In this special Zoho Day 2025 episode, Robert Kramer sits down with Ramprakash Ramamoorthy, Director of AI Research at Zoho, to dive deep into AI’s impact beyond automation, challenges and the future of AI in enterprises, and how his interests intersect with his groundbreaking work in AI at Zoho.

Key topics discussed:

• The concept of contextual intelligence and predicting the “next best action” in enterprise workflows.
• Enhancing productivity by embedding AI into daily business processes without compromising privacy or sustainability.
• The importance of data and process alignment in achieving AI-driven productivity gains.
• The role of digital maturity in determining AI implementation success.
• The long-term vision of AI at Zoho—focusing on privacy, sustainability, and customer-centric innovations.

Watch the video here:

Listen to the audio here:

Disclaimer: This show is for information and entertainment purposes only. While we will discuss publicly traded companies on this show, its contents should not be taken as investment advice.

Transcript:

Robert Kramer: Hello, I’m Robert Kramer with Moor Insights & Strategy. I’m here with Ram, and Ram is director of AI research. Ram’s last name is Ramamoorthy.

Ramprakash Ramamoorthy: You got it right.

Robert Kramer: I think I got it right. But we’re in Austin, Texas, right? This is an analyst conference. It’s fantastic. How’s it going so far for you?

Ramprakash Ramamoorthy: So, far so good, Rob. A couple of days have been lot of information out, lot of information in, so wonderful.

Robert Kramer: Absolutely. So just taking a look at your background, it’s really interesting. So, you’re from Chennai.

Ramprakash Ramamoorthy: Yeah.

Robert Kramer: I’ve actually been to Chennai. It’s a really beautiful city. It’s on the ocean, right?

Ramprakash Ramamoorthy: Yeah, right across the beach, and the Dosas and filter coffee are perfect. You should come over.

Robert Kramer: I should come back over there. So Ram, you’re also… I took another look at you, you love aviation, and turns out you told me right before this that you were at Reagan International right before there was an incident there.

Ramprakash Ramamoorthy: Yeah, pretty unfortunate incident, but I was just about two and a half hours ahead of the incident.

Robert Kramer: How did you get involved in your hobby or your huge interest in aviation?

Ramprakash Ramamoorthy: Right. So I mean I joined Zoho fresh out of college, and you had so many things to do in college. There’s a lot going on, but then when you get on to work, it starts getting monotonous, and then I did everything that everybody does, I went running, I went cycling, I tried swimming, and then, I started taking up a couple of flights early in my career and-

Robert Kramer: Flying yourself or just flying?

Ramprakash Ramamoorthy: No, no. Just flying commercial aviation. And then I realized that it’s an industry with a lot of constraints. So every gram you are add onto the flight is something, and every part is manufactured to stringent quality standards and everything is hyper-regulated. So I was able to draw a lot of parallels. The frugality of aviation is something that we should all apply to whatever we do. I mean, there’s nothing infinite there, but then it also keeps evolving. You could have the best service in the world, you could have built the best fleet in the world, but all of a sudden you’re able to filter flights by carbon emissions.And think about it, you’re able to do the best of everything at great prices, but your ESG scores are bad. So I love the way how constrained it is and how competitive it is and how regulated it is, and despite all the regulation, how do you innovate. And very little differentiation you can do in the hard products, because it’s a duopoly. You’re either an Airbus or a Boeing, but how do you differentiate? So lot of such interesting management dogmas I got from it. And then I also love flying. Thankfully, my job takes me to places, so I try different airlines and I try different aircrafts. I have a log of everything. I have a gold subscription with Flightradar and I’m not watching cat pictures when I’m idle. I just open up Flightradar and look at the flights around.

Robert Kramer: So what’s your favorite airline?

Ramprakash Ramamoorthy: That’s a bit of a controversial question I would say, but I think I love United for their touch points for their digital platform. Everything just works-

Robert Kramer: They have a great app, by the way.

