Brainwave: How will AI and Low-Code work together? Tejas Gadhia, Zoho Creator

Problem Solvers — Podcast / vCast for December 19, 2020:

Full transcript of the podcast:

Charles Araujo: okay. If 2020 has taught us anything, it’s that there’s never a shortage of challenges that the enterprise must be prepared to face and overcome. There are always problems to solve. And there are two classes of technologies that have gained increased prominence in the last few years, for their ability to help enterprise leaders solve some of their most pressing, pressing problems, AI and low code.

My name is Charles Araujo. I’m a Principal Analyst with Intellyx and today’s host of the brainwave problem-solvers podcast. The question that we’ll explore today is what happens when you put these two powerful and emerging technologies together, and how can their combination help enterprise leaders solve problems in all new ways?

I’m joined today by Tejas Gadhia, product manager for Zoho Creator, who is going to help us dive into this fascinating intersection. Tejas. Welcome to the podcast.

Tejas Gadhia: Thanks for having me.

Charles Araujo: All right. Well, I think the place we need to start is with AI, right? It’s hard to open your email or, you know, talk to anybody almost without something about AI.

Just sort of hitting you in the face today to say that there’s a lot of hype around it is a massive understatement. So can you start maybe by helping us separate some of that hype from the practical value than an enterprise leaders can reasonably expect to get from AI? Yeah, that’s a really good story.

Tejas Gadhia: Or, I mean, I would say the AI term in general. It is a fairly all encompassing one, a lot of times, AI ML and machine learning, deep learning neural networks. These are all terms that are often thrown around interchangeably, especially in marketing context. So identifying, I guess, what does, what is, is pretty important?

Generally speaking, the way I look to look at it is, AI is kind of like the overall concept. And then inside a guy is machine learning. Machine learning is more of a technique, that achieves a level of AI, but at a basic level, AI is just. Basically having a computer or a machine or a software, do something that we previously thought required some human to basically do.

and we’re either automating it. We’re making it a process that, the machine essentially can do, make some intelligent decisions ultimately, and just be smart about, the actions that we thought only humans were able to kind of. Foreign, when you call, make a decision upon. So that’s at a very basic level, I guess, just understanding the baseline of what AI is and kind of how we’re categorizing it.

but I mean, in terms of what value you really get, ultimately, I think you can kind of break it down into two main categories. The first one is productivity. Ultimately, any kind of AI you’re trying to leverage is we want to make something more productive. whether. at a very basic level, it could be, you know, building something quicker than you thought you normally would be able to do it.

or just making sure that you’re more efficient with your time, essentially because the computer is handling some of that decision-making, Workload on your behalf. And the second half of, AI, I guess a practical value is really along the lines of more data and insights kind of understanding. and that might go in a little bit more in the machine learning category, but it’s more taking a look at maybe historical trends of data that you have, and then giving you some inferences and ideas about what decisions you can make for the future based on that.

So that’s kind of the two main categories. I’d put it in productivity and then, insights.

Charles Araujo: Yeah. And I think it, I think it’s fascinating because the, the, the secret of good hype is having those kernels of truth. Right. And I think a lot of the stuff we see is talking about elements of that, and there’s, there is so much truth in it that I think the trick is, is where people sell it.

Like, it’s this out of the box. You can just turn it on and off it goes. And that’s, that’s sometimes where we see the challenges. I mean, is that sort of, when you see the difference between hype and value, is that some of what you’re saying.

Tejas Gadhia: yeah. Yes and no. I think sometimes there’s, there can be AI implementations that are happening behind the scenes that you never even see or witness that kind of just help you guide you along a process, even better.

And it’s not something that, you know, normally you would ever consider. So maybe you take something like, you sign up for some web service and, you know, you get the introduction and the onboarding kind of screens and you’re going through it. And, you know, those screens you think might be, you know, kind of static and everybody kind of sees the same screens when they sign up.

but there might be some kind of algorithm happening in the background that based on in my location, maybe based on my demographics or whatever information they might’ve gathered about me before. Or during that signup period, they’re going to dynamically change those screens up and have that kind of information happen.

