recognize that your order is delayed in shipment, and automatically offer you an incentive or solution to make up for that delay.
- Some AI agents can even use things like detected sentiment and customer intent to intelligently decide next steps, like offering an incentive to a dissatisfied customer to avoid cancellation or suggesting an upgrade to a long-time VIP customer. Then the agent can take it a step further by actually processing those tasks, thus taking work off of a human rep’s plate.
Example of How AI Chatbots and AI Agents Interact Differently With Customers
Here’s an example I like to use to pull it all together.
Imagine you’re visiting a shoe store in person. Here are your different interactions.
Chatbot interaction: The AI chatbot is similar to the greeter at the front door who says hello, asks if they can help you find something, and then points you in the right direction. While they can detect things like your intent and your sentiment, if you asked this person which brand you should buy or if a shoe runs true to size, they’d have to pull in another employee to help you.
Agent interaction: The AI agent is more like the shoe store employee who specializes in the type of shoes you’re looking for and can be consultative in their guidance. They can answer your questions about brands or functionality, pull sizes from the back for you to try on, suggest a pair of matching socks to go along with your new shoes, and then ring you up at the register to complete your transaction.
Conclusion: While the chatbot and agent both helped you on your journey to buy new shoes, their roles and responsibilities are limited to their level of knowledge and their ability to execute on certain tasks.
Learning and Adapting
How AI Chatbots Learn
AI chatbots typically require more manual training and review, and any findings that you uncover can also require manual updates to the decision trees and bot flows.
For example, if you uncover a question that the bot is struggling to answer, you’ll need to go in and review that customer question, see how the bot is responding, and train the model on a better response. You may even need to A/B test question and answer flows to see which performs better.
Since AI chatbots are constantly learning from conversations, the more conversations the bot is having and the more feedback that you give it, the more the bot adapts its future responses (and its suggested replies for your team).
Pro tip: Your support teams should also be training the model by giving it feedback on the conversations. Most chatbots will have a place for a human to do things like approve, deny, or edit a suggested response as well as rate a conversation that the bot had.
How AI Agents Learn
AI agents typically come with built-in models or algorithms that help the agent learn and adapt.
As with the chatbot, the more conversations your agent has, the more it is learning and changing to fit your customers’ needs.
Once an AI agent is trained on your datasets and connected to your tech stack, it can do things like look at past conversations or interactions that your human reps have had and offer similar responses.
As your company’s knowledge base, product offerings, or policies change, your AI Agent can automatically adapt its responses to reflect those changes, removing the need for manually updating tags or conversation flows.
Training an AI Agent still includes the responsibility of reviewing customer interactions and providing feedback, but the beauty of an agent is that it usually doesn’t require a manual adjustment (like adding a new response to multiple bot flows) in order for the bot to implement the changes necessary.
Additionally, many AI Agents will continually analyze data to find gaps and make suggestions for new or revamped flows. Some agents can even go beyond just suggesting changes and can actually write the new flows and implement them for you.
Level of Knowledge
Chatbot Knowledge
Your AI chatbot is only as intelligent as the information you feed it. Your bot’s knowledge domain will be limited to the information you’ve given it access to via site mapping, integrations, or APIs.
When possible, using federated search provides a better customer experience, but it requires more lift to set up with an AI chatbot.
To enable federated search in a standard AI chatbot, you may be looking at doing things like manual tagging and even requesting help from your engineering team to connect various data sources, training the bot, and defining the rules for relevant content types.
AI Agent Knowledge
In contrast, an AI agent not only scans your website, data, and integrations for answers, but it can also use broader language models, external resources, and historical context to inform the answers it provides.
But it goes beyond just pulling that variety of information — it can then combine those inputs to make a logical assumption, creating a more robust and nuanced response for customers.
Additionally, vertical AI agents are likely to increase in popularity due to the increasing need for industry- and domain-specific knowledge in AI workflows.
AI Agent Use Cases
You’d be hard pressed to find a company that isn’t trying to balance scaling their organization while also ensuring they deliver a great customer experience. AI agents excel at helping organizations achieve both of those goals simultaneously.
The market for AI agents in tech is expected to grow at least 45% over the next five years, with Boston Consulting Group predicting that businesses will transition to a model that includes having much smaller teams of humans working alongside an agent — specifically for complex disciplines like customer service, analytics, and software development.
Next, I’m going to look at a few ways AI agents can benefit different teams across an organization.
Use Case #1: Customer Service
Customer service and support teams continue to look to AI to help enhance the customer experience while decreasing ticket volume and response time. A recent study by Gartner found that 80% of companies are either using AI or planning to use AI in customer service by 2025.
