Dreamforce 2025 | Rohit Khanna, Smarsh

Clips
More from Dreamforce 2025

Rohit Khanna

Chief Customer Officer

Smarsh

play_circle_outline Ensuring Compliance: Smarsh's Role in Capturing and Archiving Financial Communication Data for Regulatory Surveillance

play_circle_outline Enhancing Misconduct Detection: AI's Role in Managing Massive Data Volumes and Reducing False Positives with Smarsh's Intelligent Agent

play_circle_outline Enhancing Customer Experience Through AI: Smarsh and Agentforce's Strategies for Support and Case Assessment Optimization

play_circle_outline Navigating AI Integration in Regulated Industries: Emphasizing Change Management, Community Support, and Enhanced Productivity in Compliance

Info
Transcript

Rohit Khanna, Smarsh

Rohit Khanna

Chief Customer Officer Smarsh

In this Dreamforce interview, theCUBE’s Gemma Allen speaks with Rohit Khanna, chief customer officer at Smarsh, about governance and trust for AI in highly regulated industries. Khanna explains how Smarsh captures and archives communications across 100+ channels and investigates misconduct in line with FINRA/SEC rules, including seven-year retention. With copilots and agents now summarizing conversations, he raises a key compliance question: should synthesized outputs be treated as regulated records that must be captured?

Khanna outlines why AI is esse... Read more

explore Keep Exploring

What services does Smarsh provide in the financial industry regarding communication data? add

What challenges does AI address in identifying misconduct in communications within a regulated industry? add

How is AI being utilized in customer experience and support? add

What is the impact of AI on productivity and workforce roles in regulated industries such as finance, government, and insurance, especially regarding compliance checks? add

bolt Powered by CUBE AI

Rohit Khanna, Smarsh

search

Gemma Allen

>> Welcome back to theCUBE, live here at the Moscone Center in San Francisco. We're here for Dreamforce, where we're delving into all things AI and enterprise. One key topic that we're going to talk about right now is governance and trust. Joining me have Rohit Khanna, chief customer officer at Smarsh. Welcome, Rohit.

Rohit Khanna

>> Thank you very much.

Gemma Allen

>> So, tell me, you work with some very highly-regulated organizations. And admittedly, there is some skepticism and concern around what the future of AI means for enterprises, especially as it relates to security. Talk to me about Smarsh and where you really add value for your customers in that space.

Rohit Khanna

>> Sure. Well, first of all, thanks a lot for the invitation here. We are all having a great time at Dreamforce. Smarsh mainly plays in financial industry in the highly-regulated space. We provide intelligence from communications data, which we capture and archive. That's the main part of the business, is capturing the data and archiving the data. Now, while we are capturing and archiving, we are also doing surveillance on the data. We are also looking at any misconduct happening in the data. And that misconduct can happen in any channel. It can happen in your SMS, or it can happen in your email, or it can happen in your Teams, Zoom, doesn't matter what it is. So, by law, every regulated user in a financial industry, so this is most of the bankers, investment bankers or regular bankers, they have to capture their data and they have to archive the data for about seven years. And after that, it's up to the bank what their policies are. Now, back to your question. Now, this regulated industry follow the regulators, like FINRA or SEC. Regulators define the regulation or direction, and then they want to make sure that they're not fined. Right now, AI is a big topic with regulators. I'll give you an example. Copilot is out there now, or any AI agent is out there and Copilot, Gemini, you decide. Now, the summarization is happening with many of these agents. Is that summarized data regulated or not? That's a big topic with the regulators. Should we capture the data or not? That's, again, a topic which regulators are deciding. You see, for example, we are talking, there's an agent along with us. Now, agent is looking or hearing us, and agent is now summarizing and/or adding its own statement. Is that regulated? Should we be capturing it, right? Because every data you capture, there is a cost to capture. There's a cost to archive. So, AI in a regulated industry, it's absolutely a big topic. Second thing is in a regulated industry, what we are trying to do is find misconduct in financial crime space. Now, misconduct can happen anywhere in any communication. So, in context, conversation is very, very important. So, we think AI will fundamentally change the way we are capturing, archiving and finding misconduct from this data. Why? Today when we are looking at the data, less than 3% of the data is being surveilled because there's so much data, especially after COVID, people are working from home and they have all kinds of collaboration tools. It's just not emails anymore. It is email, SMS, chat, you name it, 100 plus channels. How do we find what's happening in each channel? Or is there something happening off channel, point one. Point two, if I'm having a conversation with you, let's say, and I send you an email and say, "Hey, don't tell anyone. Let's talk about it at a cafe outside the building." Now, is that misconduct? Maybe, maybe not. Maybe we are planning our colleague's birthday party. However, when the system look at these things, it thinks there's a change of menu because we want to execute on some misconduct for a financial crime. So, I'm just giving you an example of that. There are many other examples like this where what we call false positives happen, where our conversation raises an alert, but that alert is really a false positive because today, people don't understand the context. Now, here comes AI. AI is contextual, as you know. If we employ AI agents the right way, which by the way, Smarsh has rolled out its own AI technology, we have rolled out our compliance agent, we call it Intelligent Agent, and we are now just about rolling out of a new Deep Search Agent to go find that needle in the haystack of that misconduct.

