Video Interview for Zoho by Jason English
At ZohoDay 2026 with Tejas Gadhia, head evangelist for AI and Developer Platforms at Zoho. Intellyx analyst Jason English (“JE”) interviews Tejas about the intersection of their low-code Creator suite for Citizen Developer automation and their Zia #AI capabilities, small and specialized language models instead of LLMs, learning future development and innovation-proof job skils, Ferraris and popcorn, and more.
Watch the interview video below or visit our YouTube page at: https://youtu.be/VpSVv1yw2J0
Full transcript of the interview:
JE: Hi, I’m Jason English, a principal analyst here at Intellyx, and I’m here at ZohoDay 26, where we’re exploring all the different solutions that Zoho has for the enterprise audience as well as end users and partners. And it’s been a very interesting conference so far, but I’m here with Tejas Gadhia. He’s the AI and development evangelist for Zoho. So Tejas, what is your new role? Until I know, like I know, executives at Zoho kind of take a round-robin approach or a zone defense approach where you’re constantly shifting into what area you cover. So why have you moved to this new role and what does it entail?
Tejas: Good question of what it entails. I think things are changing so fast, it’s hard to tell what’s actually going to be permanent. So every week or month, it seems like a different kind of a thing. I think the main thing is really just around, like embracing what’s new, finding new solutions and figuring out ways to improve not just internal processes, but also for customers, giving them, I think, a little bit more what I believe to be like poor, practical, probably like a grounded, simple maybe strategies, best practices, future considerations around like a lot of tools that they’re adopting nowadays.
I kind of feel like we’re in an age of it’s almost like back in, let’s say, 2012, 2013 when like Uber first came out, where delivery apps first came out, you know, everyone was addicted. You’d get Ubers for like 4 or $5 and you felt great. And now, like an Uber is more expensive than a taxi. And like, inevitably, a lot of these tools right now, they’re just getting the users a lot of subsidized things.
So it’s great to experiment and take advantage of essentially subsidized software or technology of any kind. But at the same time, most people, especially our customers, are very like value oriented. So they want to make sure that even if you adopt these tools, if you become, I don’t want to say addicted, but kind of close to it and you become dependent on it, then are you protecting your own financial side of things, your risk side of things, your privacy side of things, your compliance side of things?
And right now I think we’re kind of seeing explosion of experiment. But what is the next phase of when the the pricing model start changing? When the efficiency starts changing, when security starts becoming an issue bigger and bigger, privacy concerns start popping up. Those are like problems that will pop up next year, in the next five years that I think are kind of on the back burner because everyone’s like, try and do everything you can, automate whatever connected all and do it all.
But then someone with the with the same mind will probably step in and be like, all right, we got to figure out a long term strategy here.
Yeah, yeah. I noticed that you were one of the first companies to even suggest the idea of specialized language models or small language models as an alternative to LLMs, which everybody is trying to race toward the best, strongest, largest LLM model.
And it’s just not going to be sustainable in the long run. So how did you come to that conclusion? You could orient the company toward these more specialized models, and how do you sort of evangelize that practice within your customer base?
Yeah, I think a lot of evangelizing within the customer base is kind of almost like invisible AI, which is what people really want.
It’s not so much about talking to something or triggering a chat or having to like, think about using it. It’s the stuff that happens magically in the back end that may or may not be appreciated, but is taking advantage of on a daily basis. Even if it’s crazy things like spam detection or performance monitoring or those aren’t like the coolest things that people feel, but it’s things that they take advantage of and essentially don’t appreciate on day to day basis, but are all powered by a lot of intelligence models have been doing for a very long time.
And I think the issue around the large models is, I think the analogy is generally like – I don’t know if it’s relatable or not, but it’s kind of like you don’t have to take a Ferrari to go check your mail. Yeah, I don’t have a Ferrari, but I mean, I assume if I did, I probably wouldn’t go to the mailbox.
But before if I had a Ferrari, that’s a different story. It’s kind of like you got to use the right tool for the right job. And right now it’s like, maybe the better example is like, you have the world’s smartest human trained on the world’s data, and you’re going to ask them, how long do I pop my popcorn for?
