Verint® Systems Executive Interview
Click image below to view the ebook
In an interview with
Sheri Greenhaus, Raj Balasundaram, VP at Verint, discusses the role of AI in
enhancing customer experiences. They explore how AI components work together,
address various customer challenges, and emphasize the importance of domain
expertise. Raj also highlights the potential for AI to transform contact
centers into revenue centers and the significance of data management for
Please tell us a bit about your background and how you came to Verint.
Raj: I'm electronics engineer
who started in robotics; back in the day, I did what was called neural
networks, which is the old school artificial intelligence. That's always been
my passion and funny enough in the days when I graduated, you didn't tell
anybody that you did robotics or artificial intelligence because people would
make fun of you. Since robotics engineers were considered as nerds; we would always
say we were in mechanical engineering. I’ve seen the evolution of AI through
the years, especially in the last 10 years. I've built HR platforms for SAP,
before that I've worked in AI in the field for Oracle, especially in the SAS
world. I've been in the software service industry for a while now, especially
in the CX area.
is revolutionizing the CX automation market, especially with AI and data at its
core. I am on the commercial side of engineering, providing solutions to
customers. My team works very closely putting together architecture, solution
design, and helping the customer implementation.
AI is omnipresent. You can’t open up your email without getting an invite to a
webinar that discusses AI. What is Verint doing that is different?
Raj: The number one thing
about AI is it’s not just the algorithms or the models, or how sophisticated
the tech is. The most important thing is the application of AI and what outcome
it actually provides. And that will only
come from two things: 1. understanding the depth of how AI can be applied and 2.
how the industry works. It’s a combination of fully understanding the domain, understanding
how AI works, and fusing it together.
where Verint comes in. Verint has been in the field for the past 25 years. We
have over 60 to 70+ patents associated with Verint AI. The depth of tech, and
the depth of domain knowledge help to create what we call bots, which helps
with business outcomes. And that's very unique. AI open source is pretty much free at the
moment. There are ridiculous amounts of models, but that doesn't make you a
pioneer. What makes you a pioneer is how you apply the tech to make sure it's
available to consumers and the people who are serving the consumers in such a
way that they don't even notice that they're working with AI.
Sheri: It’s about
the application. What applications do you find are of the most interest in the
contact center? Are they looking to save costs? Are they looking to increase
productivity? How do they want to use AI?
Raj: From a consumer's perspective, we're
looking for the best service; from the brand's perspective, you want to do the
same. And most importantly, you want to do it in the channel in which the
customer's requesting that service, whether it be over phone, SMS, chat or
another app. This puts a different strain on brands where they have to staff
people in a certain way.
This brings up
two challenges: with shrinking budgets, organizations cannot staff all channels
with adequate level of service that the consumers are expecting. Traditionally,
to reduce costs, service was reduced. But what if with a hybrid (agent and bot)
workforce, you could do both; the staff stay status quo but service level
increases? That's where AI is coming into play. AI Is taking the automation to
a whole new level where we're finally at the point where the hybrid workforce
can come together. And that's where we are with AI this point in time.
Sheri: For a
company that says: “I'm buying into it, I want to do something with AI, I want
to compliment my agents.” Where do they start?
Raj: For customers, we do what is called a
maturity assessment, which is a discovery. We work with a customer in
understanding what their complete CX map is; how they are interacting with the
consumers, where they have gaps, and where they need extra help, either with
staff or with AI. We'll start with low-hanging fruit then advance to the higher
complex and more sophisticated projects. The next step is retraining or helping
the existing agents to work with bots. It's not a question of changing what
they do, but helping them to work with another entity, which can help them work
example, a customer who calls in to change a policy. The agent is new and would
normally not be able to work with that level of complexity. From a bot’s
perspective, the bots have seen all the patterns. They're always up to date -
that's exactly how AI works. When you have an agent assisted bot, the bot using
the knowledge base, recognizes the pattern. It understands what the customer is
asking. With the bot readily supplying the information, the new agent is not
wasting time looking through databases or asking for a supervisor to find the
answer. Bots work in tandem with every agent. And perhaps the best part, the
bot actually adapts to the need of that agent and their level of skill sets.
We're training the agent to work with the bot to provide a superior level of
does the bot know the level of the agent?
Raj: There are a couple of indicators. Let's
go through the back end of what's happening. As the agents are listening to the
consumers, the bots are also listening to the consumers. The same way we
transcribe in our head what the customer is saying, the bots are also
processing the information, except bots are also looking for what the agents
are doing. The bots can tell if the agent seems confused by watching the agent
search for answers, be non-responsive to the customer or even doing something
incorrectly. That’s how the bots actually figure out how to help. In the pharma
industry, for example, if somebody's reporting a symptom, you have to go
through a very compliant workflow. The bot recognizes if the agent is doing
something incorrectly and alerts the agent.
Bots look at
the patterns of successful agents and learn from them. If an agent is not
exhibiting those patterns, the bot course corrects the agent by advising them
on what to do. The bots see all those patterns and help the agents to be
the case where the agent's not in compliance. Does information pop up on a
screen to let the agent know what to do?
Raj: Typically, information is available on
the desktop all the time. There might be indicators, pop-ups, color screens. It
automatically brings up data points about the customer, such as the person
calling is an important customer and has cases pending. It’s a true blended
workforce and it's not replacing anyone. It's mainly enhancing.
a blended workforce (chat and agent) is seems that the information is delivered
quicker with more accuracy. Is that correct?
