What is CX Automation?

Michael Krigsman: Welcome to CXOTalk, where we discuss leadership, AI, and the digital economy. I'm Michael Krigsman, and we're exploring CX automation with Dan Bodner, Founder & CEO of Verint.

Dan Bodner: Verint is a CX automation company. We focus on helping brands increase CX automation. We have most of the Fortune 500 brands in the world as our customers, and we're focusing on helping them in the contact centers and CX operations to elevate the customer experience. At the same time, we aim to achieve this with the same budget and resources.

Michael Krigsman: Dan, what is CX automation?

Dan Bodner: Customer experience is very important to any brand. This is where they delight their customers so they can create better loyalty and grow revenue over time. CX has always been a focus for brands. However, it's hard to do. It takes a lot of resources, and that is a tremendous drag on operating profit.

Historically, as brands tried to improve CX, they would hire more and more people in their customer service operations so they could better respond to their customers. But that is unsustainable. The industry today already has 15 million workers doing CX, and that's a $2 trillion spend annually.

With the introduction of AI, it's now possible to automate CX. CX automation elevates CX not just with people but with a combination of people and AI working together. This collaboration automates business processes that result in a better customer experience and lower operating costs.

Michael Krigsman: Dan, you said CX historically has been difficult to achieve. Can you tell us why? What are the key challenges?

Dan Bodner: Customers call the contact center to get service, answer questions, and do business. The technology was very much focused on telephony and enabling phone calls.

Whenever a customer calls, a person working for the company has to take that call and respond. This increases the resources that brands have to hire to provide quick and contextual responses.

You have to have enough people to respond to all the phone calls that you expect. You have to have people who are knowledgeable, capable, and have the skills and empathy to respond well. You need all of this to be available so that the end customer feels that they are receiving the right experiences.

The technology was very much focused on people and their skills. Now there's a change, obviously, where AI can augment people.

Michael Krigsman: That raises the question of how does AI augment people in the contact center and what are the benefits that this brings?

Dan Bodner: As you consider deploying AI into an operation, make sure that your workforce is not disrupted because the point is you want to help the workforce to be more effective. You want to increase agent capacity with AI, but you do not want to create disruption to the workforce. This could happen by trying to completely eliminate them and getting customers to talk to bots, which sometimes works and sometimes doesn't. Or, when eventually the agent (the person) is on the call, they will have to completely change the way they operate.

AI needs to be embedded into existing workflows, so people will continue to do what they've done before but they will be assisted by AI. This assistance will help them do their tasks faster and shave seconds from the call, reduce costs, and improve the customer experience because customers will get better responses from agents.

Michael Krigsman: There is a strong efficiency argument, and you just touched on improved customer experience, such as personalization, for example. Can you elaborate on that?

Dan Bodner: One of the processes that the workforce needs to go through after every call is creating notes for the next person. When the end customers have already hung up, the agent needs to summarize the call. In case the customer calls again, those notes are available to continue the discussion where the first agent left it.

That effort can take a minute or sometimes longer. Typically, agents are not very good at writing notes. This is not their strength. Very often, those notes not only take a minute from the call, but also eventually, they're not really serving a good purpose for the next agent. They do not effectively provide the right customer experience by saying, "Hey, I know you called before, and I'm here to continue that discussion." That's an existing workflow that has been around for many, many years.

With generative AI, the process of summarizing the call can now be done by AI. The AI shows up and just writes the notes for them. Instead of spending a minute, they spend a few seconds just to read the AI summary and to approve it. This immediately cuts the call by a minute, which is great because it reduces labor costs.

But also, AI creates much better and more accurate summaries of the call for the next agent. So that elevates the customer experience the next time you call. You are getting a very contextual response, "Let me solve the problem that the prior agent couldn't."

Michael Krigsman: This is a very important point. On the one hand, you have these efficiencies and cost savings. At the same time, you're providing a benefit for the person calling in because you have that continuity. You have that personalization, and the call becomes more efficient from the point of view of the caller.

Dan Bodner: Fifteen years ago, there was an attempt to reduce the cost by moving many agents offshore. That resulted in a negative customer experience, and after several years, those resources came back onshore. Every time the industry has tried to reduce the cost, there was some element of tradeoff with the customer experience because the technology was not there to assist. This time, AI can do both, and that's why we see AI as the big driver behind CX automation.