Ramprakash Ramamoorthy: Yeah, great app. Great app. And there are instances where there were weather delays, but everything was on my app and my bag reached on a different flight, I reached on a different flight, but at no point I was left aloof. I like United for their digital products. I fly a lot of Emirates, because I mean, India is connected to the world via Emirates. I like Emirates for things just work, okay? I mean, no big app, but things just work seamlessly. And the third I would say, I love Singapore Airlines because whenever I fly east from India, their service is amazing. So top three, that would be the top three.

Robert Kramer: Great. So obviously that’s taken you to the AI world, which you love, and you spend a lot of time with research and development with AI, and Zoho obviously has a lot of interest in developing AI and automation and new things. So how do you envision AI creating the most value for the organization, but beyond automation? How can AI take Zoho to the next level? Because I think when people look at AI, they say automation or it’s going to do tasks more efficiently, but what do you see the vision for it at Zoho?

Ramprakash Ramamoorthy: Right, right. Great, Rob. I think I’ve been doing this for 13 years now. When I joined the company as an intern, I was tasked to see if AI would really fit the enterprise, back then it was ML, and it was a very consumer-grade technology back then. The only people who were talking about ML and AI were Netflix and the Amazons of the world where they were doing recommendation engines, they were doing people who got this also got, and my job as an intern was to see if it would really make sense in the enterprise. So we tried building a sentiment analyzer for our customer support product. Back then it was called Zoho Support, now Zoho Desk. The idea was to see if we can identify the sentiment of incoming emails, and then if a support manager has only about time to read 10 tickets a day, then maybe let him read it from the angriest of his customers. So this was how we started, and then we spent quite some time… And this was also the time where the transition from monolith infrastructure to things like MapReduce, things like Hadoop where you started commodity computing on distributed mode, that was taking over. So the whole notion of big data, small apps was evolving. We looked at a couple of things and the first set of launches that we did were nothing fancy, but we tried replacing existing tools that use statistics to use machine learning.

Robert Kramer: And how long ago was this?

Ramprakash Ramamoorthy: This was in 2014, ’15 ballpark. So we tried replacing our anomaly detection features, which were based on static thresholds. You call something an anomaly. If it’s in the bottom 5% and the top 5%, you just flag that as an anomaly. But a Monday morning 9:00 AM is not the same as a Saturday morning 3:00 AM. So now can you learn the behavior from the past data and then flag it as an anomaly? This was how we started, right?

Robert Kramer: Sure. Makes a lot of sense. That’s how people should look to differentiate data and that’s what you’re looking at. Okay.

Ramprakash Ramamoorthy: So this was how we started. Again, back then the conviction was, okay, we didn’t want to miss out this bigger trend that is coming in, so let’s try and do it. I mean, it was more of a spray and pray approach for us when we started. Let’s try and see if it really provides value. And then fast-forward to 2017, ’18, we had recurrent neural networks, LSTMs coming into the picture, the deep learning architectures started taking shape, and then that’s when we started a receipt digitization project where you click on a photo of a receipt and then you extract all information. Then we started looking at grammar translation. And we had a mode, especially for grammar and translation, this was a time when GDPR became very prevalent and a lot of economies, a lot of regions were trying to bring in their own privacy laws, and a lot of our customers were using third-party browser extensions, tools that were built for the consumers to translate. They were just copy-pasting sensitive enterprise information into those consumer-grade tools so that they can correct the grammar, they can translate it to a language that they’re speaking to their customers or colleagues. And you see the privacy differentiation there. Now, when all of that sits within your enterprise app’s editor, now that’s a very big privacy differentiator.

And this was also the time when we had the likes of Alexa, Siri and all these chatbots going big. That’s when we launched Zia. I mean, we started branding things. We were doing a lot of AI, but then bringing it under one cohesive brand and we were building up on the machine learning strategy, the deep learning strategy, and then come 2022, what happened, ChatGPT and the whole generative AI started coming in, and the dynamics changed because the algorithms of 2012, 2013 were so complex. I mean, you had to have a data scientist to tune the parameters, the number of trees, the random parameter that you set, and it was a difficult job. But today, you have a model architecture throw your data, throw some compute, and there you have a good AI recipe out there. So we are transitioning to the LLM era. And the thing is, this technology changes really fast, but if you look back and see what are the constants, one thing that has been constant is the availability of ever-increasing compute. Compute has exploded in the last 20 years, and there are reports that suggested compute becomes 10x cheaper every five years for the same amount of compute to inflation-adjusted dollars.

