Best Practices Nice inContact

Chatbots in the Contact Center

Subscribe

Chatbots in the Contact Center

/ Technology, Assisted Service
Chatbots in the Contact Center

A Q&A on chatbot usage, trends and developments. Views, best practices and advice from the experts.

With consumers increasingly expecting anywhere, anytime access to conduct simple transactions, more businesses are deploying chatbots to handle routine and repetitive customer service tasks. Automation has enormous appeal as an efficient, affordable option for companies that lack the human staff to effectively manage the workload in the contact center (one of the top 3 challenges cited by participants in our recent Challenges and Priorities survey).

As the technology continues to evolve with advancements in natural language, AI and machine learning capabilities, the chatbot experience is becoming more human-like and increasingly popular among consumers. Although Gartner had predicted that, by 2020, the average person would likely have more conversations each day with chatbots than their spouses (hello, Alexa), it doesn’t mean that chatbots will be replacing humans in the contact center, but rather working alongside agents to assist and enhance service delivery. 

To gain a better understanding of chatbot applications in the contact center, we reached out to chatbot solutions providers for their views on current usage, developments and advice for successful deployment and use. Our experts for this Q&A panel are: Chris Connolly, Vice President Product Marketing, Genesys; Emma Furlong, Lead Content Strategist, Clinc; Dave Hoekstra, WFM Evangelist, Teleopti; Scott Kolman, VP of Product & Corporate Marketing, Five9; Alok Kulkarni, CEO, Cyara; Jen Snell, VP of Product Marketing for Intelligent Self-Service, Verint; Shellie Vornhagen, SVP Marketing and North American Sales, Astute; and Charly Walther, VP of Product and Growth, Gengo.ai.

What are the most common applications/uses for chatbot technology in contact centers today?

Chris Connolly: Our customers are primarily using chatbots on owned digital channels, such as web and asynchronous messaging, to determine customer intent and, when appropriate, deliver the interaction to a suitable agent. Our findings show that 60% to 70% of web interactions that engage with a bot are contained within this experience.

Emma Furlong: The main focus of chatbot technology in the call center today is focused around replacing IVR systems. IVRs are traditionally the biggest pain point for customers seeking fast service or quick issue resolution.

Jen Snell: The most common applications of chatbots and IVAs are to assist in self-service efforts and alleviate contact centers from high-volume, tier-1 interactions.

Shellie Vornhagen: Many companies approach chatbots as the next iteration of a website FAQ, enabling customers to easily find answers to common questions. Some view them as a replacement for contact forms, letting customers contact companies without needing to hunt all over the website. A great chatbot will deflect most tier-1 customer questions and issues, be available across multiple communication channels (site, app, SMS, Facebook Messenger, etc.), and enable easy escalation to a live agent when needed. This addresses consumers’ expectations for instant, omnichannel support while allowing contact center agents to focus only on the cases where a human touch is truly needed.

What is the biggest misconception that business leaders have about chatbot capabilities?

Chris Connolly: The largest misconception that business leaders have about chatbot capabilities is that a stand-alone chatbot is enough to provide superior customer service. Following this misconception are smaller misconceptions regarding the effort it takes to deploy a well-functioning bot. In the first instance, we have found that CX leaders find it necessary to connect the bot experience seamlessly with the human experience. On the other hand, companies that are simply seeking to “tick the box” on chatbot often rely on upfront stand-alone bot experiences with little to no connection with the contact center—leaving customers with a broken journey that ultimately leads to further frustration and damages brand perception.

We have also found that many customers we engage with believe they either need highly specialized skills to deploy a chatbot, such as that of a data scientist, or on the opposite end of the spectrum, that IT can do it with existing talent. In our experience, the answer lies somewhere in the middle—technology has matured to the point where specialist skills are no longer required, i.e. you don’t need a Ph.D. However, the effort required to satisfactorily maintain a working bot is also understated. We have found that many IT departments that do not understand user interface design have attempted to deploy bots only to be left with an end-user experience that is, well, robotic. Similarly, we’ve seen CX leaders identify new roles, such as “bot content curator,” which are charged with ensuring consistency of persona and experience delivered by the bot, drawing a parallel with Voice User Interface Designers (VUIs).