Michael Krigsman: Dan, you mentioned the business outcomes, and it seems that a very important aspect of this is embedding the AI and embedding automation deep into the core processes of the contact center.

Dan Bodner: Absolutely. AI technology needs to be embedded in the existing workflows so it's readily available for customers and agents to use. It also needs to be continuously trained in the data chain.

AI is only as good as the data it trains on. If you deploy AI initially after you train it very well on data, it can perform well. However, data is changing very rapidly in the contact center. The conversations are changing from time to time based on new products that the company has launched or new policies. So it's very important that the data will be kept fresh and the bots will be training 24/7 in that data chain. If you bring the right technology with the right data into the workflow, all those three things are the secret sauce behind strong AI business outcomes.

Michael Krigsman: How does CX automation, as you've been describing it, change the contact center?

Dan Bodner: The contact center is going through the biggest transformation that I remember over the last 20 years. Creating more capacity within the workforce now allows the brand to choose between taking some cost out, elevating CX and creating customer loyalty, or training the agents to upsell and cross-sell so they can increase revenue. Because when you have a very delightful experience, you're very open to listen to ideas of what other business you can do with the brand. Brands are exploring different ways and different approaches to how we get AI business outcomes.

One of the great things about the Verint platform is that we designed the platform to be completely open. This means that you do not have to rip and replace your entire contact center in order to introduce AI. You can add AI into your existing ecosystem. You can add just one bot, and that one bot would give you a certain value. As you see that value, you will increase the consumption of the bot.

For example, we have a healthcare company that tried the Wrap Up Bot with 300 agents. They saw the results over several months, and now they've increased it to 30,000 agents. Customers are able, with AI, to consume it based on the value that they see, and they pay for what they use. That openness in the platform provides customers the opportunity to avoid these very long and disruptive rip-and-replace programs. They can start to embed AI into existing workflows without disrupting the workforce and see great value very, very quickly.

Michael Krigsman: They get clear results and, if the ROI is positive, then of course, they're going to come back for more and want to implement this more broadly.

Dan Bodner: Correct. Their IT departments are experimenting with AI, and they are looking for use cases. The Verint approach is quite the opposite. We have the use cases in the contact center that we have addressed for many, many years with other technologies. So we now embed AI into these existing use cases, and we can very quickly provide tangible business outcomes as opposed to an experiment.

Brands are looking for many, many different benefits. Of course, those benefits are beyond the contact center and beyond CX. But having a very mature, robust solution that is already proven to deliver AI business outcomes is what the customers are looking to deploy now.

Michael Krigsman: Well, Verint has such a long history with CX and with contact centers, so you know the dynamics, the challenges, and the issues around these use cases.

Dan Bodner: A lot of our customers already have quality monitoring programs. Those programs are very important because if you consistently assist the quality of your workforce or the quality of the conversations, you will find areas for improvement. You will find agents that need coaching, and you can motivate agents for good behavior. There are a lot of benefits to the quality monitoring program. However, one of the drawbacks of this program is that supervisors need to listen to calls, and they manually pick up calls randomly. They listen to the call, and they create an assessment. That's time-consuming and, quite frankly, quite boring.

With the Quality Bot that Verint introduced, you can use the same quality monitoring program you always had, but supervisors do not have to listen to any calls. The bot listens to the calls and automatically assesses the agents across all the criteria that you already designed for your agents. But it uses a bot to automate the listening and assessment. Plus, you're not just randomly picking calls.

Agents used to complain, "Hey, you just picked a bad call. I'm usually a great agent, but you just listened to a random bad call. Now you're judging me, and it's not fair." Well, the bot listens to all the calls, so the results of the bots are very, very accurate. The bias is not there, and agents are happier with these objective results.

You get more engaged agents; they're more open to receive coaching and receive feedback. Of course, you're eliminating the cost of the supervisors listening to calls, and so on and so on. A very good example of how bots are just embedded into the existing process and they're providing value.

Michael Krigsman: This is such an important point. The overarching process remains the same, and then you're inserting the technology at strategic points to improve that process. But the overall workflow remains undisrupted.