So the thing is these models are getting bigger and there’s a lot of productivity coming in. I gave you more of the technology landscape and how we evolve. So what is our mode? We see that the technology is a lot hyped, but we are trying to play the long game here and we’ve done the same with everything else. Think about it. When I joined the company in 2011 as an intern, my first shock was, “Hey, why would you invest so many people and so much money into building data centers? Why would you…” I mean, it’s a collocated data center, but we get the bare metal from vendors and right from the operating system that runs in these data centers is custom-made. Why spend hundreds of people to put together an operating system? And it didn’t make sense to me back then as maybe I was naive, why not run on AWS? Why not run on GCP? You can scale up, you can scale down. But today, 10, 12 years down the line, if I look back and connect the dots, it’s a game-changer. It’s a differentiation that we have, and most modern day SaaS companies are just resellers of AWS.

Robert Kramer: It’s also sustainability advantage as well, or asset of sustainability, and it goes back to what you were talking about with aviation and the carbon emission. I mean, this all kind of coincides with your personality and your strengths and what your strategies are, but continue, and then I got a few more questions for you.

Ramprakash Ramamoorthy: Sure. Yeah. So yeah, so you look back and see that the data centers are remote now. They’re a differentiator now. Same thing with the AI strategy as well-

Robert Kramer: They’re changing. There’s a lot going on there.

Ramprakash Ramamoorthy: Yeah, and we call it contextual intelligence. The idea is it’s not standalone intelligence. It’s embedded deeply within the context of the product. If we were to summarize it just one sentence, it’s about predicting the next best action At any given point in time. A high value lead has signed up to evaluate your product. What’s the next best action? There has been an escalation in the customer who has not been happy and they would possibly churn. What’s the next best action? So in the broader scheme of things, we don’t want to hype AI, but we want to give the next best action. You’re going to a meeting, can I get the summary of all the records with the customer? You’re doing a standup, can I get all the summary of details and similar projects, and prep you for the meeting. 30 minutes before the meeting you get an email saying, “Hey, these are the key points. This is what you want to discuss. This is what you discussed last time.” So we wanted to be so seamless, so embedded into everyday enterprise workflows without affecting privacy, without doing much of carbon emissions and bringing in context. So the platform is the broader story, the AI is just an add-on to it.

Robert Kramer: Exactly. Excuse me. A little cough. Okay, so what’s also interesting is my background also flows into data, and I’m a huge advocate of change management. I feel like companies, and I have a longer question here, I feel like companies don’t understand the importance of data, the importance of change management, which the change management probably is first. So when you talk about AI, do you involve the change management process to get the data management correct at companies in order to take advantage of everything you just talked about?

Ramprakash Ramamoorthy: Right. I think data is the critical piece. I mean, this has been said over and over and over, but one thing that I’ve observed is processes are also as important as data.

Robert Kramer: Good point.

Ramprakash Ramamoorthy: Right? And so the first thing that we did, we launched agents yesterday and the first thing that we did six months back when we were prepping these agents was to deploy these agents in our own infrastructure, into our own company, and the joke is anything that goes to legal team, it’s going to take at least two weeks before you get any response, and so the first thing we did was we built a agent and we deployed it in the legal team.

Robert Kramer: To do what?

Ramprakash Ramamoorthy: Okay, so the legal team uses Zoho Desk and whenever you raise, it gets raised as a ticket, and then what this agent did is it understood the context of the ticket. Now, let’s say there is a ticket from the Middle Eastern-Africa region and it talks about privacy there. Now, who are the experts who solve… Who are well-versed with Middle Eastern… Let’s say Saudi Arabia. Who are well-versed with Saudi Arabian privacy laws? And let’s say it’s coming from a particular salesperson. Whose account manager is this salesperson’s? So there is a Saudi Arabia context, there is a salesperson context, and then there is a past information on who’s there, and then there’s a current context. Let’s say I’m going on a vacation and I’m the Saudi Arabian legal expert. I’m on a vacation for the next couple of weeks. So who is the next best person? It starts off with auto-assigning. We were monitoring KPIs, like mean time taken to respond, number of tickets that were solved in the last week, and we could see that there was around a 30% improvement in productivity.