And they basically know that, Oh, this guy’s from Texas. and you know, he makes X amount and works in this industry. Maybe we’ll show them some examples or use cases that he relates to a little bit more. And so, you know, he might sign up for our software or pay us money basically. So that’s like some behind the scenes things that.

As a user, I would never see, but I do get it and see, I see value out of that because I’m seeing a more personalized approach to an onboarding session that might’ve been a little bit more generalized. but then there’s the. More hands-on approach to AI, where I have control over something. I’m going to say, look, this is my process.

This is my data. And I need to be able to infer information or I’m trying to build something out. And is there a process where maybe a certain stage is getting stuck and it’s taking seven days to move to the next stage versus three, or maybe a certain sales rep or a employee? It’s taking, you know, 12 days to verify a record versus four days and identifying those kind of anomalies and things like that can all be helpful as well.

So that’s like me having control over things and seeing it and being able to. Adjust based on that information. And then sometimes just being on the receiving end of it, can be beneficial as well, too well.

Charles Araujo: And I, and I think that one of the challenges with AI is that it is one of those that, that at least from my standpoint, one of the things that sort of separates the hype from the reality is.

It is exactly how challenging it can be to build the models and to put them in deployment and all of that. And that actually leads us to what I think we really want to dive into today. And then what happens when you combine AI and low code to maybe remove some of that complexity. And so while we’re in the mode of defining things, why don’t you maybe kind of give us your definition of, of what a low code platform is for people who maybe aren’t as familiar with the term?

We’re all kind of starting from the same place as we examine this.

Tejas Gadhia:  Yeah. I mean, I think it’s pretty fitting that the topics are AI and low-code because both are, I think are pretty, Pretty modally defined terms in terms of that’s the popular word that everybody’s associated with it, but they mean a lot of different things and encompass a lot of different stuff.

at a high level, any kind of Low-Code tool is basically similar to it. Yeah. Something that increases your productivity and enables you to do something faster, quicker, more efficiently in an organized manner. Batching, normally wouldn’t be able to be, be able to do, unless you were kind of. Maybe using it or building that with more Raul or scratch tools.

So instead of coding an application completely from scratch, creating a database, setting up a server, provisioning your security and updates and all that kind of stuff, that would be manual work that would happen. Low-code is like, look, we’ll take care of all this backend baggage for you. and all you gotta worry about is generally business logic is kind of where low-code wants to be at.

So defining processes, defining how data is kind of organized. Creating workflows things along those lines. the ultimately enable productivity, but also empowers a lot more people to build things. So before building applications and solving problems was something that was only in the hands of maybe an IT team or, you know, a techie type person.

But now. Somebody who is not very, you know, technically forward, but understands the basics of maybe an Excel spreadsheet or something like that. You know, they see how data is kind of organized and columns and rows. And, you know, you click on something and something else happened. Now I can build something that, you know, it makes my own life easier.

Every single day. It can make my team’s life easier every single day, departmental share information, that kind of stuff. It’s just scalable way of solving problems. Long story short. Right, because we don’t have to be worrying about knowing the syntax of, of having to code. You’re able to just focus on logic.

Charles Araujo: So, yeah. Okay. Well, so I think that helps. so I want to step back and things you said earlier that, that AI is sort of this broad term is, you know, I, I believe that as well, or, and that there’s all these sorts of different types. So when we’re thinking about this specifically, in terms of the intersection of AI and low code, w what are the types of AI that we should be thinking about, or kind of aware of.

Tejas Gadhia: I’m sorry. I think it goes back to kind of what I hinted on a little bit earlier. There’s generally, we kind of categorize it in two different buckets. One is platform level AI. So that kind of goes maybe to the onboarding example or, you know, helping someone build up a more efficient process.

And so an example of that would be, let’s say you’re building up maybe a new employee onboarding process.

And we’ve got an understanding of that through our HR applications. We’ve got an understanding on that based on the millions of other applications that have been built on the local platform.