That’s like, not what you should be using that for you. That’s like expensive to hire a consultant that knows everything about everything. And you’re going to ask them a very simple question. You got to find the right popcorn export or somebody who’s trained in that domain who has that knowledge. And what’s happening right now in the market is with these generalized models.
They have a lot of broad knowledge. But then what is everyone doing today? They’re building skills and agents. And those are just really prompt wrappers that establish a lot of context. And that context is all manually controlled because you have to define that context because no vendor is giving you that context being exposed. Wrangle is more that context should be ambient across the board.
When you’re using multiple Zoho products, your CRM, your help desk, your email, HR, finance, whatever it is, all those things, context is intelligently there. Automatically. Your context layer is not manually defined now it’s ambient. Context can be like the basic stuff like records, maybe like customer information fields and data. That’s all table stakes I would say through like API calls and programmatic access.
But the real context comes from like I would say, like organizational culture, organizational processes that are not really documented that, well, like let’s say we’re planning this interview, maybe we’re going back and forth on chat, a couple of emails here and there, but it’s not like the most structured way each situation will be very different. So how do you figure out all the activities that somebody is doing?
How do they interact with each other? Are they talking like casual or not casual? So when I say like if me and someone at the office in Austin is talking, it might be way more casual talking to somebody from a different office that’s, you know, doesn’t make different that I don’t talk to you that much. Maybe the context is a little bit different.
So what I want to define that every time. Hey, I don’t know this person that well they’re from a different office in a different region. So where did something this way or present something this way? No, that should be all ambient, automatic. I don’t want to sit there and write rules for my organization, because I’m never going to be able to hardcode every aspect that is almost undefinable, which is kind of the entire value of the unified context layer that we could have been talking about.
JE: Yeah. Another interesting aspect is since you have a history of we’re here at the 30th anniversary of Zoho, which is kind of wild to think about where you started out, and you have this huge suite of automation tools that have been very productive for customers, as well as low code tools and and other abilities to build applications. You know, on top of this substrate of the data that you have.
But where should people start thinking about drawing a line between where should you use AI and does good old automation still have a big place in this discussion? And how do you how do you help people make that distinction?
Tejas: Well, that’s a good question actually. I would say like the age of automation is not over. It’s not like workflows from apps and stuff are not going to be going away.
You still need basic automations. You don’t need to add for that stuff. Maybe you want to add like an AI layer for some added logic on top, but to actually execute this stuff, you don’t need to again, burn like an expensive model to be able to execute an email send. It’s, you know, do some logic, figure out how to frame it.
And then the actual execution layer happens through like a normal workflow engine. So I don’t think I don’t think it’ll cannibalize them. I think if anything, our general sense, especially in the developer ecosystem or even in the world, everyone thinks everyone’s going to go to everything. That’s kind of an extension of workflows as a whole, and it’s kind of like, well, it’s probably going to happen.
It’s actually going to happen is you’re just going to broaden the pie of people who feel like they can now build. It was kind of like low code back in the day, gave like citizen developers a taste of being able to create something that was seemingly meaningful. And once you start low coding and you’re dragging and dropping, maybe writing a few small scripts, you go deeper into the weeds and you start learning about a little HTML, a little bit of JavaScript.
It’s not like those languages went away and we abstract everything away. We basically just made it easier to interact with. So today you might be in a chat interface or even in like a terminal, but some of these coding tools, we’ll have abstraction layers above that that make it easier. So it’ll be drag and drop and visual and it’ll make it more approachable.
So the pie of builders and I would say just creators, builders, whatever you call it will only grow. And actually it’ll be a lot more, I would say like the creative types rather than the process types, because those are the ones with the ideas who can articulate things and have like really like a good, yeah, creative, out of the box thinking kind of a mind.
But if you’re very like process oriented and just following like a script every time. Those are the ones that could be a little bit more automated away. You’re not going to find a lot of new ideas. It’s kind of like the troublemakers, the creatives, the ones who are like, man, it would be cool if this happened. That’s the one who will.
Whether you know how to do it today or not, you know that it’s possible. And with that creative idea, you go down a rabbit hole and you start building crazy cool things that started as maybe pet projects and then start making big impacts on your organization.
