Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Beating PwC & Infosys: How Shubham Garg’s CodeVyasa Won Fortune 500 Clients Without VC Funding
0 Playsin 1 hour
How did Shubham Garg bootstrap CodeVyasa to $XXM ARR without a single dollar of VC funding?
In this episode, we uncover the contrarian playbook behind India's fastest-growing AI-powered IT services firm and why enterprises are ditching SaaS for custom software. Shubham Garg is the Founder and CEO of CodeVyasa, a 600-person bootstrapped IT services company serving unicorns like Mamaearth, UpGrad, and Yatra, plus Fortune 500 clients including HDFC Bank and major oil and gas PSUs. What makes his story remarkable is the journey itself - from failing at a supply chain SaaS startup in 2019 to pivoting into high-value outcome-based engineering services that reached $XXM in annual recurring revenue, all without external funding.
In this candid conversation with host Akshay Datt, Shubham reveals how CodeVyasa helped UpGrad slash their AWS cloud bill by 40% (saving $2M annually), replaced SAP's enterprise software for a government PSU in just two quarters, and is now capitalizing on the AI revolution by selling GenAI implementation services. He shares hard-earned lessons of scaling a services business with no VC money, and his controversial thesis that the era of SaaS dominance is ending as enterprises shift to building custom AI-powered tools in-house.
This episode is essential viewing for bootstrapped founders, service business operators, enterprise sales professionals, and anyone navigating India's IT services boom in the AI era.
What You'll Learn:
👉How Shubham Garg scaled CodeVyasa from zero to $XXM ARR and 600 employees without VC funding using pure customer revenue and cash flow discipline
👉Why outcome-based pricing beats traditional time and material models, and how CodeVyasa competes with giants like Infosys and PwC for Fortune 500 contracts
👉The build versus buy revolution - why enterprises are replacing SaaS platforms like SAP and Salesforce with custom AI-powered software built in-house
👉CodeVyasa's AI enablement strategy including data pipelines, LLM operations, and identity stitching that now represents 20-25% of revenue
#ITServicesIndia #BootstrappedStartup #AIServices #ProductEngineering #OutcomeBasedPricing #DevOpsConsulting #GenAIImplementation #BuildVsBuySoftware #SaaSvsCustimSoftware #IndianUnicorns #DataEngineering #LLMOperations #IdentityStitching #ChinaPlusOneStrategy #FortuneIndiaIT #EnterpriseSales #B2BTechSales #ScalingWithoutVC #CashFlowMasterclass #TechStartupIndia #QAAutomation #CloudOptimization #ReplacingSAP #BootstrappedGrowth
Recommended
Transcript
Reducing AWS Cloud Bill: A First Principles Approach
00:00:00
Speaker
Let's take an example of Upgrad. right Back in the day, it had an AWS cloud bill of $5 million dollars every week year. And we were able to bring down their cloud costs by more than 40%. What made you choose that unsexy startup instead of going to VCs, pitching an idea, raising millions of dollars? I am not a services guy. So my approach of doing this was just not even know if there is a services bill. Just always first principles. We are not even talking about human arts build, all of that. We are playing a role that is outcome driven. Shubham Garg is the founder of Code Vyasa. Code Vyasa is an AI-enabled product engineering firm serving unicorns like Mama Earth and Upgrad and has achieved double-digit million dollar ARR without external funding. No external funding, 10 to 15 million ARR. Most people who started DevShop don't hit this kind of scale so
Shubham Garg's Journey to Founding Code Vyasa
00:00:52
Speaker
fast. How did you do it, man?
00:01:01
Speaker
Shubham, welcome to the Founder Thesis podcast. ah that There are so many... yeah companies you worked with the who have come on this show, it's ah quite a coincidence that like you ah just take me through a little bit of your background, pre-entrepreneurial background, and then we can talk about the the journey of building Code Vyasa, which in itself is a phenomenal story. Got it. you Awesome, Akshay. Thanks for having me. um Great to be here. Like you rightly said, I think we've had so many customers. So,
00:01:35
Speaker
of my you know mentors already on the show. So yeah, great to be here. So my name is Shubham. I am the founder CEO of Code Vyasa. ah Prior to starting Code Vyasa, talking about my early days, I got into IIT Rootki eventually, you know kind of did graduation from Delhi College of Engineering.