And so when you go and build it, you might create it a certain way and we’ll give you suggestions saying, Hey, you know, maybe you forgot to capture the candidates birthday. Maybe you forgot to capture, this, like, what do you call a benefits package information? And we’ll give you those suggestions along the way that you normally wouldn’t.

No about, so that’s kinda more along the guided, platform level stuff. And then there’s the data oriented things, which I think people find a little bit more tangible or, or hands-on, and that’s things like image, classification, forecasting, anomaly detection, image classification. what you call text-to-speech or an image-to-text OCR recognition, things like that, you kind of can see with your own eyes that I put in this information and now the computer or the machine or the software is basically spitting out some level of insights for me that I don’t have to do myself.

Charles Araujo: So let me try to process and translate this, make sure I get it because actually, I’m someone that has done quite a bit of study around and I don’t think I’ve heard it categorized in these ways. I’m actually quite intrigued by this. So you’re saying the data AI is sort of the, probably more traditional use cases or the way we think about it, of looking for patterns in the data.

To identify an insight that I might leverage in some ways. Is that what that is?

Tejas Gadhia: Yeah, I think that’s what most people probably equate it to a distance the most easily relatable and identifiable cool use case. Okay. And then platform AI, you’re saying, I actually don’t know that I fully was following that.

Charles Araujo: So, so explain platform AI, meaning it’s like embedded into a platform. Is that what you’re saying?

Tejas Gadhia: Yeah, I would say it’s more behind the scenes, things that are happening. To make sure that, what you’re building is good. So in the, in the low code space, generally the biggest problem that we end up seeing is that the people there’s a pros and cons of democratizing this application development.

The pro is more people can develop apps and people who aren’t very technical can build apps. And the harsh reality is, is that people who don’t know how to build apps very well, they tend to build pretty bad apps. They are just not able to design things very well. And it’s not a knock on them.

It’s not their primary skillset and their primary job function. So it’s not a focus that they do. So at the platform level, we have to basically help that person out to build applications better. so that’s suggesting relationships suggesting workflows, identifying issues, preventing duplication of data, cleansing the data that comes in organizing all that backend stuff that you might need them to take for granted and not realize that it’s happening, and it magically makes your application building experience better and ultimately yields better results.

When you go in. Run reports and figure out what this app is doing and playing it a little bit to myself here, because as my kids will tell you, sometimes I have to hear the same words twice to understand them. So could you basically said the same thing and it’s like, it just finally clicked.

Charles Araujo: So what you’re really saying is, so platform AI is you’re talking about this AI that for instance, you guys have built it into your platform to help me as a user.

Build an app more easily. So from, from a local perspective, it’s sort of I’m the, as the, as the app builder and low-code, I’m the consumer of this AI versus the data. AI is the distinction that you’re drawing here to say that that’s AI, that I’m applying in my app for my ultimate consumer of that app. Is that okay?

Tejas Gadhia: Yeah. Two sides of the same coin that ultimately make you. Got it. All right. It makes perfect sense. You just had to tell me twice so I could get it through my thick head. Let’s dive a little bit deeper into the use cases then here. I think that the platform AI I get, are there any, when we think in terms of low code, I mean, obviously that would seem like.

Charles Araujo: That’s when people should be looking for platforms that are embedding AI into it, to help solve some of those, the problems of people developing apps that you talked about. Are there any special gotchas or special parts of the, this from a use case perspective that I should be looking at?

Tejas Gadhia: Yeah, I would say the primary one that we see people generally. get the most benefit out of is really data modeling. And so just the idea of you go and create a low-code application, and most of them are data oriented or database oriented, where you’re tracking and a bunch of information and organizing it in some way and probably wanting to report on it in some way.

And it could be pulling it from a couple of different applications or data sources. And so what ends up happening is that, the reason you’re pulling put, organizing this information. Is not just to have it, you know, neatly stacked in a certain place. It’s to really be able to run reports, analyze it, see charts and graphs and sorting and organization of that data.

and the way that that can happen, the way you can get better insights with those better reports is if data is related and organized properly, And so, even at a very basic level, let’s say we take what most people can probably relate to as an Excel sheet and you have a bunch of numbers in a column, right?