Robert Kramer: Okay, that’s substantial.

Ramprakash Ramamoorthy: And if you see, you just open Twitter or X, the way AI is hyped is like, “Oh, I’m getting 10x productivity. I fired all my employees and then I’m able to do the same level of service,” but I think the reality in a bigger enterprise like us was a 30% productivity improvement, and I think it’s a start of the hockey stick things that would come, but it does not starts with 10x or 30x. It starts with 10%, it starts with 30%, and this is because you need your data and processes to be streamlined. It’s not about how clean the data you have, how much of processes and how do they talk even in a company like ours. So think of something like a COVID vaccination certificate. So whenever I raise a travel request, it was super important three years back, or four years back, five years back. Oh, five years already?

Robert Kramer: Sure, it was a long time ago.

Ramprakash Ramamoorthy: It’s irrelevant, but you have that in your form, right? Are you vaccinated for COVID-19? Yes. You see that the process has drifted, but then the forms that we used to collect information has not been drifted. And we are also seeing generally these SaaS vendors sell to departments. There is a vendor for the marketing department, there’s a vendor for the finance departments, and these vendors’ tools don’t talk to each other natively, and so these departments don’t talk to each other natively. And this is why things like CRM for everything, which we launched the previous year, is a good differentiator. Now, we are all working to close a customer. I’m the salesperson, I have complete access on the CRM. You are the legal team who’s drafting the privacy agreement, and then there is another accounts receivable person who is working with to collect the money from the customer. So now three have three different tools and we don’t see the actual ground truth which might keep changing. Now, can we bring them onto CRM? So can the process be streamlined so that eventually your data is streamlined? And we’ve seen that give the same AI stack to three different companies and they will have three very different outcomes from it. It all depends on digital maturity. I’ll give you a very tangential consumerish example. We talked about Chennai, and I was in a self-driving Waymo yesterday, okay? Now, take the same-

Robert Kramer: You’re here, so it worked.

Ramprakash Ramamoorthy: Yeah. Take the same Waymo, put it on Chennai’s roads.

Robert Kramer: Different story.

Ramprakash Ramamoorthy: It won’t work, I mean.

Robert Kramer: Well, that’s a little different. It wasn’t trained for that.

Ramprakash Ramamoorthy: Right. Right. No, but the AI is the same, but the maturity of the ecosystem, how well-marked the lanes are, how disciplined the other drivers are, I don’t think you can ever train a self-driving car for Chennai’s driving conditions. So the idea is same AI stack could work very differently depending on how digitally mature the organization is. So the pursuit for digital maturity where you streamline processes, where you streamline data, where you invest in existing rule-based automation, if you don’t have any rule-based automation, don’t expect to AI and come in and do an overnight magic. It’s a step-by-step improvement. It’s a percentage improvement, then a multiple improvement, and that happens better when you are digitally mature.

Robert Kramer: Kind of getting off on a tangent here. I’m a big fan of industry 5.0, which is a collaboration between robots and humans. So do you see people being replaced by AI or the jobs being more strategic with AI for the human to interact? Because all the things you just mentioned, the 30% efficiency gains, were those monitored by humans or is that pure AI, or both?

Ramprakash Ramamoorthy: I think the riskier job, especially when you talk about industry 5.0, now let’s say you’re dealing with heavy machinery, it’s better a robot does it than a human risking their lives and trying to do it. So there are certain aspects of where we would want AI to be doing the heavy lifting, but we are not there yet. Let’s talk about a desk job. Very simple thing. So today we are talking about the ticket assignment in Zoho Desk. Today, let’s say there is a person who reads the ticket and understands who to assign it to. Now, let’s say he’s on a smoke break, he’s on a coffee break. Now there’s a 30-minute delay. There’s a super emergency ticket that has landed in your customer support system, and the delay cascades, because you didn’t assign it to the right person, then you assign it to somebody, but then you have to go look up in the HRMS system, are they available? You have to go look at their pipeline, will they have bandwidth to take this up on an emergency note? Now, if you can automate that.