Emma Furlong: I think the biggest misconception around chatbot capabilities is that chatbots can’t handle context or messy human language. While many chat solutions do struggle with this, it doesn’t mean that the technology doesn’t exist. Clinc’s conversational AI was born out of a research lab at The University of Michigan which innovated on several key problems in the field, context among them, to see if it was possible to create an authentically human experience with AI. This research endeavor uncovered new approaches that make things like contextual awareness, conversational healing and the comprehension of messy language possible, and we’ve already deployed chatbots with those advancements in the market today. The technology is out there, you just need to know what to look for.

Dave Hoekstra: The biggest misconception is that chatbots will eliminate the need for skilled workers. When chatbot technology matures, it will be able to automate repetitive tasks and foster better customer experiences than IVRs and other self-service technology have done in the past. Yet the ability for a human to empathize and make decisions in a “gray area” is nearly impossible to replicate. In fact, the growth of chatbot implementations is instead placing a greater focus on the frontline employee and their competence. With chatbots taking on more of the mundane, automatable tasks, employees are instead presented with more complex cases that require emotional intelligence and a higher level of skill to bring customers the solutions they need. Emotional intelligence and problem-solving skills are particularly needed in the case of “bot-failure,” when a customer doesn’t get their question resolved by the bot and is transferred to a human agent. Then companies need to make sure they have knowledgeable, empathic employees available to turn customer disappointment into a positive experience.

Scott Kolman: Probably that it is the panacea of all interactions. Like any other interaction channel, it is important to understand and assess what types of customer issues are best served over that channel, as well as what customers expect to accomplish.

Jen Snell: One of the biggest misconceptions is that all chatbots are created equal and will easily deliver value for your business.

Enterprises are inherently complex, and when you’re truly looking to solve real business problems or achieve goals with this technology, it’s never just a matter of building a brain and plugging in an API. It’s a matter of integration with systems of record, continuous improvement and evolution of language models, designing and deploying to deliver measurable ROI to the business, along with the ability to extend and grow with ease overtime, and so much more.

Shellie Vornhagen: A common misconception about service chatbots is that you will need to author every response the bot can give. But advanced bot solutions can take into account any existing content you may have, like knowledgebase articles or FAQs, as well as gather information via integrations into other systems or websites.

Another misconception we see with digital self-service in general is an attitude of “set it and forget it.” While some chatbots can automatically learn and improve based on what your live agents do, you will also need to regularly review reports that show what gaps exist in the bot’s knowledge. And since products and policies tend to change, having an easy, non-technical tool to update content is absolutely necessary.

Lastly, some companies make the mistake of launching a chatbot without an in-channel escalation path to a live agent. If a customer has chosen to interact with you over a website chatbot or via Facebook Messenger, they want to stay in that channel even if further help is needed—and they definitely don’t want to repeat what they just told the bot! Make sure that immediate, in-channel escalation is available at any point during self-service interactions, and that agents are given all the context of the interaction when assigned the case.

How are companies using chatbots internally to support employees?

Chris Connolly: Chatbots are beginning to be deployed for internal organizational requests such as facilities management, employee guidance and, most significantly, we’ve seen a strong uptick in organizations reaching out to Genesys to understand the assistive bot technologies within a given agent’s desktop. The number of inquiries has risen along with the level of sophistication of questions from customers. We now find that educated customers are coming to us with practical implementation questions whereas, just six months ago, the inquiries were more orientated toward future visions.

Emma Furlong: There are several different use cases for internal chatbots but they all boil down to giving employees quicker access to information or helping them achieve solutions more efficiently for their customers.