Dan Bodner: Yes. There are processes that were not in place before because it wasn't possible to do them manually.

If you're an agent, and you're on a shift between 12:00 and 4:00, and you just got a call from school that your daughter is sick and you have to pick her up, you call into your supervisor and say, "Hey, I want to opt out of my schedule." The supervisor will say, "No, we need you. We don't have anyone else. Our customers are going to wait in line if you're not around."

There's not much flexibility in this job, and that creates a lot of issues with work-life balance for agents, low engagement, and, quite frankly, an attrition issue. There's very high attrition with the contact center workforce. You really want to keep your best agents because, over time, they really become good at answering and responding to customers.

It's important to provide this agent flexibility, but it wasn't possible to hire enough scheduling supervisors to get all these requests for change. It was also difficult to change the schedule in real time so that we can replace you with someone else. This is a process that just didn't exist.

We created a bot. We call it the Time Flex Bot. That bot basically gives you unlimited flexibility. You want to change, you can change. The way we do that is to look at the schedule the way it is now and, if you opt-out, obviously that schedule will be somewhat of lesser quality. But there could be a midnight shift that is already really bad quality, so the bot will automatically offer you an option. You can get out of this shift, but you have to sign up for the midnight shift. When you do, overall, the schedule is getting better, and you get to choose which shift is best for you.

It wasn't possible without AI to really look at all of these different shifts with thousands and thousands of employees and find the best, optimum schedule manually. Now customers are implementing this and creating a whole new process around work-life balance for their employees, and it's all because of AI.

Michael Krigsman: It seems what you've done, given your very extensive experience with CX and with call centers, is dissected and analyzed the various aspects of call center operations and made choices about where you can apply this kind of AI-based automation to make the various kinds of improvements that you've been describing.

Dan Bodner: That's correct. We have been, for decades, improving processes based on the technology that was available. AI is just a game-changer. It's accelerating the opportunity to automate many processes that, before, we could optimize but not automate. That step-up in technology is creating a step-up in the business outcomes.

Ten years ago, I would be happy to talk to our customers about a 5% efficiency improvement with our software. Now we're talking about a 20%, 30% increase in agent capacity when you deploy 3-4 bots together. If you are a contact center with 10,000 employees, a 25% increase in capacity is $100 million in savings. It's a huge impact on the bottom line.

But of course, many of our customers are choosing not to just take the impact to the bottom line. They are also choosing to improve CX by giving more time for employees to show empathy and, of course, upselling and increasing revenue. So the transformation in the contact center is more than just efficiency. It's also, because of the amount of efficiency, it provides an opportunity to turn the contact center into a revenue-generating center.

Michael Krigsman: How is this different from existing methods of automation such as RPA (robotic process automation)?

Dan Bodner: RPA is used in the contact center for things like cut and paste. These are horizontal automation capabilities that are applicable in the contact center. Verint leverages RPA technology. We do not try to redesign it.

But some of the examples I gave before show that we are automating very unique contact center processes that RPA doesn't handle at all. We do this with deep knowledge of how contact centers work, how CX is different than what people do in HR, IT, or legal when you write documents and you need to cut and paste fields from one document to the next. The bottom line is RPA is an automation technology that is important to the contact center, but it's only addressing a small fraction of the automation opportunities in the contact center.

Michael Krigsman: Well, you're going far beyond the repetition of small, granular tasks such as cut and paste, for example, as you said.

Dan Bodner: Yes. With AI, obviously, one of the biggest improvements is in self-service—the ability to speak to a bot and get that bot to address your needs without even transferring to a person. That requires the bot to be really intelligent. They need to understand the intent of what you are asking, and they need to be able to find the answer and provide you that answer contextually so that you will be satisfied with the response.

Because a lot of these bots sometimes frustrate customers, we built some self-awareness into the bots. So, when the bots can contain the conversation because they have the answer, they will provide the answer. But when they don't, they will transfer the call to a person.

It's not just a routing exercise. The call will be transferred with all the context. So, you've been asking the bot questions. Those questions will be captured. That context will be provided to the agent so they don't get the call transferred and then, "How can I help you?" or, "I need to authenticate you." You've already been authenticated once by the bot. Now the agent can just jump in and keep responding and concluding the call.