Robert Kramer: Fantastic.

Ramprakash Ramamoorthy: So I think the idea is you take 10% of the whole pipeline and bring the 30% improvement there, and that is all about it. And it’s still the human who is going to solve the ticket because we have seen customer support AI, and we don’t want to work with it. We want to talk to a human

Robert Kramer: Right. We’re trained to do that.

Ramprakash Ramamoorthy: Yeah, we’re trained to do that. We all want to talk to a human. So I think the productivity improvement is there, but it won’t be replacing. It’ll augment humans than replacing humans.

Robert Kramer: Good point. So where do you see the most critical future of AI and Zoho applications and challenges you might have with that type of a development? And do you see that changing drastically over the years? Because technology has moved quickly. We were talking about five years ago filling out a travel form. Now today, I’m sure people don’t do that, it’s completely automated. But AI is making such a world of difference, and like you said, it’s based on processes and data and even change management, but also culture of the company. So where do you see things from your perspective at Zoho? Because the company is making a lot of progress, a lot of changes, and actually getting deeper into the enterprise market. So what’s your thoughts? It was a kind of a large question, but I think you could probably have a-

Ramprakash Ramamoorthy: I always give large answers, Rob, so you know it by now.

Robert Kramer: So here we go.

Ramprakash Ramamoorthy: Right. So I think the way I see it is there are companies that have become an AI company overnight. No, but I think it’s a broader part of the digital transformation. We’re still doing digital transformation, and we saw this in cloud, we saw this in mobile and now we seeing this in AI. So it’s a natural evolution of technology and technology’s job is to bring in productivity to whatever we do. Let’s take a very tangential example. Take something like the music industry. We started off with vinyl records, we went to cassettes, we went to CDs, we went to downloading MP3 files. Today we are streaming, but the challenge here is the distribution changed. There was an efficiency in distribution. For example, selling CDs to make 10,000 people listen your album is very difficult when compared to 10,000 people listening to your album on Spotify or Apple Music. So distribution changed, but the time taken to create music almost remained constant. There were tooling improvements. You don’t have to have an orchestra now. You can have a bunch of iPads generating all the music for you, but still, there’s only 10, 20% improvement.

But now this is impacting creation. Today you can create music in minutes, you can create apps in minutes. So there is a big boost to software development productivity that is coming in, but there are certain things that do not change. For example, the structured information in your database, there are the compliance laws around like GDPR asking to forget the customer. These things don’t change. Now, how do you marry the space of innovation with the ground reality that we have? Seems to be the next five years for me, where here is a tool that can give you insane productivity boost. How do you rightly engage that with the enterprise? How do you rightly marry that to the enterprise? Because I was reading an economist survey that said 87% of employees don’t want their employers to know that they’re using AI. I’ll repeat it. 87% of employees don’t want their employers to know that they’re using AI. It’s such a big productivity tool and I’m sure employers will be happy-

Robert Kramer: Because it makes them look better.

Ramprakash Ramamoorthy: Yeah, it makes the broader output better, so you result in happier customers, more fulfilled, more satisfied employees. So it’s a thing to do. So we want to make sure that companies are able to ride this AI wave without any apprehension. I mean, we help them ride the cloud wave. We were shipping software in floppy disks and today we are a SaaS company. We are streaming software. So we have seen waves and we see this as one of the waves. It has an impact, but it’s just one of the waves, and tomorrow there’ll be another wave and we will ride that too as well. It all boils down to solid fundamentals. The investment in infrastructure that we have done, the long game where we have not taken any VC funding, and the goal is to put insane value to customers than getting fancied by technology and pivoting everything that we do, and-

Robert Kramer: Because everything is about the customer. They drive all industries, they drive the outcome. It could even be sports, it’s all about the fans. Did you think that technology… And this is kind of an interesting question. Do you think that technology is too much? I mean, is it going too fast or is it where you want it to be? Because you have to train people to use it and the aptitude to do that could vary and the companies could be either leaders or followers and they might not be as innovative to take the technology on if it’s so extravagant or so complex or so dynamic, but some companies are on the leading edge and they’re dying for it. But AI seems like it could fill all those voids, but what’s your thought on that?