JE: Yeah, I’m kind of curious about how sophisticated you’ve seen development get in your customer base, because when you have a lot of companies that really did start out with just one person, probably just building process oriented tools and some low code apps, and now they might have a 10 or 100 or a thousand people soon.
And so you have these bigger companies, and so do you see a broader range of skill sets that not just those citizen developers, but some actual developers interfacing with your team? And how is that going?
Tejas: Yeah, I say the number one skill set probably in life is being able to think and plan properly. Like that’s number one issue.
So even with low code world, you know, we made it easy to build tools and a lot of people do the same thing doesn’t mean people build good tools, doesn’t mean they have good data models, doesn’t mean they thought through all the processes and the nuances and the quirks and exceptions and whatever things that down the line right now, it’s almost even worse because leasing a local platform or any software platform, you have some guardrails, some bounds.
You can’t go too crazy nowadays if you’re building completely, you know, just thrown out there in one shot in some kind of bad coded app that things got problems you don’t even know about and you don’t know how to fix them cause you don’t even know what you don’t know, at least like the the SaaS products out there, that the guardrails keep it maintained and make sure you have structure, provide advice so those, you know, guardrails and tooling and like best practices will get built into more of these agent kind of building opportunities.
But for now it does require some level of a of a caution or probably just responsibility. Probably a better way to word it.
JE: Yeah. I mean, I think just like in the opening session we were talking, there was a lot of talk about and this has been going on in the marketplace right now, talk about how AI is going to replace SaaS or something like that, when in reality, it really seems like just having this foundation of SaaS becomes like a better tool for AI’s to use in building workflow.
And as far as having one source of truth for what they’re doing, as opposed to grabbing random things out there that are unauthorized. And so, you know, what’s kind of your perspective on that when people, I guess they could be critics or naysayers who talk about, SaaS going away because of AI, I just don’t – I’m not seeing it. But what’s your perspective?
Tejas: I mean, I think there’s probably like a wide spectrum trying to make that kind of a blanket statement on one end for like the actual like I would say, the vibe code hype. I do think there is a threat for SaaS in a sense that you will have an explosion of just products in the market for people to choose from.
And right now, let’s say you’re like a mom and pop candy store with a retail front, right? Your choice for managing your customers might be a full blown CRM. So it’s your own other competitor, whatever it is, right? But that’s like not made for your flower shop or your candy shop or whatever you have. Right? So you go down the vertical path, which we’ve seen in the market for quite some time.
Will that vertical solution do exactly what you want to do? Probably not, because it’s usually customized for that vertical, but maybe not for your specific use cases. And so then you’ll have an explosion of these super, super niche products that have a very probably a small addressable market, but are now being able to be served extremely well. You have a lot of people take advantage of those people.
So I think there’ll be some negative consequences as well. But you would have a specialized CRM for maybe like candy shop owners in Texas. Maybe the total addressable market of candy shop owners is like 1000. But if I can get 500 customers on board, that’s a pretty successful business for like one person shop. And so you’ll like everyone thinks about this big addressable market and billion dollar thing.
You can look very good by having a very niche focus that you didn’t think was a sustainable, I would say like a market or even like a business idea, and they’ll be served extremely well. They’ll be taking care of someone with deep knowledge of candy making, I guess, or whatever that bad example I used was, you’ll get like a lot of expertise there.
I think you’ll get a lot of to take advantage of that. So you’ll have like to frame it back. You’ll have like these super niche specialized ones that will provide a lot of value for very small customers because it’ll be very niche, very simple and what they need. Then you take a step up and it’s like, all right, we’re going to make a new helpdesk on AI.
We’re going to make a new CRM or new HR or whatever it is that’s a little bit more risky because now it becomes more of a mission critical core system information. Am I really trusting that? I think what people that one thing that people kind of discredit and you kind of mentioned earlier was the 30 years mark, is there’s a trust involved when you when you choose any kind of thing you buy, it doesn’t matter if it’s closed, the car, the software, whatever it is you’re buying, not just today’s product.
But does this company have a history of innovation, a history of serving their customers well, the history of doing the right thing, a history of delivering good value, bringing prices down, offering more features, whatever it is, you know, everyone has a trend of kind of their perception and that trend or like a that a character of the organization or person builder, whatever you’re buying from that’s established over a lot of time.