That's again cutting the time from the call, which is good from a cost perspective because the agents spend less time asking you questions. But also, it's elevating your customer experience because you don't have to repeat yourself three times to get something done.

Another interesting development with AI is to bridge self-service and assisted service to be integrated and not just two separate functions. When self-service fails, the only choice you have is you hang up and you call again. With a Verint bot, you don't have to hang up. If it fails, it's because the bot is not smart enough, which all bots have a certain limitation. But then the bot will transfer. We have another bot called the Transfer Bot who will take their call from the Containment Bot, transfer it to an agent, and the agent will continue from the same point.

Michael Krigsman: What about the technology environment? What are the tech pieces that need to be in place to implement a system like this?

Dan Bodner: As you implement AI in the contact center, a very important premise is that you want to be able to leverage the latest and greatest of AI when it's available, and closed platforms do not allow that. They only give you the best AI at the time you purchase the system. With the fast pace of AI technology, an open architecture to AI is a critical choice for our customers.

Once you know you can bring the latest and greatest AI, you have to have an architecture that trains the AI continuously on fresh data. That data needs to be available in a unified data hub so it's readily available for those bots to train.

Now that you know you have the latest AI model and it's always trained on the latest and greatest fresh data, how do you bring that AI to the fingertips of the workforce? That requires the ability to architect that AI into your existing ecosystem. Customers have different technologies, so again, an open platform that has a lot of integration points will also help in that regard. Whatever your practice is, you don't have to change it, which could be very disruptive.

Michael Krigsman: Can you talk about the data that's required to make these systems work effectively?

Dan Bodner: The data that is needed by the bots is behavioral data. The best way for the bots to learn is from conversations that people had with customers. If you collect all the data about voice conversations, chat conversations, emails, surveys that you got back from customers, all of this is behavioral data that contains very important information about how the workforce behaves and how the customers behave.

That data, while it's a lot of unstructured data, it's multimodal data. There is voice, video, text. But if you bring all this data and allow the bots to learn, they're going to learn from the best practice, they're going to learn how people respond the best, and then your bots are going to be as good as your best people. This is really the best outcome you can expect.

The bots may be able to operate faster because they can move data faster, they can operate faster. But eventually, their behavior needs to be as close to your best people as possible because it requires the ability to listen well, to show empathy, to understand the context, and to provide the quick, contextual responses.

What Verint does in our platform is we bring all this behavioral data that is usually siloed in a lot of different places in the enterprise. But we bring all this data into one data hub, and we allow the bots to train in that data hub. We call it the Data Gym.

Michael Krigsman: Dan, you've alluded to this but can you elaborate on the collaboration between humans and bots (in practical terms today)?

Dan Bodner: Yes, I'll give you an example of the Coaching Bot. Supervisors are not there in real-time during the call, so the Coaching Bot is available in real-time during the call to assist the agent. We introduced the Coaching Bot the first time about four years ago, and we sold a few. The agents turned them off. They didn't want to use them. Their frustration was that the Coaching Bot showed up at the wrong time, giving me the wrong answer. "I already know what I'm doing. I don't need it."

This was a failure of technology because it was not implemented with this concept of people and bots working together. We learned from that mistake, and we created our next-generation Coaching Bot, which is designed to only help people that need help.

If we know that you are struggling with empathy because you're constantly getting bad feedback from your supervisor that you're not empathic, you have a very low score in customer sentiment, the Coaching Bot is only going to show up for you and point out, "Hey, you need to change. Your customer sentiment is negative. You need to change your tone. You need to slow down. You're speaking too fast for this customer. He's an old person." These are the kinds of suggestions that we only provide to agents who need that.

On the other hand, if you're constantly getting very high scores in customer sentiment, the bot will not provide you with any suggestions. It will not disrupt your call because, again, calls are a real-time environment. Agents are busy trying to solve problems; they don't need bots to disrupt them.

Every bot that we have only has one job. That job is very, very well-defined. So, once you know the job, and you know who is the person they are assisting, then you have to design that workflow so they assist the person in a way that the person is receptive to that assistance.