Ramprakash Ramamoorthy: Yeah, this technology is evolving faster, and then there’s also a lot of hype, right? And we have seen this with AI. It’s a technology as old as 1950s and there are AI summers and there are AI winters. I think this is the longest AI summer we have had. It started off sometime in 2012, 2013, it’s been going really strong. I mean, one year we talk about LLMs, the next year we’re talking about agents. So it’s like the train… But I think this is just innovation at its best because I think we have very strong digital foundation now. We all use apps, we all generate a lot of data. We all have heart rate sensors, sleep sensors, and so everything is data-driven now, so it’s going to evolve at a much faster pace, but companies should take a level-headed stance. Your customer is not going to bother if you’re using AI or not, but as long as they’re happy, as long as they see value in your offerings, then you’re good.

So this applies to us, this applies to our customers as well. But use it. It’s a productivity booster. If you can reduce your mean time taken to respond to your customer support tickets by just about 20 minutes, that’s a win to start off with. So find out areas, and this is the general advice that we tell them also, find out these 80% problems. Let me explain what 80% problem is. If you want to solve a problem with 100% accuracy and you’re deploying AI there, you’re going to be disappointed. Find out the 80% problems, which AI is really good at. Something about the customer ticket assignment in Zoho desk is an example of the 80% problem. It’s okay if you assign 80% of the ticket automatically. If there is a mistake, I can always reassign. So it grows, okay? Find those 80% problems, get used to it, then go to 85% problems, get used to it, and if you are starting off with 100% problems, you’re going to be disappointed.

Robert Kramer: I agree. The interesting thing, we were talking about customers, but to them it’s all about business impact. What’s the business outcome? Do I care about the features or do I care about the conclusion? The conclusion is, like you just said, start at the 80 because I want to see results. The features are great, they sound good, but the business conclusion, the impact, the outcome is what I’m looking for. So we’ve got one more question for you. So obviously you love aviation, you told us why, and how do you think that’s influenced your role for developing technology and being interested in AI? Can you kind of parallel it and show us or tell us a little bit more about how you see your research with AI and aviation kind of overlapping?

Ramprakash Ramamoorthy: Right. I think the constraints in aviation where you have to go by the rules at the same time you have to differentiate. And today, if you think about Airbus and Boeing, you can say the same about ChatGPT and another comparable LLM vendor. There is not much of a differentiation. Is there a clear differentiation when I use one LLM against the other? No. I mean, they’re all borderline same. Something have their strengths, but others have their strengths as well. So in an era where you cannot differentiate much and then you stick back to your convictions, safety is a conviction. Better service is a conviction. Value is a conviction in aviation. Same thing. Privacy is a conviction. Playing the long game where very stable we are here, we operate in emerging markets, we price our products in local currency there, and that’s a conviction. That’s a long-term conviction. Unfortunately, SaaS has been chasing aggressive growth, chasing quarter-on-quarter results. No, it’s okay. I mean, play the long game. So what does it took when we invested in data centers? Would that have resulted a better share price? Would that have got us more funding? Absolutely not. We were able to do that because we were privately held and we had a long-term deeper conviction. Same thing with building our own foundational LLM. Does that make a business impact today in our balance sheet? No, but in the long term, when LLM becomes a crucial piece of technology and when we have the deep technical know-how to do it, that is a differentiation. So I think this long game, having a good track record… I told you about the top three airlines, right? I trust them for doing things. So we want our customers to trust Zoho for privacy, to trust Zoho for value, to trust Zoho that we will not ditch them when technology changes and we will have it as a natural evolution of things. So these are the parallels that I draw and the way we see business at Zoho.

Robert Kramer: Well, this is fantastic. Thank you for joining me today. Robert Kramer with Moor Insights & Strategy, Ram with Zoho, Head of AI. Actually, Director of AI Research. And it’s been a pleasure and we’ll look forward to connecting hopefully in the future. Thank you very much.

Ramprakash Ramamoorthy: Pleasure is mine, Robert. Thank you.

Robert Kramer: Cheers.