Gemma Allen

>> That is really so interesting, especially the synthesis of LLMs, right? I remember the first time I saw that happen on a WhatsApp chat, where a chat between my husband and our neighbor about our kids' behavior, it was summarized, right? And it read so much stronger than probably the incident was. But one thing that I think is really interesting about synthesis as well is thinking about how we as employees can also search and discover our own data. And I know that's a big, big element of conversation here right now. And I think Agentforce and this agent-to-agent opportunity is huge for that. But these systems, and we've talked offline about some of your competitors that I have some experience with, they can be clunky and they've been clunky for a long time.

Rohit Khanna

>> Correct.

Gemma Allen

>> Tell us about the customer experience, about the UX, about that discovery element of Smarsh.

Rohit Khanna

>> Very good example. Great question. So, how are we using agentic AI today? What we believe in customer experience and customer outcome is customer success end of the day, right? So, when we started implementing AI, from a customer experience perspective, we first utilized a use case or we utilized AI on a use case for internal purpose, that is for our customer support reps. We have about almost 400 support reps all over the world. We gave them an AI agent. We said, "Okay. Utilize this agent to find answers to the complex situations, to the complicated questions." That's how we started. Once we got a little comfortable, then we rolled out other use cases, like case summaries, like automatic case assessments, creation of knowledge base articles. So, once we got comfortable, we said, "You know what? This is the time." We switched it open for our customer. So, we used to have chatbot before. We replaced the chatbot with Agentforce agent. Now, I always say we got lucky, and let me tell you how. About four or five years back, we embarked on creating KCS, which is knowledge-based articles. We were using AI to create those knowledge-based articles, and we had created incentive for a lot of technical support reps to create that knowledge. Fast-forward four or five years back, we had millions of articles in our database, which was cured, curated, and ready to go. Now, we did not know about this Agentforce coming up. It just happened, Agentforce came. Now, you may have heard many AI projects are failing and/or are in pilot phase and not gone production. The reason why, it's not the technology, it's the data. Do you have curated data? Do you have the right data which Agentforce can go and learn from? So, when I say we got lucky, we had the data. And so, when we put the agent, it just worked out fantastically. So, now today we have Agentforce for our customer giving the best customer experience, and then we have agent four for our internal use case too. And by the way, I just want to let you know we named them too, and in the beginning there was not a lot of adoption because customers were not sure whether they should use this agent. But guess what? We named those agents. So, our external agent is called Archie. Like from archive, Archie. And our internal agent is Emma. So, as soon as we named it and we personalized it, suddenly the adoption rate went up by 76%.

Gemma Allen

>> Wow. So, tell me a little about data, because we all know it's a fuel for AI. We also know that there's a lot of challenges, and you work with a lot of companies that have a lot of legacy data, I'm sure, especially in that regulated space, in banks, in government, in areas where there's always been challenges with agility. AI has suddenly now just come upon us and there's this expectation that you get with the program or you become dust, right? How though are those bottlenecks that have existed for decades suddenly going to be fixed as it relates to structured data and the opportunity to turn data into true fuel?

Rohit Khanna

>> You ask a great question. I think that is the fundamental piece of what I would call the change management. Look, AI is here, and AI will be here. It's going to change everybody the way they are working. It's going to change every system. There will be agents and agentic workflows all over our regulated customers, both financial industry and the government and insurance, they know, they understand that this is coming. Now, what they are looking at it is from productivity perspective. They know it's coming. They're very risk-averse, as you would know. They're following what regulators are saying. But at the same time, they're saying, "Hold on. Is there a way we can perhaps not have that many level-one people who are looking at the data for false positives?" Many of the financial institutions, for example, they have hundreds and thousands of people all over the world literally looking at the false positives, thousands of false positives every day. And they say, "Nope. No, it's not right. It's not right. It's not right." Now, even if we can give them 20% productivity gain, that's millions of dollars of cost takeout for many of these financial institutions. I think that's what AI would do for them.

Gemma Allen

>> Let's talk a little bit about those humans who are spending that time on compliance checks, working in those roles where there has to be some level of human intervention at times. In this agent-to-agent world, one of the key challenges, I think, is how do you keep humans in the loop, keep those guardrails and then also act in a way that has speed, agility, promise, all of the good stuff that we're hearing about?