Dave Hoekstra: The spotlight is often on how chatbots can enhance customer communication and experiences, but they also have the potential to deliver great internal value to an organization by offering timely, effective support to employees. Chatbots can come in both as a form of colleague to employees and as an extra channel for the company, getting the right information out to relevant employees, at the right time. For example, with workforce planning processes, chatbots could send direct messages to employees about the upcoming possibility of overtime hours or time off that they can choose to take. This creates an efficient but friendly dialogue with the employee, offering them greater options over their working time, while empowering the company with more flexibility and transparency for real-time operations.

Jen Snell: Companies are using chatbots to support both primary and non-primary job duties. For example, integrated chatbots to support Human Resources or IT Helpdesk, assist employees with getting the personalized information they need instantly. Instead of searching for PTO or business policies, employees can simply ask the chatbot or Intelligent Virtual Assistant; which will provide them with the exact answer they need. Chatbots are also being deployed as a resource assistant, supporting and assisting employees throughout their day.

What should contact centers do to prepare for chatbot adoption?

Chris Connolly: We suggest forging alliances among key business stakeholders that are jointly responsible for the customer journey. Specifically, we advocate for contact center executives to reach out to sales and marketing colleagues to start a conversation about the important role the contact center has in closing and retaining customers, along with the enormous amount of useful customer data that marketing and sales teams may not know about. For example, interaction analytics have been proven to provide a rich set of real-time insights about marketing campaigns as well as sales insights. Additionally, and more tactically, we recommend that contact center operations consider a centralized knowledge management function and toolset that can be leveraged by voice and chatbot technologies to help accelerate adoption and ensure a baseline set of content for such projects.

Emma Furlong: The best thing a contact center can do to prepare for chatbot adoption is to select a full-service conversational AI platform provider that handles everything from data curation and collection, to performance metrics and analytics. From the customer communications perspective, it’s important to demonstrate the value of the chat technology being implemented, and make it clear to customers where and how to access the chatbot.

Dave Hoekstra: Before an organization can utilize a chatbot, a deep understanding of the decision-making process is required.  Mapping out how decisions are made is a great exercise for a contact center in general but can also significantly reduce the amount of time it takes to leverage chatbot technology. Equally, like any software deployment or operational change, all employees that will use a chatbot by their side or directly interact with it for answers should be properly onboarded, rather than thrown in at the deep end.

Scott Kolman: A key first step is to understand what the customer is trying to achieve when engaging with the business, then determine the best way to handle. How do they prefer to communicate with your business? What is the intent of the interaction, and is this a reasonable candidate for a chatbot or is it better to have the customer engage directly with an agent? For example, if the customer is looking to find out the status of an outstanding order, you might want to handle the interaction from a chatbot. Conversely, if they are calling you to make a major purchase, or cancel their service, it may be best to have them engage with an agent specifically skilled to address that issue. Key to all of this is to provide a seamless transition from the chatbot to an agent—should the need arise. Make sure that the transcription of the interaction is made available to the agent, along with guidance on how to move the interaction forward to satisfactory conclusion.

Alok Kulkarni: Chatbots have actually become one of the most important ambassadors for your brand, so they should be designed for continuous improvement. This means your business and technical teams will need to embrace techniques that foster and enable rapid innovation—and there are two key ways to enable that. First, your business team must very specifically define the specific customer intents your chatbot will handle. While this may be painstaking work, this step is critical for then designing training and testing protocols to ensure that it meets those requirements. Second, in order to plan for continuous evolution of your chatbot’s capabilities, turn to an Agile/DevOps approach. This will yield tremendous efficiency benefits by enabling an iterative process supports frequent feature and capability updates. Here’s a tip from one of our customers: Start small. Take a service you’ve already automated and recreate that service in a chatbot interface. This gives you a clear, simple use case—one that you know can be automated—to learn from and work out the kinks.

Jen Snell: The real secret to adopting chatbots and IVAs isn’t technological; it’s cultural and philosophical. In traditional IT cultures, resource planning was based around the notion of completing one project and moving on to the next. In contrast, AI-powered technologies such as IVAs and Chatbots are predicated on the notion of continuous improvement through a combination of machine learning and human oversight. To prepare, it’s most important to have long-term business and IT vision; clearly defined goals and a collaborative culture.

