Answers to your questions about artificial intelligence in contact centers

September 2, 2021

Artificial Intelligence (AI) is the veritable reigning champion and current all-star industry buzzword. Most people interact with AI daily in some form or another. But how much more would you like to understand about how businesses are using and deploying AI and automation in their contact centers?

To help, we launched a series of content, events, and webinars designed to take you through the 101-level understanding up through the masterclass, where you’ll learn to put theory into practice. Our goal is to help you understand how to best leverage smart contact center technology, and how to make the case for AI in an everyday contact center setting.

Below, we’ve addressed the great questions you’ve been asking in three live webinars:

If you’re reading this (and haven’t yet attended yet) but want to get at least 30% smarter about AI, we’ve linked the on-demand webinars and helpful content from our educational AI-series below. Don’t miss out!

Real-life chatbot examples

In “Successful Chatbots: Customers and experts share the secret sauce,” a panel of contact center leaders and CX chatbot experts discuss what makes a chatbot successful and explore some real-life chatbot examples. Watch the webinar now.

Here are questions asked by the live audience:

What is the difference between AI chatbots, virtual agents, and FAQ-based chatbots?

These could all be the same thing—depending on the complexity and the technology behind the scenes. Definitions can vary across the industry because so much of what a chatbot is relates to how it’s programmed. This is how we distinguish between these terms:

Rules-based chatbot (aka FAQ bot, menu/button-based chatbot, or decision-tree chatbot)

This is the most basic type of chatbot that answers questions by offering menu options and buttons to the customer who inputs their need, and the bot provides a response. These are programmed with rules, for instance, “If customer selects ‘Order Status’ then confirm order number, match the number to the database and input corresponding status.

These are good for high volume, easily repeatable tasks such as ordering, answering FAQs, and routing to live agents. Although they can receive inputs and respond to customers, they can’t understand them or reply outside of their programmed commands—they can only do tasks or respond where there are fixed outcomes.

In their most basic form, these rule-based bots aren’t typically using any artificial intelligence, just business rules that are defined by humans—although more advanced applications might combine AI technology with rules-based models.

AI-based chatbots (aka conversational AI chatbots)

This is an umbrella term for any chatbot programmed with artificial intelligence technology, such as machine learning (ML) or natural language processing (NLP). AI-based chatbots help facilitate more complex interactions than rules-based bots. Contrary to rules-based bots, these can better process, understand, and respond to a customer based on the context of the query outside of fixed outcomes.

For instance, AI-based chatbots programmed with one type of AI technology, natural language processing, use keyword recognition to understand a freeform user query, so a bot can understand and respond to the various ways a customer might ask a question, like “what’s my order status” or “where is my shipment?”

AI-based bots can facilitate a variety of tasks from password resets or collecting and processing payments, reporting balance inquiries, handling claim statuses to confirming, setting, or rescheduling appointments or reservations.

Virtual agents

Virtual agents (aka VA or virtual assistants) fall underneath the umbrella of AI-based bots. These are the most advanced that support both voice (IVR) and text (chat), combining technologies like NLP/NLU with machine learning and robotic process automation (RPA). These can follow engage in two-way multi-turn conversations, can process and understand speech, text, and get smarter over time.

If you’d like to better understand AI technology in a fun way, download “Contact Center AI Taught Through Pop Culture” to explore AI examples from TV and movies.

Got advice on how to improve a chatbot?

To improve chatbot efficiency, constantly analyze performance metrics for optimizations. For the highest impact improvements, consider these chatbot key performance indicators (KPIs):

Survey for customer satisfaction score (CSAT) or customer effort score (CES) – One way to measure customer effort is by pushing a quick survey in the chat to understand how easy it was for the customer to solve their problem or at the end of a customers chatbot or cross-channel interaction beginning with a chatbot to ask for the customer’s level of satisfaction.
Session duration – Customers expect a quick resolution, so monitor and try to decrease the amount of time it takes for a chatbot to resolve customer issues. Another way to optimize is to look at the average length of chatbot interactions compared to the ratio of users who make a purchase or become a lead after interacting.
Increase First Contact Resolution (FCR) and containment rate – Customers want to speak to one person, so you ideally want an interaction to be wholly contained within the same channel in which the customer reaches. Trends in interactions where customers begin with a chatbot and are unable to resolve are great ways to optimize your chatbot.

To effectively improve your chatbot, it’s best to monitor and measure 100% of interactions and have an understanding of the full end-to-end customer journey. And to make it easier to analyze and visualize data, it’s helpful to have AI-powered Interaction Analytics that categorizes interactions based on topic or customer sentiment and makes it easy for you to identify trends and clues to what’s causing the most common issues, including churn.

Does NICE CXone have a chatbot demo?

Here’s a quick 3-minute overview of conversational AI-powered IVR and self-service bots:

If you’re looking for a more extensive in-depth demo of CXone including self-service features and more, follow the link for a 20-minute video tour.

Is it better to optimize chatbot scripts and content based on questions gathered from an IT Helpdesk, common questions that customers ask, or a hybrid of both?