AI agents can transform the support experience by not only handling Tier 1 support requests, but by helping customers fully resolve their own issues (which leads to quicker resolution times).
As we’ve discussed earlier, AI agents can revolutionize the customer experience by doing things that typically require multiple steps, systems, or human intervention, such as:
- Helping a customer add more seats to their software plan.
- Rebooking a customer’s hotel reservation for a new date, or upgrading them to a bigger room.
- Freezing or pausing a customer’s subscription upon request.
- Surfacing videos, dev docs, or training courses to help with a customer’s specific question.
- Analyzing customer requests and pain points to surface themes. This could help identify potential improvement opportunities across the company including trends in staffing, gaps in support documentation, and more.
Use Case #2: Ecommerce
Implementing AI agents can help retailers drive more revenue and enhance the customer experience. AI agents can also help retailers make buying decisions by providing real-time analysis on customer data and inventory.
AI Agents could help ecommerce teams with things such as:
- Creating personalized product recommendations for customers based on their previous purchasing history.
- Proactively suggesting a sizing guide to a customer when they’re shopping for an item that is commonly returned. This can help companies avoid losing money to things like “bracketing,” where customers buy multiple items because they’re unsure what size will work best, and they return what doesn’t fit.
- Analyzing historical data on items that are commonly purchased together and then creating proactive upsell opportunities for customers as they shop.
- By detecting sentiment or intent, the agent can offer appropriate next steps that proactively increase customer satisfaction. For example, if the agent recognizes that a customer’s order is delayed or lost in shipping, the agent can automatically offer a discount or ship out a replacement order.
Use Case #3: B2B Account Management/Customer Success
Ask any customer success manager what they wish AI could take off their plate, and chances are they’ll have an entire list. If you can find an agent that helps account teams spend more time managing relationships and less time on admin work, you’ll have a happier team!
Josh Schachter, the CEO of UpdateAI, told me that in his experience, “AI agents are redefining customer success and account management by automating busywork, scaling customer knowledge, and delivering white-glove service at scale.”
Similarly, Erin Gehant, an account manager at TrueLearn, said that their AI agent streamlines their internal processes and does things like “transcribing meetings and summarizing key takeaways, and then drafting follow-up emails based on meeting insights.”
Gehant says this has “helped boost efficiency in my workflow which gives me more time to focus on relationship building.”
Here are a few ways account teams can benefit from AI agents:
- Automatically surfacing at-risk customers based on a drop in sentiment, product adoption data, or health score.
- Suggesting (and then drafting) activities to engage customers based on customer sentiment, drop in MAU, or other signals of risk.
Example: An agent could notify a CSM that their top account has dropped in product adoption, then offer to draft an email inviting that customer to an upcoming webinar the customer could benefit from.
- Scan customer interactions like emails and meeting recordings for keywords or sentiment that may signal an upsell, then notify an Account Manager.
- Based on certain triggers, an agent could compile a deck or report for an upcoming QBR with the customer’s data.
- Automatically drafting a follow-up email after a meeting with a summary and/or key points that the CSM said they would follow up on.
Use Case #4: Marketing
When it comes to finding creative ways to use AI, I’d bet money that marketing teams already have a long list of clever use cases. From lead scoring to data analysis to content creation — the possibilities are endless.
Today, between 40-50% of marketers are already using some form of AI in their workflow and the majority of marketers surveyed expressed excitement about what AI can help them with in their jobs.
Here are just a few ways that AI agents are being used by marketing teams today:
- Predictive modeling. AI agents can forecast customer behavior and trends, which enables marketers to make changes and enhancements to their marketing strategies in a timely manner.
- Personalized content. AI agents can analyze customer data and create curated content and experiences for them. The agent can also surface missed opportunities for conversion, low-performing campaigns, and other themes in customer behavior that you can use in creating campaigns.
- Curating communication. As I talked about earlier, agents are powered on advanced language models, so having an agent draft your emails, website copy, webinar titles, and more gives you a great starting point. You’ll want to go in and review it to make sure it sounds human and matches your company’s brand voice, of course, but the more you use features like this, the more the agent learns and adapts to your brand voice.
- Surfacing compelling insights. If an AI agent has access to data like customer feedback, NPS scores, product adoption, and customer lifetime value, imagine what insights it could surface for you. If an agent can find correlations between things like buying a certain feature and quicker time to value realization, you can use those insights to target future customers.
While this list highlights just a few different teams that can benefit from AI agents, I have no doubt we’ll continue to see new and creative use cases for AI agents expand across entire organizations.