Shellie Vornhagen: A chatbot represents the contact center team’s opportunity to get front-and-center on your brand’s digital properties, so early and frequent collaboration with your web team (and app team, if applicable), is a must. Think about bringing together a crossfunctional team that will run point on the project, ensuring every phase runs smoothly.

Speaking of phases, it’s important to agree on phases for your chatbot deployment—which use cases will you address first? Which channels will you deploy? What will the bot be able to do right away, in six months, in a year? Asking these questions from the start of the project will set the stage for a successful chatbot.

It’s also important for contact centers to understand what is being automated and how that information ties into the knowledgebase. Not only does this encourage confidence in the chatbot, but it serves as a reminder that the chatbot and the knowledge from which it draws information are intrinsically tied.

Last but not least, don’t forget about escalation. Map out how the escalation process from chatbot to agent will work, including situations where you would want to proactively suggest the customer engage with a live agent.

Charly Walther: Since chatbots are designed to make communication easier, it’s absolutely critical that they have an excellent understanding of language. In particular, there’s one scenario that chatbot users are desperate to avoid: the situation where the chatbot might have the right answer to a question ready to go but is unable to deliver the information because it didn’t understand the way the customer phrased it. Almost every question you can think of has multiple variants, all of which your chatbot has to understand in order to perform effectively. Without due preparation, your chatbot can get stuck because the customer said, “Can I return this if I don’t like it?” rather than, “What’s your returns policy?”

To build an effective chatbot, it’s important to move away from thinking about one list of canonical question-and-answer pairings. Instead, your data should contain as many variations as possible of the questions you want the chatbot to answer. Gathering 20 variations of a single intent such as, “Tell me more about the returns policy,” will allow your chatbot to serve a much wider range of clients. It will also help it to learn the meaning of new phrases that it might encounter while serving your customers, thus building a more effective knowledge base.

A good place to start gathering these intents might be your existing customer records. Take a look at the different ways in which people ask the same question and try to incorporate them into your chatbot’s training dataset. This will give your chatbot a stronger foundation and make it a more valuable asset in the long term.

How can businesses overcome customer reluctance to use chatbots for service and support issues?

Chris Connolly: All business leaders contemplating the deployment of bots should consider mindful design of the user experience. Bots need to have utility, but they also need to get out of the way quickly when they are not wanted by the end user or if they are unable to answer the customer inquiry. Putting a bot in the experience with the intent to simply contain the user to that experience will do more harm than good. We recommend that, first, business take the steps to roll out bots internally to an expert audience, and then progressively expand the invited parties before exposing the capability to end customers. We have found that the most successful customer adoptions have focused on specific and narrow tasks. This must be clearly communicated to the end user so as to set clear expectations of what can and cannot be done.

Emma Furlong: One of the main reasons customers are reluctant to use customer service chatbots is that they lack trust that the bot will handle their issues successfully. If you can deploy a virtual assistant that defies mainstream expectations by understanding natural language and resolving issues quickly and with accuracy, any reluctance will diminish, if not disappear.

Dave Hoekstra: Make the chatbot a guided tour instead of an open-ended journey. Customers have an inherent distrust of chatbots; not because they don’t trust their abilities, but because experience has taught them that chatbots won’t do what customers need. By making the chatbot gently guide the customer to the queries that can be performed, customers will slowly gain trust that their queries will be answered and their time won’t be wasted.

Scott Kolman: A few things can be done here to lessen resistance and encourage participation. First, don’t try to fool the customer that they are talking to a live agent. Make it clear that they are speaking to a chatbot. Second, make it easy to escalate the interaction from a chatbot to a live agent, and when you do so, ensure that the text of the conversation is provided to the agent to avoid the need to start all over. In fact, in a recent Zogby Research consumer survey, respondents were asked their willingness to interact with a bot. A relatively small portion were open to it, yet a much larger percentage of respondents (regardless of age) were open to it as long as they can speak to a live agent, if needed.