All of the above. Questions your support team routinely answers and questions customers most asked are both great ways to optimize your chatbot flow. Our recommendation is to formalize a knowledge management program to capture, organize and curate all organizational knowledge, including answers to customer questions and helpdesk tickets, that you can then use to supercharge your chatbot.

Is a chatbot only optimal for simple answers? How effective is a conversational chatbot when dealing with complex or multilevel troubleshooting issues?

Chatbots are ideal for simple, easy-to-repeat tasks, such as answering routine questions or checking order status. For more complex use cases, you’ll want to consider investing in a conversational bot that can process and understand human language, engage in multi-turn conversation, and handle more complex tasks. Certainly, the sophistication of AI-powered virtual agents makes supporting more complex interactions 100% realistic and feasible. Like, a chatbot that helps you book your flight, plus reservations for the hotel and car rental too.

And we say this with a cautioning. Our benchmark findings indicate that 90% of businesses believe chatbots and virtual assistants need to get smarter before consumers are willing to use them regularly. So although it’s entirely possible to implement a successful chatbot customers like, it requires thoughtful and strategic implementation and maintenance. A poorly designed or configured chatbot isn’t an effective asset.

Here are some chatbot best practices:

Don’t try to have your bot do too many things. Instead, have several bots that each performs a certain task and are linked by your IVR
Design a pilot and determine how it will scale
Evaluate how the solution fits into your existing process
Integrate agent escalation and equip agents with customer context
Consistently test and train. Chatbots need ongoing support just like human agents

Can chatbots be used to present content relevant to the page the user is on?

Absolutely. And any opportunity to proactively meet a customer’s need is an opportunity to create a loyal customer.

Let’s say, the customer searches Google attempting to troubleshoot their Whatchamacallit device and lands on a support article for the Whatchamacallit website. The article is close but doesn’t quite answer her question. But there’s a chatbot that pops up with a proactive offer, “I see you’re having an issue with your whatchamacallit. Here’s another article that may help.” There’s a link to a relevant article and maybe even an option for the customer to choose a callback or video call with a live agent for support.

The AI bot described here used an awareness of context to better understand the customer’s intent— where they came from (i.e., from Google, from a display ad, from social media, etc.) and what kind of content they’re looking at or interacting with (e.g. troubleshooting). Based on the context, chatbots can present relevant content and offers—versus the generic “Sign up for our newsletter!” that can be a real turn-off when all you want is to find your answer as quickly as possible.

Is it better to have one bot handle everything?

Rather than having one bot handle everything, it’s more efficient and effective to match the bot to the job. You’ll want to deploy bots that are designed to handle a specific use case (e.g. fetch FAQ answers from a knowledge base, reset passwords, or schedule service calls).

Is your use case for customers or agents? Helping customers self-serve, or automating an internal workflow? The best approach is to identify the task/goal, then decide what kind of bot or automation can best support that use case. For instance, you can use an RPA bot to help agents fill in details on forms, and deploy an NLP-powered chatbot to answer customer questions about applying for a mortgage. Different functions need different bots.

What’s the average, or benchmark, for chatbot containment?

A recent benchmark survey determined that approximately 40% of customer service experiences and interactions happen in self-serve channels.1 But successful containment within self-service (where the customer gets what they need and there’s no need for an agent) only happens when the bot is effective at successfully resolving the request. If you start with relatively simple narrow use cases for high-volume requests that are easily automated (i.e. don’t require a human), then it’s not unreasonable to expect even just a 5% containment increase! Read some great CXone customer success stories here:

Can a chatbot help organizations that want to scale to 24/7 service without adding additional headcount?

Yes! Chatbots are an ideal solution for 24/7 support because bots don’t need to sleep like agents do.

When an agent is available, ensure there’s seamless elevation to the agent, including routing the context of the customer interaction to the live agent
When an agent isn’t available, A chatbot could answer questions, collect details from a customer, and even schedule a callback.
implement outbound callback scheduling, incorporate an option to send an email, or other alternative means of solving their issue as soon as agents are available.

Are issues like data governance, personally identifiable information (PII), and data privacy handled by a typical contact center chatbot experience, or separately by IT/IT SEC/data governance?

It depends on your chatbot application, as well as your contact center technology. Like anything, chatbots are only as secure as you make them. It’s always a good idea to consult with internal or external security specialists. For deeper insight, check out this post, “Security and Ethics of Contact Center AI: When is AI Creepy?”

How do medium-sized B2B companies make the case for chatbot implementation?

Automating routine uses cases (e.g., opening a ticket or placing an order) with self-service options is a much cheaper way to do business. One way to help justify is to calculate the quantity of interactions and time spent by human agents to determine what you’ll save. In ContactBabel’s most recent edition of “The Inner Circle Guide to AI, Chatbots and Machine Learning,” is a table that breaks down the average cost per interaction by channel:

average cost per interaction by channel table

Of course, making the case for an investment is a much more comprehensive process. For more advice and insights into showing ROI for chatbots, see these resources:

10 ways AI can improve the contact center experience

There are more benefits to artificial intelligence than simply just automating contact center processes. AI has the potential to drive lasting customer loyalty and shape your contact center into a well-oiled machine.