AI Chatbot Use Cases
When it comes to how consumers interact with a business, chatbots have become table stakes in today’s digital landscape. Customers not only expect to have a chatbot option, but in some cases, they prefer using a chatbot, with 74% of internet users saying they prefer a chatbot for help resolving simple questions.
However, currently only 58% of B2B companies and 42% of B2C companies are using chatbots, which I find surprising. If the majority of consumers want (and expect) a quick, 24/7 chat experience to be available, then it sounds like companies have some catching up to do.
It’s worth noting that chatbots aren’t limited to just helping with support questions — businesses can use chatbots to increase conversion and revenue as well. Data shows that on average, businesses that use chatbots see more high quality leads captured, an increase in transactions, and more upsell revenue attributed to chatbot interactions.
Let’s look at a few ways that teams can use AI chatbots.
Use Case #1: Customer Service
When you use an AI chatbot to manage repetitive and commonly submitted support requests, you free up your support team to handle more complex customer issues and deliver white-glove service.
Additionally, by creating these support flows with an AI chatbot, you provide your customers with 24/7 support, which, for the majority of consumers, is the most helpful feature of a chatbot.
Here are a few key ways that AI chatbots can help customer service and support teams:
- General support. I recommend beginning with a “Get help” chatbot flow that’s modeled after your FAQ docs. After a few weeks, review the reporting to see what areas remain unresolved, then train and update your model accordingly.
- Self-serve for customers. This one is a no-brainer. Connecting your AI chatbot to your knowledge base, academy, and even dev docs is a great way to surface relevant content that can help customers find answers to their questions. Not only can your AI bot deliver the content, but it can use the content to formulate the appropriate response.
- Suggested replies. Most AI chatbots will offer suggested replies. Have your team spend time training the model by actively using these suggested replies. Whether your reps edit them, accept them, or reject them, the only way to make them better is by engaging with the feature so the model learns as it goes.
- Onboarding new support reps. Access to an AI chatbot is a goldmine for a new support team member. Support reps can use features like suggested replies to expedite their responses or search for past tickets by topic to see how previous reps have responded to similar questions. This allows support reps to jump into the queue without needing quite as much ramp time.
Use Case #2: Product Setup/Onboarding
Offering your customers support with their new purchase is a great way to make setting up a new product, or onboarding to a new software, feel personal. If the product or service you sell requires your customer to set it up on their own, an AI chatbot can offer quicker time to value for your customers and take some of the lift off of your implementation teams.
Gardner sees even more opportunity here: “I foresee AI expanding more into onboarding and customer success in the near future. We can use AI to think about helping customers set things up or be more strategic in their implementation.”
I’ve seen how onboarding bots offer value in both the D2C and B2B space:
- In B2B, CX leaders are looking at how they can leverage AI to enhance every stage of the customer journey.
- In the D2C space, not having a streamlined setup or onboarding strategy could cost you revenue, since data shows that 54% of consumers say they’d return a product if it was hard to install.
Here are a few ideas on how to use a chatbot for onboarding:
- Design a “Getting Started” or “Setup” chatbot that hand-holds users through setting up their product or service.
Pro tip: Go one step at a time, set a positive and encouraging tone, and require the customer to engage with the chat before moving on to the next step. Avoid just linking help docs in this bot — actually write out the steps and embed video tutorials in the chat flow if a step is especially technical or complicated.
- Design a “Troubleshooting” chatbot that’s based on common troubleshooting questions that arise after a customer starts using your product.
Pro tip: Take time to review this and see how it’s performing. I would be sure to ask myself: Is this bot helping me drive down support tickets?
- Once a customer completes the setup flow, or upon their next visit, launch a “Tips and Tricks” chatbot that shows the customer more cool features and functionality they can take advantage of.
Pro tip: I suggest you get creative with this by pulling in your customer marketing or customer advocacy team to feature existing customers showing off their tips and tricks.
Use Case #3: B2B Sales Teams
AI chatbots are a great tool for helping businesses move customers through the different stages of the sales cycle.
Chatbots can capture and qualify leads, offer a prospect the chance to book a demo right away, and curate content that resonates with prospects no matter where they’re at in the funnel.
Here are a few key ways AI chatbots can benefit sales teams:
- General lead capture. Instead of making your site visitors fill out a clunky website form, offer them a chatbot experience that feels conversational. AI can enhance the lead information by flagging things like sentiment and intent for you.
Pro tip: If you set qualifying criteria in the chatbot flow, like company size, business email, etc., you could have the chatbot drop your calendar for qualifying visitors and give them the opportunity to book time with you.