00:01:55
Speaker
and post that coincidentally you had all the startups which were picking up in India. right So that was in hindsight the most interesting time to kind of come out of a college, right which is when you saw great opportunities to learn in fast-paced startups.
00:02:12
Speaker
um As it were to happen, right I got into Crownit, worked with Mr. Ashish Munjal for a couple of years, in fact more than that, and that really got me set in terms of you're understanding how and what it takes to run fast paced startup.
00:02:30
Speaker
um So, I mean, most most of us would know what Crownit is. So it was a consumer tech company and we were like one of the category creators back in those days. And my role was predominantly around product growth, bit of GTM.
00:02:46
Speaker
And then eventually I worked for um you know companies like Haptic, which was later acquired by Jio. And then for Muglix, which is... which is also into a similar space these days. So yeah, I've worked for about five, six, five years or so before i um you know started my own company, which is... What was the trigger?
Early Customers and the Simple Pitch: Automating QA and DevOps
00:03:10
Speaker
like ah You started, i think, almost in the middle of COVID... Just before that actually.
00:03:16
Speaker
Just before. Okay. Yeah. Okay. So what was the trigger? Like what made you want to quit Mowgliks, which was a unicorn by that time? Absolutely. And I was one of the initial members there, right? So it was a great place. It always has been a great place to work. ah Look, I always wanted to start a tech company. ah I mean, that that was always the idea. Right. I would say while I was working at Crownit, I always wanted to kind of start out, but I'd always look at it the other way around, which is it took me another three years to actually start my thing is is how I look at it.
00:03:53
Speaker
But yeah, I think honestly, in terms of space, I wanted to do B2B tech. Having worked in the ecosystem for five, six years, I had even back then a great amount of network to tap into for... um early adoption in terms of getting those customers for customer feedback. you know When you're kind of coming up with different use cases, particularly in a services context, you need that validation, that that feedback on your hypothesis. So that really came in generously.
00:04:28
Speaker
right so we had Who were the early customers? like whom did yeah yeah And what was your pitch to them? what were you ah like Absolutely. i think initial few customers were Policy Bazaar Upgrad, Sunstone, ah you know, Ashish always coming to the rescue. And then we, ah you know, started working with larger companies, Uber.
00:04:49
Speaker
And what was the pitch? Like, what were you? Absolutely. So, um look, back then, um I think our pitch was very simple that we are... specializing in two or three use cases. Back in the day, automation QA was ah you know theoretically a massive impact that you can create if you are able to deliver well. So we did that really, really well for upgrade.
00:05:13
Speaker
Then we you know you know did various use cases around DevOps for all of these mass market consumer companies. And that was driving crazy impact. So just to give you an example, right? um um What does DevOps mean for people who are not from the world of software?
00:05:30
Speaker
So DevOps would have multiple facets to it, right? One of the things is, let's say, um let's take an example of Upgrad, right? Back in the day, it had an AWS cloud bill of $5 million dollars every year.
00:05:43
Speaker
Now, if you were to think of it as, let's say, a warehouse to optimize for in a simplistic example, right? And we were able to optimize, make the ah you know ah the cloud robust and while optimizing it to that level, and we were able to bring down their cloud cost by more than 40%. So think about it. We were helping them save literally a couple of millions of dollars every year, and we are not even charging a recurring fee for that. So we came into um the ecosystem as experts of doing that.
00:06:15
Speaker
We did that project for, let's say, in a phase of couple of quarters. We charged a fee for that and that kept giving them this high a saving ah forever almost.
00:06:27
Speaker
So then we know we kind of doubled down. So we did multiple use cases around platform engineering. um Another example could be... like so What does you know platform engineering mean? like Let's keep breaking down the jargons as we go along. Absolutely. So if you think about it, right let's let's talk mobile, for example. So in a traditional sense, or even now, right you have so many companies who have a different native mobile app and a hybrid, sorry, a different Android app and a different iOS app.
00:06:56
Speaker
you have two different teams building that. um Invariably, you would see Android team is at least couple of releases ahead of iOS. the The consumer is not sure why the app looks different on different devices. The support team did not understand why a customer is asking a thing which is not there now because it's there on the old app.