Rohit Khanna

>> Of course, I'm smiling because I'll tell you a... So, when we named our agent Archie, I often told my team, "Hey, just want to make sure that Archie is a teenager. Archie is still learning. What do teenagers do? They get drunk."

Gemma Allen

>> Archie's hormonal? That's okay.

Rohit Khanna

>> They get drunk, they get their DUIs, which is the hallucination, what I'm talking about. So, we have a team in place for making sure we have a human in the loop to make sure that we are checking every answer. And yes, we are correcting. I would say it started with almost about 72% accuracy. Now, it's around 90%. Still not 100%, but we are checking. We are making sure it's learning. So, Archie is growing. Look, I have two kids. Now, they're in college. I remember their teenager age because you have to be around. The human in the loop needs to be there. For example, you would not let your teenager just go at whatever he or she wants to do, but you cannot let your agent just go at it. So, now we have our team who's taken care of Archie. In my eyes, our Archie's about 19 or 20 years old now, just getting to be an adult-

Gemma Allen

>> Still living at home.

Rohit Khanna

>> Still living at home.

Gemma Allen

>> Still having his mother do his laundry.

Rohit Khanna

>> Absolutely. But it's getting there. I can see that our Archie's getting ready and same goes with Emma. Emma is our internal agent. She's also getting there. I think another year, they are ready to go and I don't think we would need that much of . I mean, right now I have a team in place, almost about 10 people. I think that team is going to be about two to three people, will still be always there. Actually, if you don't mind, you raise a very, very good point because I think you talked about it is the data governance. So, Archie's learning, answering questions, but sometimes Archie says things or gives information which we do not want Archie to give, internal information because what Archie or Emma they do is they go inside the data, their own LLM at the same time inside Smarsh ecosystem and get the answers. And we have seen, "Oh, my God, we should not be exposing that to our customer." By the way, I'm so happy that finally Salesforce have policies engine, so we can create those policies now, so that when the data is being rendered by the service agent, those policies kick in, and so that policies stop. If it's like, again, "Hey, do not say this. Do not disclosure." So, the policies stop Archie to say things they should not be. I think I'm so excited about that feature. That's the data governance features which Salesforce just came out.

Gemma Allen

>> Well, let's shift gear to Salesforce for a second then, because we're here at Dreamforce. It's an incredible setup.

Rohit Khanna

>> Fantastic show.

Gemma Allen

>> The vibe is unbelievable. The amount of people, it's like mind blowing. Right?

Rohit Khanna

>> Right.

Gemma Allen

>> Tell us about your experience with Salesforce partner, customer. Tell us what excites you most about the announcements from today and yesterday, how will that affect Smarsh particularly?

Rohit Khanna

>> Sure. Look, Smarsh is a SaaS vendor, very much like Salesforce. Obviously, Salesforce is the pioneer, and we are all in this space, are learning a lot from Salesforce. But this is the way you do the conference, no doubt about it. Last year when I came over to Dreamforce, that's when I learned about Agentforce and my team. I mean, that's how we learned where to go, what to do. So, Dreamforce, for me, it's just not about party, concert, great food. It's all learning. It's about understanding what other people are doing because we go back home, we are in our own bubble, we do our own things,. We have our own objective, but we do not get time to go look at what others are doing. We do not get time to inspired by other stories. So, for me and my team, we look at stories. We go hear what other people are doing, we reach out to people and say, "Hey, you know what? I heard you on the stage. Tell me more about how did you do that?" So, Dreamforce for me is like a turbo education in two days, three days, where you're taking all these ideas, you go back home. For example, I have a whole strategy meeting next week with my team where I'll say, "Okay, what did you learn? What are we going to implement?" This is how we implemented Agentforce. We came here last year, like I said, we got our objectives. Q1, we did a proof-of-concept with Agentforce, and Q2 we were in production with Agentforce.

Gemma Allen

>> Well, I guess above all, it's about community too, and this is definitely an exemplar of that. So, last question. You said it's not about party, but I have to ask you, are you going to see Metallica or Benson Boone tonight?

Rohit Khanna

>> Metallica, for sure. Definitely looking forward to seeing Metallica. All my team members, they're all-

Gemma Allen

>> You and me both....

Rohit Khanna

>> fully excited about that. But no, what a pleasure and what an honor, we'll be seeing them performing.

Gemma Allen

>> Well, thanks for coming on theCUBE.

Rohit Khanna

>> I appreciate it. Thank you for the invitation.

Gemma Allen

>> I'm Gemma Allen, here broadcasting live at Dreamforce in San Francisco. We'll be back very shortly with our next guest.