- White glove treatment for inbound prospects. If your chatbot has a way to recognize when visitors from your prospect list land on the site, then offer those prospects a personalized chat experience that includes the option to chat with you or book time on your calendar.
Pro tip: I’ve seen AE’s create hyper-personalized chat experiences for their “white whale” accounts, including a chatbot that presents a video of the AE addressing that company by name and introducing themselves.
- Curated content. Use your AI chatbot to create curated experiences for visitors, like relevant case studies for returning visitors, or specific experiences for inbound visitors who fit your ICP.
- Close the deal. If your chatbot can recognize when a visitor from an open opportunity lands on your site, you can create a more personalized chatbot experience that helps move them through the sales cycle.
Use Case #4: Industry-Specific Use Cases
AI chatbots are a great resource for every business, and while I highlighted a few different teams above that could benefit, I thought it would be worth visiting a few additional industry-specific use cases.
Service and Hospitality Industries
- Restaurants can use AI chatbots to help diners find a location and book a reservation.
- Stylists can use a chatbot to help customers choose a service and book an appointment.
- Hotels or B&Bs can use chatbots to help customers get answers to frequently asked questions, request additional services, or book add-ons.
Pro tip: If Millennials are part of your target demographic, you should really have a chatbot option. Data shows millennials overwhelmingly prefer a chatbot option for contacting businesses, booking services, etc. (I’m a millennial, and trust me — we don’t pick up the phone unless it’s an emergency.)
Recruiters
- Consider using an AI chatbot as an intake for interested applicants. You can customize the questions to reflect the information you need and set qualifying criteria (just don’t ask for personal information and be aware of data privacy laws!)
Realty/Property Management
- Depending on the complexity of your chatbot and how it integrates with something like an MLS, you could use a chatbot to allow visitors to search for properties by zip code, capture information for interested buyers, and set appointments for tours.
How to Choose Between AI Agents and Chatbots
Now that I’ve covered the major differences and use cases between AI agents and chatbots, how do you decide which one is the best option?
In general, I think it comes down to a few key factors like budget, bandwidth, your specific use case, and your desired outcome.
Reasons to go with a chatbot:
- AI chatbots are a great option for anyone, but work especially well for those who are just starting to use AI to help scale their organization.
- Small teams, small businesses, and those on a tight budget also benefit from choosing an AI chatbot.
- If your use case is fairly simple, and aligns with the use cases above, then an AI chatbot is a great option for you.
- Chatbots are likely more affordable than an AI agent, so if you’re working on a small budget, this will likely be your best bet. (I’ve seen AI chatbots greatly vary in capability, too, so if you’re looking for a more advanced bot, those options definitely exist.)
Pro tip: In my time working for a leading chatbot company, I’ve found that many companies will throw “AI” in front of their rule-based chatbot and call it an AI chatbot.
But just because a chatbot connects to your knowledge center and can surface help articles does not make it an AI chatbot. If you truly want an AI chatbot experience, look for a tool that uses NLP, can respond and adapt to changes in language, and does not rely solely on rule-based decision trees.
Reasons to go with an agent:
- AI agents are great for companies that are trying to solve for efficiency or headcount constraints.
- Agents are great for larger and mature organizations because there’s typically existing infrastructure to help build, integrate, and manage an AI agent to maximize its capabilities. (That’s not to say that an AI agent isn’t right for a small, scrappy company that has the right team in place to implement and manage!)
- AI agents are great for semi-complex to complex use cases — situations where you don’t just want to ask AI to pull data for you, but you also want a tool that can synthesize that data, make suggestions, and execute tasks for you.
- If you’re an organization that wants to leverage AI to reduce manual tasks, create efficiency, identify ways to improve the customer experience, and analyze data to make more informed decisions, then an Agent is the best option for you.
No matter which option you choose, here are a few final recommendations to consider before you sign the dotted line:
- Review integration capabilities with your current tech stack (especially your support ticketing system if that’s your use case).
- Ask about things like the time involved in training the model, guardrails, and data privacy.
- Read the reviews and ask around before you commit.
Make Your Choice: AI Chatbot or AI Agent
Whether you choose an AI chatbot or an AI agent, you’re sure to see gains in efficiency for your teams and an enhanced customer experience.
In my humble opinion, I believe if you’re just starting out or you have a really simple use case you’re looking to solve for, then you can’t go wrong starting with an AI chatbot.
However, if you find yourself leaning on external AI tools (ahem, like ChatGPT) to pull up information, synthesize data, or automate other tasks, then it’s time to explore an AI agent.