00:07:16
Speaker
And we could we could do that, which is move them to a hybrid platform, let's say ah Flutter or React Native. And one of the things they would leverage a company like Codeway for that would be that if you were if they were to do that in-house, that would probably take them three quarters with all the rework and those kind of things associated with that.
00:07:37
Speaker
But as a company that specializes doing for you know the very use case for mass market consumer companies, we were able to do that in almost half the time. So there you have impact in terms of your voodoo market, in terms of your how
Outcome-Driven Projects vs Traditional Models
00:07:51
Speaker
cleanly and tightly you know the migration is done.
00:07:55
Speaker
And then you could extrapolate that across different use cases across platform. Okay. Okay. So QA, automation, DevOps, product engineering, this was the early days, like the pre-AI offerings that we were doing. Is there a reason why you didn't want to do a funded startup? Because you had worked all previous startups you worked in were all funded, building products, and here you are doing...
00:08:23
Speaker
Something which is normally seen as ah not so sexy. Back in the day, yes. Yeah, you know you like that service delivery kind of a thing where it's a man-hour-based billing. Correct. Not seen as a sexy startup to build. Correct. What made you choose that unsexy startup instead of like going to VCs, pitching an idea, raising millions of dollars? Correct.
00:08:46
Speaker
Absolutely. um And you know such a great question and relevant question because if you think about it, the whole hypothesis is now changing the other way around. I'll come to that later. But you know on the first part, right um look, the approach you were taking right was not a people-centric approach, which is you know just multiply, put a dollar value for every human hour and invoice that to the customer.
00:09:09
Speaker
Like I told you, in terms of the examples or the initial first use cases you were doing, right we were attached to an outcome. We honestly thought we were ah you know ah great at delivering those particular outcomes.
00:09:23
Speaker
And it's about identifying customers who are looking for that use case to be done. And the moment, even back in the day, Akshay, when we made ah the discussion around outcome based versus a TNM, timing material kind of a base, you garner a lot more command even in customer conversations, right?
00:09:46
Speaker
And I do hear you when you say this being almost like ah the not so sexy area. But like you said, I think this was even something which I realized after getting into it was that apart from this being a really great business in terms of rep repeatability of ah revenue, in terms of the the Profit margins, yeah right most things which ah you know get less spoken about, particularly in the starting of the journey, these are things that are real problem statements. right And if you're able to you know not spread yourself too thin, particularly early on in your journey and focus on delivering those outcomes, there's a great amount of repeatability in terms of customer revenue.
00:10:30
Speaker
And that's what okay kind of followed throughout, right? Over the last five years, we have not tried to do hundreds of things. ah More often than not, we have tried to stay away from a time and material model model. Again, not in all the cases you can do without it because there are dynamics. Let's say if you work with a large bank, um banks and their executives are attuned to buying in a certain way, right?
00:10:57
Speaker
They wouldn't want to do... ah an outcome based thing, particularly when the project sizes are more than three quarters. right And then it becomes a time and material. But even in that, right when you have similar projects, live projects that you can demonstrate, that again garners a great amount of command, respect, right and and credibility as an organization in the sales cycle.
00:11:23
Speaker
So the traditional way in which this industry works is the pod model, right? I believe one pod has like a bunch of engineers and a business analyst and a couple of such ah diverse skill sets, like a bunch of anywhere from five to 10 people or could be bigger also. And you charge ah like per month, this pod will cost you so much and that pod is available to solve the customer problem. um How does the...
00:11:52
Speaker
outcome based pricing work? Like like the pod model pricing is clear time and money kind of a pricing, but sorry, that time time and money, is that what you call it?
00:12:03
Speaker
Time and material. Time and material. Okay. um How does this outcome based pricing work? Right. um So outcome based pricing is fairly simple. I mean, you talk about the outcome that you intend to deliver to the customer. um You subsequently talk about ah your, like you rightly put it, the part or the competition of it and the tenure they will work.
00:12:28
Speaker
But the moment you ah are kind of linking your invoices to an outcome, you are giving such a great amount of confidence to the customer that unlike a traditional IT services model where the customer gets invoiced irrespective of the outcome, here there is skin in the game from the vendor side, which is CodeBiasa. And the customer already has looked at of the live projects where we've been able to deliver it.