Like you can’t fake that, you can’t do it. So when you have a lot up and comers, how do you build credibility? How do you know that this person is going to support this application, that they vibe coded five years down the line? How do you know it’s going to be interoperable? Do you know it’s gonna last. And how do you know it’s going to be accessible?
Probably not. So you have to factor all those risks in for that. Like the SaaS killer. Kind of like a fear mongering, I guess. And then upmarket, I would say. And for more like mature companies, even just mid to large organizations, we’ll have probably a decent amount of like I would say like top down automations and integration central IT teams.
And then we’ll probably be trying to deliver like productivity agents and stuff. I think that’ll be the first step. But really where this all kind of goes is truly hyper personalized kind of experiences. How I work, how you work is very different. And so why should even though we both use the same software and the same CRM, let’s use that as an example.
What I need to see is very different than what you need to see. And why should I see the same thing as you? So we’ll have like maybe one core system of record and maybe 5000 apps for 5000 employees that all do exactly what that employee wants. Maybe there’s some mandates from central or whatever. But if I need to just see last five emails, a few customers I got to follow up with, and a couple of tasks, why should I not just have my own thing from that?
I have access to the data, obviously. So I mean, I can now develop an interface that makes me more productive for me. So you’ll have a lot of personal productivity, I think. And you’re not going to be bound by what the vendor offers. Hopefully that vendors extensible enough. And and I think that’s when people feel like I’ve got something for me, which kind of goes back to that niche example of a the candy store or whatever.
It’s made for me, and I’m making it for me. I know how I work, I refine my process, just like you have like folders, an email that’s made for you. It’s not mandated from the organization level of like how I organize my email. I trust it on you. Same way if I want to have automations in my email, that’s not me.
It’s not the organization whose pushing that down my throat. However, you leverage folders and automations in email, that’ll be your own personal productivity gains, and that’s on you to kind of find the best way that works for you. So yeah, long story short, hyper personalization for a lot of people, a lot of value for like very micro businesses, I think from well-intentioned developers and a lot of risk for up and comers that may not have the longevity and the trust kind of built up.
And not to say that they’re not trustworthy, just that it takes time to build up that trust. And so you don’t want to put all your eggs in one basket without knowing what the long term future is.
JE: Yeah, yeah. It’s almost like because you have so much detailed expertise for each at these very low levels you have, that becomes the context for the AI itself to act upon.
And so like we talk about AI all the time with other companies in a, in a general sense, like, okay, instead of having an API gateway to talk to, we’re going to put MCP server up there now and that’s going to replace that. And then it’s just going to talk to that. And then it’ll magically sort of know what it’s supposed to do next or something like that.
And so I mean, your approach is very different because you are the system of record and you have so much rich data for each particular user, even down to the customer level, how do you recommend developers approach this if they’re really trying to get into the space, prove some value quick and do something that’s meaningful for their companies?
You know, what do you recommend when you when you talk to them?
Tejas: Yeah, I usually tell people initially you kind of want to do something that feels like it has like big org impact. You want to build this agent that’s going to do tickets for you or this like SDR thing that’s going to follow up with customers. And you think it’s like some major thing.
Instead, scale it back. All these like super ambitious projects. I think I saw something, I think it was like MIT study, if I’m not mistaken, something like 95% of AI projects have failed in the last year. And it’s crazy. And so the reason is because usually the larger the company, the more ambitious in scope. And it’s like we’re going to automate a bunch of stuff, start super small, find something like very basic, maybe like automatically routing tickets.
It’s not. It goes back to like it’s not the coolest thing, but it probably saves a lot of headache or maybe like some sort of summarization thing that’s like truly contextual to you and not like just taking what the vendor gave you and slapping it on there. As a vendor ourselves, we have to have our we have to offer out of the box AI functionalities.
So let’s say we have something that like summarize notes for example, and various products. Now summarizing notes is like essentially what we’re doing is we’re taking the content of the note, wrapping it in a bit of a prompt, and then sending off to the LM, do the math, come back and give me the the summary. But what I need to see in that note summary, what you need to see.