So Xcel, if you type in, five in a cell and press enter that number goes to the right hand side, right. And Xcel is basically identified that this is a number, not a text or something. And now on that five cell, I can do some mathematical functions that I wouldn’t be able to do. If I type, you know, the word Charlie in itself, right?

The same types of functions aren’t happening. I wouldn’t call that necessarily AI that’s a little bit more just, you know, classification of objects, but that gets extrapolated into a bigger sentence. And so, being able to run reports and the analytics and insights based on information in a lot of different places.

That information needs to be connected generally through like unique identifying keys. And it could be a whole manual process. There’s whole database administrators whose full-time job it is to basically maintain these relationships and structures. And the local platform will basically take care of all this for you.

So. Kind of ranting here, but I guess a very practical example would be, let’s say you have a CRM application and this low-code application, maybe your low-code application of some commission calculating based on deals that are closing in your CRM. So your CRM sends your information to this low-code app and based on maybe, you know, a deal ID or the deal name or the amount it’ll connect it.

And we can have that commission linked to that CRM. The Lincolns are extremely important and people who make bad links make duplicate links like incorrect links.  reporting you can do because the reporting is only based on connected data. We can’t connect data, that’s not connected or report on it. And so identifying good data models is.

The core foundation of a good application built most new applications, even if you’re coding from scratch, starting at the database layer and organizing that data structure, is the first step. Any professional developer will tell you that they need to tackle before they get into logic and design and all the other things that are just kind of layers on top of it.

Charles Araujo: Yeah. You know, it’s funny I’ve seen some of that firsthand and, and I think it, it addresses there’s two problems really for a citizen developer or a business user developer, who’s trying to build low-code app. And the first is, is maybe not understanding things like data models, right. Just having that, you know, lack of, of, of understanding of sort of the science behind that, of why that’s all important and that you have to choose them carefully, as you’re saying.

So I think that’s great for that. The other is just ignorance. Like, I don’t even know what. Out there and that right, because we sort of were suffering from that same problem where we have all of this data. And in this case, it’s the data model itself is in effect metadata. Right. And being able to figure out what’s what and how it applies is difficult.

And so having something like a platform AI, to help do that, you know, kind of helps me bridge that gap. And, and well, what I’m assuming is ultimately what you’re all about is just create a better app faster. Right. And so, you know, that’s what this is coming down to, Sorry to cut you off when they got to just quickly add there.

Tejas Gadhia: The best part about these low-code apps is when you, when we have more people building things, we have people who are closer to the problems. Solving them with low-code application. Generally, what would happen is I got to hire a developer or talk to my ID team that it guy has no idea what I do in my day-to-day life, the types of problems I faced, the things they’re relying on my description of that problem, which evolve over time.

Or maybe I just can’t describe it very well. And I can’t describe it in terms that they understand. And so that, that real life experience logic is really where you’re. Developer, I would say your skillset really is. Everything goes below it, the baggage of data models and workflows and things like that.

That’s not what a low-code developer wants to worry about. They’re solving a problem because they have deep intimate knowledge of a process, from my firsthand experience perspective. and the rest of it is just basically building blocks to help them translate that knowledge into a digital process or a digital storage system.

Charles Araujo: So let’s switch gears. Cause I think, cause then the next question is, how do I use AI to build better versions of those apps and to make it a better experience for the people that are using the apps I’m building on this low-code platform. And that’s really what the data AI is. And I know for me, so I’m a Zoho Creator user, I’ve played with it and used it.

And I’ll be honest. I see those little at the bottom, on the left side, the little data fields that are all ‘Zia’ right? The Zia. And they scare me a little bit. I’m like, Oh, well, what happens is this, what is this going to do? And I’m not sure what to do with this? So describe some of the use cases of the ways that people can actually use.

This type of technology and in your product or any low-code product that has this and, and, you know, are there, are there guardrails or the things they should be worried about? You know, and what is, what is the real benefit of being able to apply it in a low-code type setting?