00:12:55
Speaker
Right? So ultimately... is it the Easy to measure the outcome. like Like that's what I'm wondering. How do you negotiate on what you will charge a customer? Is there a very clear definition? um You know, people could say that this is an outcome which you didn't cause, my own team caused it, whatever. like like Like, how do you, how how does it work? like Absolutely. how is that Yeah, great point again. I think, um like you said in the starting, right, there are areas, there are projects where you have to fall back to a T&M time and material kind of a model.
00:13:26
Speaker
But um the moment you have these conversations, the customer even wants to kind of take a step ahead, right? Which is talk about the measurable outcomes wherein not everything is measurable, right? So you have that classic input metric versus output metric, right? You could talk about how we have gone through the entire process, right? And then measure what can be measured.
00:13:50
Speaker
right And that is how you flip the conversation. right so um and And I think over the last, ah even before starting CodeViasa, given I was in the space for good five years or so, right I had a great amount of connections and all of them, like it's any which way is a very tightly knitted ecosystem.
00:14:08
Speaker
So all of those things really came to a rescue initially. where you are kind of looking to build it ground up, you need that initial, you know, kind of two, three million in annual revenue for the company to, you know, kind of build itself into a machinery.
00:14:25
Speaker
Until then, it's, you know, mostly, um you know, kind of a founder hustle. So that initial momentum that that for you to take off was extremely generously kind of given by, you know again, great companies. So you know I'll talk about three, four companies that we have worked for a good three, four years before they went public. You know you had Pine Labs.
00:14:49
Speaker
We have been working with Pine Labs for the last four years now, one of our initial customers. You've had Yatra. you know We have been working for such such a long time. The partnership is you know great has been great. And then you had Mama Earth and various few examples. right So it's almost like you know you're kind of growing along with your customer.
00:15:10
Speaker
Your role keeps on kind of evolving. so um I think we still do pod-based um kind of hourly billing model along with outcome-based projects. So in a given state, you would have a mid-sized company like a Yatra or um ah you know Pine Labs working on both the models on multiple fronts.
00:15:33
Speaker
and that's how that's what i mean depends on the nature of project like whether it's a broad base or outcome okay and when you deliver when you deliver right the customer has a larger trust in you which means repeat business which means annual contracts you know getting renewed uh year after year so you know your business was more on the software engineering side when did the pivot to ai happen Right. I think, um again, very
Enhancing Productivity with AI-Enabled Pods
00:16:05
Speaker
interesting. um Over the last couple of years, you have had such, such massive changes in the ecosystem. right First, you had um you know NVIDIA really you know kind of giving computation power in abundance. right And then you had GPT. So if you combine both of these things, they have really changed the ecosystem.
00:16:25
Speaker
ah Our shift to AI um has been, I would say, ah on multiple aspects. So for example, i think the first adoption we had organizationally was where we started using um you know human layer along with automation tools, AI related things that amplify the productivity. right So you have, we spoke about the pod model, you have, you know great engineers from our side along with modern AI tools right that really work as a pod and really take um you know the efficiency to a different level.
00:17:04
Speaker
ah Subsequently, now you have different streams emerging. right So um we could be playing a very, very different role at, let's say, a different startup, which could be wherein we are playing the role wherein we act as a true engineering partner that enables them to adopt AI.
00:17:27
Speaker
So what that means is we do the plumbing, we do the foundational work, which could be you know building data pipelines, which could be um you know building agents, LLM ops, that really um sets the customer for for AI adoption.
00:17:47
Speaker
right And in this sense, we are creating a mode for ourselves, which is we are not even talking about human arts build, all of that. We are playing a role that is outcome driven and that lays the foundation layer for companies to really kind of build their AI systems on top of that.
00:18:07
Speaker
Okay. So there are... two ways in which how your pivot to AI has happened, if I understand correctly. One is the pod itself became human plus AI pods instead of a pod of just humans. And the second is ah you specifically started offering AI enablement services to help ah enterprises adopt AI. um I'll talk of the first first the first model or the first adoption before going to the second.
00:18:35
Speaker
When a pod gets it gets to be a hybrid pod. ah What does that mean? What kind of tools are they using? And how do you then, ah how does the TNM calculation happen then? Does still remain the same way? Or is there a premium that you're able to charge because your pod has ai tools in it?