What the sales person with the engineer, what the PM, what the CEO. They all need to see different content from that summary. So you actually need different prompts for each person based on their universal context. What does this person do when they look at the record? Am I looking at really like, oh, customer didn’t sign a Po? I don’t care if they sign a poll that’s not relevant to my summary whatsoever.
I want to know, where they’re from, what they do, what issues they had various aspects. That’s my personal summary. So like all of these out-of-the-box features are really going to be table stakes. It’s a nice way to get your foot wet, but you’re not really going to increase productivity if you use the same generalized tool everybody has.
So that the box of features that any software vendor offers, ourselves included, they’re there to get your feet wet and your creativity going. But to really take it to the next level is really about custom extensibility of any kind. You got to wrap it in like your own context, your own explanations, your own company, like, I guess, philosophies or whatever it is, your and your role goes back to that hyper personalization of everything will be very much not just generic AI feature gives me a summary.
There’s an automatic routing. It has to be, well, this is how we work and we need to add our layer of context on there. And you can manually define it right now. Or it can be ambient defined for you based on your historical work as well.
JE: Yeah, I guess for you, because I’m always really interested in how Zoho does development work itself.
And so inside your own organization, you have thousands of probably more than was it 5000 developers? Is it more like 50 now?
Tejas: I don’t even know at least at least 10 thousand. It’s got to be like 12, 13.
JE: Yeah, yeah, yeah. There’s thousands of developers in the company. And so ideas are filtering in from all over the place.
And so how do you sort of wrangle so much innovation into, into a strategy that you can execute on like a one, one step at a time? Because a lot of companies who are especially as you move into larger markets, they’re dealing with this issue themselves. And now they have to deal with the issue of AI and a lot of people who are eager to jump in, do something with it.
So so how does your approach to that differ from or is it unique to Zoho itself? I’m just curious.
Tejas: I don’t think it might not be unique to us. I would say like our approach is generally like one, like most of our product teams are very decentralized, and decision making happens at the individual level or centralized stuff.
But we try to give experimentation options to everybody to do what they want. So different teams will embrace or maybe not embrace or whatever it is, use it for different things. Some will focus on maybe prototyping, some might focus on testing, I might focus on customer service and customer facing opportunities, whatever it is. But so we don’t really have like a top down mandate.
I feel like on one hand it’s good to have a centralized like a strategy, but sometimes that central strategy will kind of box you in and prevent the creativity that you might want. So for us, it’s kind of free for all right now. I mean, I think that’s kind of general sense for most people. Free for all in the sense of don’t connect things you’re not supposed to get to, like make sure it’s secure through like our own governance structure.
But you have to experiment and find something that works for you and your team. And it goes back to that context thing again. Like if central IT comes and drop something down that’s going to be uniformly mandated. Chances are not going to work for like whatever subgroup or some kind of experiences you generally have within a developer, a product team, or a sub team within their module team or a testing team or whatever it is.
And so it should happen at the closer to the problem rather than like a big issue. But sharing the information is the most important thing. So like when one team learns that, hey, actually like this tweak helps us, helps our productivity, you got to share that stuff and build like, I think like an unofficial or official, I guess, center of excellence, kind of a term.
I don’t like that term. It feels very cooperative. But, you know, just a group of people that share information, learn from each other, get better because otherwise there’ll be a lot of repetition. If I’m making an agent that manages my email and you go make it from scratch, you are going to hit a lot of roadblocks and maybe you’ll learn along the way.
Hopefully everybody will learn along the way. But you also shouldn’t reinvent the wheel if something’s already been built. So you need to have like internal repositories and shared information, and that’ll increase the productivity of everybody to just learn and build more. And now days these tools are like, you know, you can get the ASMR, you can have like videos generated that will make it interesting.
And someone makes an agent like, I don’t want to read the code, I want someone just make me a short overview video and I’ll probably be more interested in at that point. So I think.
Maybe I won’t say the strategy is no strategy, but the strategy is generally decentralization. So everybody can experiment without major roadblocks. But you still want an overarching theme of like leverage the tools that are available to you, and it’ll help increase your productivity for your own sanity. Same way, like people used to write on paper. Then they got a computer.
And when you get the computer, you can’t hike the expectations go up because you now have a tool at your hand that you can write faster than by hand. So, you know, and if you don’t use a computer and you’re writing by hand, you can’t say, oh, I’m writing by hand. So like my output is going to be lower.