The Coaching Bot is giving compliance coaching. For example, in some industries, in the first 30 seconds of the call, you have to remind the customer of some compliance statement. Maybe the side effects of the drug. After 29 seconds, if you did not do that, the Coaching Bot will whisper to you, "Hey, make the mandatory statement." You will appreciate that because all you need to do now is read the statement.

You're compliant. Your company is compliant. Your supervisor is not going to complain that you forgot. It's very, very timely and contextual assistance that creates value to the agent and creates value to the company being compliant. That's how you create collaboration between AI and people by designing the AI from the get-go to know what it's intended to do and to do it in a collaborative spirit with the person they're assisting.

Michael Krigsman: Dan, what's the best way for an organization to implement a program of CX automation?

Dan Bodner: We tell our customers, "We love you the way you are. Now let's add AI on top."

Being able to deploy an open platform that works in your existing ecosystem with no rip and replace anything day-one, just adding AI to augment what you do, embed it in your existing workflows, obviously train this AI on your data because your data is the key, then just give your workforce and your business processes help from AI to get it done faster, better, with less cost, and with an elevated customer experience.

Approach AI with an open platform. Approach AI with augmenting your existing ecosystem. Insert AI to help your workforce to do things faster and better. And most importantly, get pace. Get AI to deliver tangible AI business outcomes now. This is not just architecture and foundation and infrastructure that will benefit you in two, three, four, five years. These are AI business outcomes that can be deployed today.

We advise customers to insist that they see very tangible business outcomes in 60 or 90 days. If it takes longer, it means that they are involved in an infrastructure project that they may not see results from.

You start small. You may start with only a small volume of data going to AI, so your cost is limited. You prove the results, and then you scale.

Michael Krigsman: What metrics or evaluation points do you suggest that CX buyers use to evaluate these programs?

Dan Bodner: You have to first measure the increase in capacity, and second is the increase in customer experience, which you can measure in CSAT or NPS or any of the metrics you're using today. But you should measure and see improvement in CX as a result of AI. The combination of both, when you're able to lower your costs and increase CX at the same time, obviously, that's the ultimate benefit of CX automation.

Michael Krigsman: These are hard metrics. These are not just quantitative and anecdotal stories.

Dan Bodner: These are hard metrics, tangible, and this can be measured by every single bot that you implement. This should not be just at the end of the program that you will see those results. For every bot, you can measure after 60 to 90 days, and you need to be measuring hard benefits to your company.

Michael Krigsman: Dan, do you have any advice on the preparation of data? Obviously, data is the foundation and that's an important part of any deployment.

Dan Bodner: Yes. When you have good data, you will have good results. When you have great data, you will have great results. The same AI, the same solution, training on more data and complete data will give you much more powerful results than if you just have good data.

My advice is: invest in data. Many companies have decided to invest big dollars in moving all the data to a data lake. It does not work for AI. The data lake is just too removed from the day-to-day of the contact center, so it's very hard to train AI in real time on real-time data where all your data flows to some huge data lake.

Many companies are now experimenting with AI on the data lake, which could be very useful when looking for trending information. But CX is a real-time environment. The most important trend is what happened in the last hour because that's your opportunity to improve CX in the next hour at a lower cost. Your bots need to be trained on real data in real time.

One of the solutions that we provide at Verint is that we collect data from all the different silos into a data hub that is used for the bots' training, and we also allow our customers to collect the data from the data hub to their data lake. We are like a gateway to your data lake, but it's not just a gateway. It's also where the bots are spending the time to train.

Michael Krigsman: Dan, any final thoughts on CX automation in the contact center?

Dan Bodner: We changed the name of the company from the "customer engagement company" to the "CX automation company" because I feel that AI is not just technology. When it's harnessed in CX, it's an opportunity to do what I've been trying to do for two decades, which is changing the industry to be able to give a better customer experience.

The industry has tried to do that many times, but it was always very costly. It always required hiring more people, which was not sustainable. I believe that we're now at the beginning of a transformation where you can (with less cost) elevate CX and create that brand loyalty that will differentiate you from your competitors.

Michael Krigsman: Dan Bodner, Founder & CEO of Verint, thank you so much for taking the time to speak with us.

Dan Bodner: Thank you. It was a pleasure to be here.