00:18:55
Speaker
help me understand how that happens. Both of these things are, do act as a counterbalancing factor, right? So one is the customer expectation now is getting higher and higher, ah right? At the same time there, um status quo, their expectation in terms of, let's call it productivity per headcount has really gone up.
00:19:18
Speaker
Now it's about how do you use different automation tools, let's say a co-pilot that can really enhance the workflow, that can really enhance the productivity of the pod.
00:19:30
Speaker
right And that really, you know, kind of has to keep going up and up, even more than what it is going for the customer in-house, for them to continue, continue, um you know, trusting on Codeway Asa.
00:19:43
Speaker
So that would be a very, I would say, pragmatic, a you know, fairly simple way to look at how that's evolving. So there you're saying you don't necessarily have...
00:19:55
Speaker
power to increase your margin by using AI because all other devop agencies or development dev shops are also offering AI powered pods and customer is now expecting you to offer an AI powered pod so without necessarily increasing your but pricing too much Yeah, I think um I'll explain how the margins have increased for us.
00:20:19
Speaker
um So, for example, wherein you needed, you know, let's say a pod of 10 people. Can you as a team, as a pod, have ways to increase the productivity, right? By, let's say, a 2x, you know, and then that does not necessarily mean the customer is going to slash the pricing into half.
00:20:42
Speaker
That does not mean that. So by doing all of this, you are increasing your organizationally, you're increasing your revenue per headcount. And that's what I mean, that AI is actually kind of helping companies like us, you know, kind of make more sense of the human resources, the AI resources that all of us have on our hands. So a pod which earlier needed 20 people now,
00:21:09
Speaker
A 10 people pod can deliver the same outcome and the pod pricing stays the same. Therefore, revenue per employee. I'll not say 20 to 10. I think another way to look at this is um what is the impact? What is the productivity you can do?
00:21:26
Speaker
So, for example, if you have, let's say a team achieving, let's say 10 story points in a given day, for example, can that be increased to 12? Because um You can't look at the customer's project being finite. I mean, there's always a new feature to build. There's always a different problem statement to solve.
00:21:46
Speaker
right So that's not shrinking. right If you can keep on increasing your productivity, your headcount or rather the revenue per headcount increases and eventually you're making larger impact in the customer's landscape.
00:21:59
Speaker
And that ah results into, you know, generally happier customers. That keeps on resulting into more and more projects is the larger, I would say, way to look at this. What is the story point thing that you mentioned?
00:22:13
Speaker
You said that if team is able to hit 12 story points instead of 10. Yeah. So for example, story points are ways, let's say story points are units to calculate the amount of engineering effort ri required.
00:22:25
Speaker
So just like you and I calculate distance in kilometers or miles, engineers would roughly, let's say, say that, hey, this website is 10 story points, this mobile app is 12.
00:22:36
Speaker
And that is their prime of face judgment of how much how many hours it will take them to go. So StoryPoints is just, ah I would say, for example, let's say for a Zepto kind of an app, StoryPoint would be like, you can log in. You can search, you can add to cart, you can check out, you can put your payment information. These these would be like story points. These would be features. These would be features. And then you could put correspondingly the story points, which is to build search feature.
00:23:05
Speaker
I would need XRs to, okay you know, change this flow, checkout flow. This would take me Y story points. So a way to kind of contextualize the engineering effort, it would go into building...
00:23:18
Speaker
any feature okay got it got it okay okay okay understood and your ah pricing has some story point targets associated with it so if you hit more story points then you are able to earn better something like that um Not really. So ah in terms of outcome based approach, it's very very objective, right? Outcome based is like pure productivity led gains. if you hit it ah got it Correct. In terms of what model, it's never because story points in itself is very subjective, right? So like you said, two engineers, at different companies have different ways to look at story points, right? So ours is, like I said, I think it's about input versus output metrics.
00:23:58
Speaker
What are the kind of you know kind of progress you're able to make on a given project, along with consistent feedback from the customer. So there comes the customer success layer we have and things like that.
00:24:13
Speaker
Okay, got it. A little bit context about your business today. um you told me you have about a headcount of 600. How is this headcount utilized in terms of a split between like say a pure TNM billing versus outcome-based billing, what's the split look like over there? I would say it's almost 50-50%, right? So 50% based, we are outcome-based, right? And 50%, the rest 50% is the pod model, which is the retainer model. I think one of the things we have taken very aggressively is how do we push the outcome based model even at large customers.