In the webinar “10 Ways AI Can Improve the Contact Center Experience,” a panel of experts discusses how AI amps up contact center productivity and improves the overall experience for customers and employees. Watch the webinar now.

Here are the questions you asked:

What AI-powered tools can determine an interaction’s primary cause for contact?

Our solution uses natural language processing and machine learning to analyze and categorize 100% of interactions by attributes like the key topic or even customer sentiment, which helps teams to better see trends and patterns that are unseen by humans. AI supercharges analytics, taking you from hunting for needles in the haystack to identifying trends and detecting root causes.

Does the NICE CXone platform integrate with Microsoft D365 and applications like Microsoft Teams?

NICE CXone integrates with both Microsoft Dynamics and Microsoft Teams among other leading UC tools such as Fuze.

Here’s a quick overview video of our Teams integration:

Are fewer customer service representatives needed in contact centers due to chatbot implementation?

No, in fact – interaction volumes –both agent-assisted and Self-Service—are higher than ever. AI doesn’t lead to eliminating resources—it augments them and adds efficiency. Artificial intelligence helps make your strategy more effective and your processes more efficient, resulting in reduced costs and resources. However, AI isn’t an agent replacement. Though many routine interactions are facilitated through automation, agents are best at supporting complex interactions, interactions where it requires more empathy, and to support the ongoing demand for voice interactions.

Also, keep in mind that AI automates processes, not people. Agents know and carry out these processes, and implementing AI upskills agents and enables them to perform higher-value transactions. The low-value tasks, such as dispatching or registering cases, are automated and taken care of by bots. But remember, the human element of the contact center is still imperative, and AI technology should only be enhancing their capabilities, not eliminating their positions.

Here's a great blog post that digs deeper, “Who can do a better job in your contact center: humans or AI?”

What AI capabilities drive the fastest ROI? With so many capabilities and features, what should you prioritize?

You’re not alone in this challenge. We hear, “How do I determine the AI use case that’s right for our contact center?” frequently enough that we designed an interactive flow-chart tool that can help you determine the AI use case that’s right for your goals and organizational maturity (Use it on a desktop browser for best results).

But the good news is that AI and automation can help you determine a distinct, measurable ROI. Here’s a blog on demonstrating the ROI of contact center AI.

It’s important to prioritize the elements that have the greatest impact on your business goals aligned to your customer journey. For instance, solving for what causes customer friction. Or because so much of CX depends on agent experience, another good practice is to consider what frustrates agents. You can look at call or contact volume regarding certain aspects, such as CSAT, to identify use cases and subsequently drive a greater ROI.

Learn more about contact center AI in the e-Book “Contact Center AI Throughout Pop Culture.”

Can you aggregate both agent and customer sentiment analysis? Can a supervisor monitor an interaction where sentiment has escalated?

Yes. AI enhancements to Quality Management and monitoring tools and analytics you can aggregate data about customer sentiment and agent behaviors. Supervisors can be alerted if the “temperature” of a call has gone off-track (e.g. when sentiment score exceeds the threshold). The supervisor could then listen in and take over remotely if necessary. Most previous processes had agents flag down their supervisor on the contact center floor when the agent was experiencing an issue, which isn’t as feasible now with remote delivery models.

Does contact center chatbot deployment need specialized hardware (a graphics processing unit, or GPU), and at what rate per agent?

If your strategy and use case calls for computer vision or processing of images, it could use a GPU, but if you need hardware or your solution incorporates a GPU is entirely dependent on your specific scenario and use case. If the chatbot runs on cloud-based software, additional specialized hardware will likely not be necessary, as there is no on-premises equipment for the bot. If your technology is hosted instead of cloud-native, you might need additional hardware.

Building your business case for the ROI of AI in your contact center

When modernizing a contact center, leaders may consider AI-powered solutions. Several CXone experts hosted a webinar, “Build a winning case for AI in your contact center,” where they weighed in on some of the key benefits of AI use cases within the contact center and how to determine the ROI of AI in your strategy.

Watch the webinar on-demand now to get the answers to the most-asked questions below.

In the realm of contact center AI, there’s a lot of information to unpack. To make this information as accessible as possible, all webinars discussed in this piece are available on-demand via the links above. If you are interested in learning more about contact center AI, explore these additional resources:

Human or Computer: Spot the Bot Quiz – How high is your AI IQ? Can anyone tell what’s human and what’s a bot nowadays? Put your instincts to the test.
Effortless Service, Happier Customers – Find out what it takes to implement exceptional self-service within a contact center.
Inner Circle Guide to AI, Chatbots, and More – Deep dive with ContactBabel into the extraordinary use cases for contact center AI, how it establishes customer trust and relationships, and much, much more.
Solve for the Human Side of AI In Your Contact Center – Learn how agents and AI work together in harmony. How are agent roles being reshaped? How is AI impacting contact centers across the board? Find out more in this educational whitepaper.

friendly positive cute cartoon orange robot

If you liked this blog, we hope you’ll join us for an upcoming webinar or event.

¹ NICE CXone: 2020 Customer Experience (CX) Transformation Benchmark

About the Author

Tamsin Dollin

See All Blogs