Jen Snell: People are increasingly adopting and relying on intelligent systems in their daily routine. To build reliability that will drive customer engagement with chatbots, it’s important to focus on transparency and trust with end users. A chatbot or IVA is much like an employee—the more you train it the better it will become. As the IVA is in training, it’s always important to keep the eye on the overall goal, which is to serve your customers. When the chatbot is unable to answer, make sure that you seamlessly connect the customers with the right person to get them the information they need. Obtaining a positive outcome with little effort will ensure that your customers do not abandon the chatbot.

Charly Walther: It’s important to remember that a chatbot should initially allow customers to choose their own destiny. If they are reluctant, the easiest thing to do is simply offer them both options. For example, while clients wait to talk to a human via support chat, they could be offered the chance to ask the chatbot some questions in the meantime. In this way, the chatbot can only be a positive experience. If it fails to answer the right questions, then the customer simply continues waiting for the support agent.

This kind of setup also allows the chatbot to take on more responsibility as clients become accustomed to it. As you to collect more data for your chatbot, you can train it to answer more and more questions. Eventually, you will be able to tip the balance and have the chatbot answer more questions than the live support agent. Through gradual expansion in a safe environment, you give customers the opportunity to choose the chatbot for themselves.

What is your best tip for successfully onboarding chatbot customers?

Chris Connolly: Make the experience simple, clear and connected with your brand experience. Don’t isolate your chatbot or leave it as a customer experience island.

Emma Furlong: The best way to onboard new customers to a chatbot is to show them a remarkable experience. If a customer sees a demo or video of the chatbot in action, and solving real problems, they will be much more likely to drop their misconceptions and try it.

What are the key trends and developments in chatbot technology that contact center leaders should look for to improve the customer experience?

Chris Connolly: 1. The technology behind bots is not limited to text/digital channels. Smart vendors are also building technology that enables conversations to happen with bots on voice channels using a common management system. For example, bots created for the web can also be used on a business 800-number or smart speaker.

2. Bot experiences are not limited to conversational dialog—they can also include embeddable “micro-applications” that improve the user experience by providing rich content experiences.

3. AI technology has expanded beyond chat and voice bots. There now exists technology driven by machine learning that matches the customer with the right experience (human or machine) based on collected journey data and digital footprints. These insights commence well before the customer has touched the brand, and extend far beyond the immediate customer communication.

Emma Furlong: There have been significant developments in the machine learning and natural language processing research communities over the past few years, but one of the key innovations is around crowdsourcing for data curation. Soliciting utterances from the crowd provides a quick and diverse data set to train AI models on, much more effectively than other data collection methods. Using this approach significantly cuts down on chatbot development time while simultaneously driving a better user experience.

Alok Kulkarni: It’s tempting to think of chatbots as the “next generation” of customer engagement. I think that all-in approach can be risky. Businesses may be ready to hop on the AI and chatbot train, but many customers are still hesitant. What I’ve noticed is this: Brands that have successfully implemented chatbots have taken a hybrid approach, with AI helping to make human interactions more effective and more impactful. They seamlessly integrate them into the customer journey by defining realistic operating goals for the chatbot, and then ensuring continuity with their other customer engagement channels.

Jen Snell: The vanguard of artificial intelligence (AI) in customer experience has of course been in customer service, where it went from cost-saving automation to making the customer service center a strategic source of data, customer information and business strategy.

We’re now seeing AI-powered chatbots and IVAs move to more direct customer engagements, especially as customers have begun to expect the convenience of automated interactions. Our clients have seen the biggest direct payoff as far as ROI on AI investments in assistants that are actually involved in sales, and for a simple a reason: AI agents actually ask the questions that create upsell and cross-sell opportunities.

Over the next year, we expect AI assistants to move even further toward the front lines of customer experience, driven as much by customer wants as much as business opportunities. At this point, companies need to be investing in AI customer experiences not to get ahead of the curve, but to just stay relevant.