00:24:57
Speaker
So like I said, I think ah typically banks or let's say other Fortune 500 customers are attuned to you know kind of buying in the retainer model.
00:25:08
Speaker
But over the last couple of years, we have had the you know large Fortune 500 companies, large banks agreeing to signing outcome-based fixed cost projects. And that is in the longer run going to be our moat, right? Which is we are not just, let's say, um a helping hand or a cog in the wheel. We are able to you know put our skin in the game and really kind of inch towards achieving a target in an engineering context.
00:25:36
Speaker
And that's what we are itching more and more. This outcome-based pricing, um has the industry also started moving towards it in general? Like say an Infosys, would Infosys also be doing outcome-based pricing or is Infosys doing more of pod model pricing? Yeah, i think it would be um hybrid. So it would be ah large pods,
00:26:00
Speaker
along with some tangible progress or measurable outcomes. And one of the reasons it's very hard to do that is, like I said, let's imagine, let's say if the large IT services providers are working with government agencies,
00:26:17
Speaker
Now, it's exceedingly hard for them to you know put a fixed cost to a project that's going to take hundreds of people and you know going to take another two years.
00:26:28
Speaker
Because the project is humongous. Now, at the outset, it becomes impossible for most human beings to put a fixed dollar value. without risking, um you know, but things either which ways.
00:26:41
Speaker
And that's why, like I said, becomes a hybrid, ah you know, wherein you do a pod, multi-pod kind of a thing, along with phase-wise, maybe tentative cost. I think that's the hybrid or the middle path, which which companies are doing.
00:26:57
Speaker
Okay. Okay. Okay. Got it. Interesting. Okay. ah Okay. So...
AI Enablement Services and Infrastructure
00:27:02
Speaker
Again, coming back to that basic question of that pivot to AI, one way in which the pivot to AI happened is pods became AI-powered pods, not just humans. And the second is the nature of business. So you are now not just selling DevOps and QA automation and... and ah engineering services but you are now selling ai adoption uh services what are these ai adoption services can you just like explain you said lln ops you said data pipeline what what do these mean correct so let's say let's talk about a large bank right or let's say talk about any new upstart uh for them to be ai ready they need to have the infra underlining infra
00:27:44
Speaker
ah ah that has to be there in place for you know any magical AI to kind of operate on top of that. So in again, a very simplistic term, I would say, think of it as a plumbing or the foundation layer for anything to kind of think of it as a launchpad, right? you are You are kind of setting the ground rules.
00:28:04
Speaker
And that could mean building data pipelines, data warehouse optimization. um It could be LLM ops. It could be. Describe this. I'm not a techie. I don't know what is LLM ops. What is data pipeline?
00:28:18
Speaker
Got it. Got it. So, for example, if you think about it, let's let's take an example of and understand what, let's say, a data pipeline or a data problem looks like. Let's start. Let's let's look at a large food tech company in India.
00:28:32
Speaker
and we work with both both the large companies, good tech companies in India. What would be their problem statement? Their problem statement would be that they have customer data across. So they have data, order data, they have cancellations, they have promotions given to different customers, they have rider data, they have you know support related queries.
00:28:52
Speaker
Now, in a traditional sense, so All of these data was being captured by these large companies. right The problem was they were not stitched together.
00:29:03
Speaker
That was not unified. So let's say a consumer, of let's say Shubham, ordered something, has raised a ticket on support, has had this transaction history, history past history.
00:29:14
Speaker
All of those systems systematically is being captured. But is that unified in real time? for let's say a customer support agent at Swiggy to make sense of.
00:29:26
Speaker
Until this is all there in one place, no amount of modern AI or or any tool can take this to a level where it will amplify the user experience for me as a consumer here.
00:29:41
Speaker
And this is practically what I'm talking about when I say that you build the foundation layer, which is you solve the problem of bringing all of these data points at one place right for them to make sense of it and then build something magnificent on top of this. So this is one, but let's say I've explained you a problem statement around data. Similarly, if you look at into a banks or an NBFC kind of a setup,
00:30:09
Speaker
what are the different triggers you have for a given consumer again, right? Different attributes, be it his civil, be it his other social signals, be it his repayment till now for the loan he has taken, ah be it, you know, let's say all the pending cases he or she has, the new cases that he or she might have had, the idea How do you make sense of all of these things, right? For...