Tejas Gadhia: Yeah, good question. Guard rails are extremely important because going back to what I just said, the person building the application has deep domain knowledge or expertise, but not necessarily AI knowledge.

And, you know, it’s kind of, you don’t open the flood Gates of AI capabilities to somebody who doesn’t. I know what they’re doing because they’re just not going to utilize it very well, themselves. And so there has to be some level of guard rails for, I would say the typical user and then some level of hyper customization for a power user who wants to really.

Who knows what they’re doing and just wants to take advantage of the, of the platform itself. At a very basic level, what generally we see is a lot of times our applications and low code are generally customer facing. And so they, some customers interfacing with this data on an internal process.

Let’s make it simple and call it, I’ve got a pizza shop and, you know, a customer places, an order through the app, they order a pizza. the pizza store gets the order updates a status like your Domino’s tracker, basically, right? It tells you which involve in it’s on the route. Click to see a map, all that kind of stuff.

And then the customer gets the piece of info or the pizza at their door. And then I said, how was your experience you get, right. So let’s just take that small aspect of feedback collection, right? generally what happens is, you know, you might have a star rating, one to five and you’ll have like a comment box.

And in that comment box, people can write whatever they want to, and maybe people. Normally, you would have to go through an individually, read through each one of these comment box submissions to figure out what’s going on. And maybe you read a thousand reviews or whatever, and you yourself in your head, can’t really correlate all these thousands and distribute them into what are the primary issues people are facing, right?

Or what are the primary things they like about us? so maybe what we can do is we can analyze all the comment box. Data that we have, and we’ll say, look, people really love our delivery drivers. They think they’re the nicest people in the world they’re super friendly. and you know, that’s a big asset that we have.

We should utilize them more. Maybe take advantage of that in some way, shape or form, but people would think that our crust is burned 90% of the time. I see the word Burke coming up all the time, crispy, overcooked, whatever. And we can take all that information. Organize it, analyze it and identify the things that people like and the things that people don’t like.

And then we can even automatically say, if someone submits a negative review in like an escalator, and I get a message on my phone directly saying, Hey, somebody left a negative review. And imagine if I could call that customer five minutes later and be like, Hey, I’m sorry that pizza’s not working out for you.

Can we send you another one? Can we send you a coupon? Can we do versus normally those might go into like an email box and you read it the next day and you know, You’re not really identifying it. so that’s probably the best real life example of, of a pizza pizza shop that is really analyzing large quantities of comment box, text, and, and analyzing it and making decisions and creating workflows based on that.

Charles Araujo: So I know we’re about to wrap up here, but I do have one last question on that one because I am, what I’m curious about is how do you help your customers overcome what I call the imagination gap? Right. Cause I think the hard part is, is that still is kind of whizzbang neat kind of stuff. Right? It’s bordering science fiction, even though we’re all used to consuming services like that.

Now we all have smart assistants in our phone or in our phones and our homes. And what have you. If I was creating that app, I would probably not think that that would be something that would be available for me to roll out. So how are you helping your customers kind of broaden their view of what’s even possible using a tool like this?

Tejas Gadhia: You’re giving me the alley-oop here Charlie, cause that goes back to the platform, AI level things. And so for us as a platform builder, a provider, ultimately, what we can do is we can identify, you’ve created a textbook called feedback, and you’re getting a bunch of responses. And now we can give you a suggestion inside the product and say, Hey, look, you’ve got this field called feedback or comments, and it’s receiving a bunch of information, right?

Did you know that you could do some analysis based on this, and basically we’ll suggest you to implement the AI for your service. and so that’s the biggest thing. A lot of times it’s you don’t know what you don’t know, like you said before. And so we, as the provider of the platform or any good low-code platform for that matter should be suggesting use cases to the user.

And once you suggest enough, then that imagination starts building in, then they’re like, well, what else can I do? Maybe I want to. Have customers take a picture of the pizza and we’ll analyze every single picture of a pizza. And if any, one of them has like, you know, you’re gonna box a pizza and it’s like that on its side.