Raj: That's right. In the industry the
channels have become very agnostic. It doesn't matter whether you're chatting
with someone, whether you’re SMS texting, or whether you're on social media
channels. Now, you can actually have a chat with a bot first, then if required,
the bot can bring an agent to the conversation. There are now three ‘people’ in
the conversation. The agent, the customer, and the bot. The bots continue to
listen. Once the call is ended and there are post-call duties, the bots take
over, and the agent can go to the next call. The communication is faster. In the past, we would have chatted with a bot
and with the bot not being able to help, the customer frustration increases,
especially when the customer has to start from the beginning with a live agent.
do you think organizations are with the process of not forcing customers, when
they go from chat to agent, to repeat what they have previously said?
Raj: It is pretty recent. I think the number
one problem is bringing the data into one central place. Since the data is not centralized,
they're not thinking about channel-agnostic conversations. Many organizations
still separate the person that calls-in with the customer on chat or social
media. It is a decades long problem.
revolutionizing that space with a completely channel-agnostic technology; one can
interact with customers across the channels. For example, similar issues come
up whether the customer is on chat or on the phone. The system will instantly
recognize the issue across channels because it identifies the pattern. Channels
are not isolated anymore, because the bots are listening across channels. They
are in one application.
mentioned that data silos have been an issue for decades. Is the bot listening,
picking up data from every channel, and bringing it to a central location? If an
agent is on a call or on a chat, are the bots ‘thinking’: "I have heard
this before, I am going to supply the information that I know to the agent?"
Raj: Essentially that's what it is, whether
it's a voice channel or any channel where we're receiving information. It's
brought into a single place within Verint across the call center ecosystem,
whether we're talking about agent performance data, HR data, or the voice
calls, and the chats, and everything comes to one place. And that is where the
AI trains and looks for patterns. AI is all about looking for patterns and
finding solutions for patterns. It’s as simple as that. I think people are
making it more complicated than what it should be because artificial
intelligence simply takes unstructured data and looks for patterns and finds a
solution. So, when queries and interactions are happening across a lot of
different channels, the data comes to a central place where you then have a
singular entity. Patterns are identified. The solution is then available for
any channel, any agent. And that's the simplicity of it.
the AI start to ‘think’ for itself as a human draws conclusion or it is it
constantly looking at data and drawing conclusions from patterns?
Raj: AI has several components. When a voice
call comes in, it uses natural language processing to convert voice into text
and it's not interpreting what the customer is saying. That's one part. There
are other entities or algorithms within AI which figure out based on a selected
data set, these are the things that could happen, like forecasting a workload,
all these different AI components; think of them as Lego blocks. We take
different Lego blocks, probably hundreds and thousands of Lego blocks, and
the voice call comes in, there's an AI entity deciphering and converting into
data: such as when customers are calling, when agents are coming to work,
etcetera. The combined data figures out
the forecast. It is a string of entities coming together to make it work but
you need domain expertise to have it work correctly.
One way to look
at domain expertise is to use the example of building a house. Let's say you have a huge workforce ready to start.
Despite the best of the carpenters,
electrician, plumbers, you cannot build a house because you don’t have the
domain expertise on how to do it. Whereas if you knew how to build a house,
instead of hiring ten thousand, a thousand may be sufficient. The domain
knowledge is huge in the application of AI.
Sheri: What do
you find customers are looking for?
Raj: Different customers have different
problems. Some customers need a rapid expansion of agents but lack the funds to
accomplish it. This can be defined as a problem of scale.
Others such as huge
insurance companies want their agents to be more effective at what they do. Even
scheduling a shift for them is a nightmare. That's a different problem. In that
case, we apply different algorithms and different techniques to make sure that
we optimize their workloads and forecasting schedules. It all starts with the
the problem in a CX ecosystem is not easy. After defining the problem, the
solution is fairly straightforward because the platform does the majority of
the solutioning itself. We have solutions for many problems. If you have an
issue of containment, where you don’t want a lot of calls coming in and you
want to contain the customers to be within a chat channel or an SMS channel. We
have a straightforward application, a bot, which goes with it called a
containment bot. For another example, if agents need to be more effective with
phone calls, we can apply a Wrap-up bot.
sounds like using something like a Wrap-up bot would be a win to start with. It
can immediately create savings.
Raj: 100%. It’s a quick calculation to show
how to save 10 minutes per agent per call. The application is straightforward. This
is also unique to Verint in terms of how we do wrap up. Because we are there
during the call the bot ‘knows’ when to take over for wrap-up.
is always a concern that AI is going to take away jobs. What are your thoughts
Raj: It will enhance not take away jobs. And
most importantly, it's going to increase the revenue. It’s going to change the
contact centers, which have been seen as cost centers in the past, into a
revenue center. With a hybrid workforce, we can serve customers better.
Customers will, in turn, be loyal to the brand. AI is going to enhance what we
do on a day-to-day basis, make us better and increase revenue.
there anything that we have not touched on that you think is important that our
Raj: From an AI perspective, how we think
about data is very important. So currently, we have a lot of enterprise
customers doing massive data warehouse projects. I sometimes say Data Lakes and
the Data Warehouse projects are the place where data goes to die. They are
customers to think about end-to-end CX platforms. Think about platforms like
Verint as an engine which collects data across the CX, which helps you to
understand the pattern. Verint is a perfect complement for a CRM platform. Within
the CRM platform, you generate data from customers from your Salesforce, and
all of the data comes in. Whereas within Verint, we collect the data
downstream, which is the interactions that we have with customers. So, instead
of putting all the data in one place, we use these platforms to bring
everything together, and they will provide way better CX in that case.