It’s like no technology is given to you. It’s there. It’s for you to embrace and then make yourself better and get ahead of the pack so you can fight it. But it’s kind of like very obvious where things are going. And it’s not necessarily about being an early adopter, it’s more about anything that can help productivity. We should always be curious about whatever kind of it is, whether it’s planning, whether it’s documentation, whether it’s prototyping, whether it’s meetings, whatever it is.
Those are all, you know, if a new something comes in that helps us be more productive, we should always evaluate if something that works for us and and try to embrace it or adapt it to kind of how we do things. I don’t know if that’s really the answer to your question, but a long ramble of various strategies, and that’s kind of the good thing about us.
We can kind of do whatever we want to in different areas and see what works.
JE: Yeah, yeah, it’s definitely a different culture. But then again, you know, the idea of just knowledge sharing and having a knowledge base that advances through this process. It’s just it’s that’s a very common theme. And one of the things that really works well with AI, you know, it’s really good documentation and knowledge sharing.
Yeah. And you know, promoting if you set it up right, it can help promote where you’re trying to go as opposed to just what you’ve done so far. Yeah. Well, I guess I think probably got as much as I can get. We could go on forever, but this has been very productive talk, and I really enjoy hearing about what you’re seeing in your line of work and how much Zoho has really embraced AI and turned it into central part of the strategy without losing their roots and real productivity and value.
Tejas: Yeah, I think it’s very easy to, like, get caught up in the hype to. And sometimes you got to just pause, see where things are kind of going, see where the I guess the puck is going or whatever the analogy is, and kind of be a little bit more patient. And I think that’s a kind of like a really good thing that like she then everybody kind of like instills everybody from the beginning that like, you kind of got to be paused and sometimes you get a little FOMO or you feel like you’re behind or whatever it is, but, you know, you kind of take a breather, see where things are going, don’t just rush in and try to jump. I think I’ll say, like last year, year before, I want to say Klarna. The by now, pay later app. Yeah, they had made some posts about like I would call him out, I guess, but whatever it was public, they were like getting rid of all their support team and they’re going to have bots and agents running everything.
So like laid off a bunch of people and then they’re like, oh, this is not working. Then they’d hire people back and like, this is not a that’s like a, you know, you jump too far and then you didn’t like, think strategically of not really replacing people, but it should be like the superpower, not the replacement for people.
That’s like the wrong way of looking about it. You want to empower the employees to make sure you’re not really replacing them.
JE: That’s definitely a big thing, I think. Well, yeah, they’re not a customer anyway. Yeah, just doing payday loans for online purchases is not what I call it. Yeah.
Tejas: Anyway, also next year, maybe it’s just like a little robot here with my voice and my context, and it’ll just do the interview for me.
So exactly will be personified in our in our AI avatars. But actually, yeah, the robot things. Me last little thing I’ll say, I think there’s a big thing of like like let’s say if I take like helpdesk automation or sales automation, right? So you have this like agent in LLM basically drafting these outreaches and emails. Right. And you think you’re like getting one over somebody.
And there’s nothing that granted my gear is more than when someone sends me like an email that is very obvious copy pasted from another one, because it’s like, you don’t respect me enough to read my stuff and like, you lose that human touch because what am I going to do? I’m going to build an alum that’s going to answer my emails for me.
So now I’m talking to this little, no one, no human in the loop. And you have to have that human touch for everything. I think the race for automation, and like the AI thing, it’s still grounded, like human connection, human relationships. That’s how sales is done. That’s how good customer support is done.
That’s how product teams learn from customers. You can’t just have like an agent go into an email and that agent comes back like, that’s not going to work out very well for you. So some people will go far down that route. But I think for history, it’s basically like, you know, the human connection is what perseveres.
And that’s going to be the X factor going forward. So when you have option between human support or human sales or human experiences versus like a robot, a lot of people will choose the human and it might even be a premium in the future, but that’s going to be what people still want. No matter how much you get, nothing beats the humans.
JE: Yeah, hopefully. Let’s just try to make sure that happens. All right. Well, thanks. This has been super interesting. I’m Jason English here reporting from Zoho Day for Intellyx. Thanks so much.