And like all mushy or whatever, then we know it’s a bad picture or there’s not enough pepperonis on this pizza. We can analyze all those pictures and say, hell yeah, that’s not enough. Let’s send you more. Next time. We’re sending you a coupon or whatever it is. Right. So it’ll suggest ideas to get the creative juices kind of flowing.

Then hopefully the low-code application developer can take those ideas and then kind of run, run with it from there to build their own use cases. Now.

Charles Araujo: Okay. And I know I’m showing that I’ve watched too many sci-fi movies, but I don’t need to be worried about AI suggesting that it, that we’d build more AI.

Right. It’s that that’s not going to lead to any bad outcomes, right?

Tejas Gadhia: not yet. We’re still, we’re still safe.

Charles Araujo: So I know, I know. All right. As we’re, we’re coming to the end of our time together, and I always want to wrap up with something both a little bit forward looking, but. Still practical. So, so here’s my question for you.

Where’s all this going. And what is the most important thing that an enterprise leader should be doing right now to prepare for that future?

Tejas Gadhia: End on a high note there? Ultimately I think enterprise leaders should really. They should really keep their eyes open to all the possibilities that are out there.

a lot of times there’s fear of kind of allowing everybody in the workforce to build applications could be dangerous and kind of scary and overwhelming, and the fear of people making bad applications. And then you have security and compliance and privacy concerns to take care of. but a lot of these low-code platforms, they incorporate those.

Governance measures for develop for, I would say enterprise IT leaders, and really, they should start with a small pilot program, pick an individual team or department, give them a chance and kind of have them be, I think a popular. The term that’s enterprisey is like a center of excellence. Whether people use it or not just call them your advocates within the organization who, you know, they test something to see if it works, see if it increases their productivity, see if it makes their lives better.

And then you kind of deploy it, just throwing it out there for the whole organization. It’s usually a failing strategy. It’s better to start small and slowly increase the scope over time. And I think people will generally see that letting more people in their organization solve problems based on their own personal experiences and deep knowledge of what they do on their day-to-day life will really increase their company productivity and make their employees happier because they’re not doing as much manual stuff.

That’s just kind of a time-waster.

Charles Araujo: Yeah. You know, not to answer my own question, but it’s certainly something that, you know, I come from the enterprise side and something that I definitely, back then, and certainly see now, as I talk with my friends that are still executives within larger enterprises, is that the more you experiment, the more you sort of open those doors and give people the tools to, as you said, start, you know, the people that are closest to the challenges start giving them the tools to address them that.

It’s a self-fulfilling prophecy, right? As they become more familiar with it, it starts closing that imagination gap. They start, like you said, asking the question, well, what else can I do with it? And so for enterprise leaders, are we able to sort of overcome some of that initial fear. And I think, what we didn’t talk about is this intersection of low code and AI are these two sort of nebulous terms.

They’re very buzzy, very hype-y. And, and that, you know, is anyone who’s been in the trenches for a while, those immediately throw off, you know, all these warning signals. It’s like, well, I need to be really cautious and careful here. And so, you know, getting past that and saying, I can do this, I can experiment and play with this, but in a controlled fashion and that the objective is that we can create real value that I think those organizations are finding tremendous success with them. And I’m sure you’re seeing that with your clients.

Tejas Gadhia: Yeah, for sure. You could have a sort of better.

Charles Araujo: All right. Well, thank you so much for being with us today. I think it’s been a great conversation.

I hope that people look at both what you guys are doing and just explore this idea of how they can bring AI and low-code together to create real value in their organization.

Tejas Gadhia: So thanks for being with us again today and always a pleasure talking to you, Charlie.

Charles Araujo: All right. And that is it for us on this episode of the Intellyx Brainwave podcast. See you next time.

 

©2020 Intellyx LLC. At the time of publishing, Zoho is an Intellyx subscriber. All dialogue in this program represents the expressed opinions of the hosts and guests, and are not necessarily the official position of Intellyx, or any company mentioned or included in this podcast audio or video.