Charly Walther: The rise of combined chatbot and human-in-the-loop solutions will have a drastic effect on the customer experience. The additional learnings that human-in-the-loop provides will result in chatbots that are more confident in their own answers. This will make them more accurate, more informative, and better able to discern when a support agent might be needed.

As this technology is ever more widely adopted, customers will come to see excellent service from chatbots as the norm. To deal with these heightened expectations, it’s important that your data covers both a wide range of questions and a variety of ways to communicate with them. Fortunately, another byproduct of the improvement in chatbot technology has been an ecosystem of third-party providers who are able to either annotate your data or build a custom solution for you. When they’re effectively utilized, these specialists can provide you with a dramatic ROI and help you to avoid falling foul of a gap between customer expectation and chatbot reality.

What advice can you offer for matching chatbot solution options with business requirements?

Chris Connolly: Come and talk to us—we have experts on staff and great customers willing to showcase what they’ve learned.

Emma Furlong: The most obvious business requirement for a chatbot solution is to deliver ROI. The contact center is one of the use cases with the highest potential for massive ROI due to the fact that it can replace/augment contact center workers, and reduce handle time. This translates to massive cost savings in the contact center, and provided that the chat experience is positive, the business value is clear and immense.

Scott Kolman: Start with your more repetitive questions and interactions as your initial candidates for a chatbot. Look for activities that have been performed quite frequently and have well defined processes, and answers. From there, make a chatbot interaction an option, while presenting up front how the customer can engage with a live agent—if desired. As customers gain experience engaging with chatbots and find their issues easily resolved, they will increase their openness to interacting with a chatbot in the future.

Alok Kulkarni: Set clear goals that match both your business objectives, as well as your customers’ objectives. Here’s a real-world case study from a business software company that offers support online and via their contact center. Typically, customers start with online support and when that fails, they call the contact center. Knowing that many customers actually prefer to resolve their issues without calling, the contact center team focused on how they could better serve those customers with self-help content. They hypothesized that chatbots could strengthen the experience for customers, and therefore, they focused their AI deployment on customizing support conversations with the aim of increasing the rate of self-service completion. The result was successful, achieving resource savings for the company, while also maintaining a strong customer NPS. In fact, their business was then able to apply the resources savings to improving other aspects of their CX.

Jen Snell: The technology is not new—find a chatbot solution that has been proven to deliver meaningful results in an enterprise environment.

Shellie Vornhagen: My advice is to just get started! Don’t wait until you have the “perfect plan.” It’s fine to initially build out just a few use cases, and then use an iterative process to add more over time.

Another element to keep in mind is flexibility: Make sure the core functionality of your chatbot can be built once, but deployed on multiple channels with minimal modifications so you can maximize the value of the bot. And remember that key technologies behind chatbots, like NLP and AI, are still rapidly evolving, so look for a flexible solution that can leverage multiple technologies as they continue to change.

Charly Walther: There are several different approaches to building out your chatbot. These range from a generic off-the-shelf solution to a more expansive, customized offering built on hundreds of intent variations and perhaps some human-in-the-loop flows. Regardless of your choice and level of commitment, it’s essential that you have target KPIs for the following questions:

  • How often do you expect your chatbot to produce the right answer?
  • What is your target NPS?
  • How often is it acceptable for the chatbot to reply with “I don’t know” or “have a look at our FAQ?”

Once you have solid answers to all these questions, you can begin to trial generic approaches to chatbot implementation. It’s important to evaluate the strengths and weaknesses of these options thoroughly. This will allow you to have more useful conversations with service providers to see if further investment makes sense—and whether it will enable you to make significant progress toward your KPIs.

Susan Hash

Susan Hash

Susan Hash is the Editor of Contact Center Pipeline magazine and the Pipeline blog. She is a veteran business journalist with 25 years of specialized experience writing about customer care and contact centers.

Contact author

x