00:30:37
Speaker
for let's say any of the modern AI stack to tell you that, hey, here is a 40% risk that this guy is not going to give you the repay the loan back in time.
00:30:48
Speaker
Now, once you build this foundation layer is how you can then amplify your productivity, your insights, or the consumer experience using some of these modern AI stack.
00:31:01
Speaker
So this is what I mean in the second business model, which is making um enterprises ai ready. right And that's a very different volume. The reason I'm putting that in a separate bucket, Akshay, is that in the first model, we are building new products. We are you know kind of um driving automation. But here in the second model, it is purely a fundamental problem that is now there to be solved. which Again, so for example, we spoke about identity stitching, right? A problem statement which was not very relevant two years ago because any which ways the identity was being stitched,
00:31:36
Speaker
probably with a latency of three hours that's still manageable but in today's context for any of the ai models to work right you have to have all of those things in real time uh stitched across the different data points and that is exactly what my identity stitching you mean like say his itr and his ah credit bureau report and his repayment history all of that being integrated and available at one place real time instead of every three hours some sort of integration happening So think of identity stitching or identity resolution as making sense of all the attributes, all the signals that you have for a consumer and putting that together to build the most appropriate persona.
00:32:21
Speaker
Simple as that, right? User profiling, KYC, simple, very simple. Now the example the the attributes you spoke about are prevalent in let's say an NBFC context, which I explained when I was giving that you know loan repayment example. But let's say in a Swiggy's case, right which is a very very different thing, right which is look at orders, look at cancellations, look at customer support tickets, look at past transactions, those kinds of things.
00:32:46
Speaker
So um just kind of making enterprises ah to be able to make sense of all the data points which they store and then let AI models kind of amplify the productivity or the user experience.
00:33:00
Speaker
Okay, got it. code And what LLM Ops? LLM Ops is, if you think about it, how do you kind of fine tune what you already have right in terms of large language models? right How do you contextualize that even more given your nature of operations? right ah Think of it as, let's say, a bit of a private cloud. right um And then there could be other areas. There could be vector DVs. There could be agents. I'm sure agents is something which everyone talks about these days right and other use cases around this.
00:33:35
Speaker
Okay, okay, okay. Got it. Understood. How would Okay, let me ask ah again one more business breakdown. So, of your revenue, how much of it now comes from the software engineering business and how much of it comes from this AI adoption business?
00:33:52
Speaker
Yeah, yeah, yeah. So to be very honest, I think this is one area which is picking up. So as of now, I think about 20-25% revenue, not more than that. And that's because um everyone is still figuring out, you know, everyone is seeing, or is this a trend which is prevalent only on LinkedIn? Or is it something which can really change my team's productivity?
00:34:17
Speaker
Right. Okay. ah But that 20% is going to increase even as our larger pie increases. Right. And that's where ah the direction of the space is in which in which you and I operate. That's where the growth is over the next few years. Absolutely. I mean, you simply look at um the last batch of Y Combinator, which is always a great reflection of where people are aspiring to start their companies, you will see more than half of them, like even even much, much more than half of them into this space. And that tells you where the next two, three years and and companies are going to be.
00:34:55
Speaker
Okay. what is ah What are like the ah areas in which companies are embracing AI and what are the areas where they are tentative?
00:35:08
Speaker
Yeah, yeah, yeah. Great. So um I think fundamentally everyone, the first answer to most people is, can we amplify the user experience?
00:35:20
Speaker
Right? Then when you kind of think through, when you kind of shop around the market, there's a counterbalancing factor, which is, hey, what is the associated risk? What if the implementation goes south?
00:35:35
Speaker
And that is where I think, and and all of these I discover and I talk to customers of different company sizes, which is as of today, most people are trying to first leverage um you know AI to increase internal productivity.
00:35:52
Speaker
and then take it to the consumer. However, I mean, when it started all of these discussions, it was the other way around. But I think it this approach makes a lot more sense.
00:36:04
Speaker
So leverage internal productivity means like, say, um using AI for managing your finance or managing your procurement or for your software team to use coding apps instead of writing code that's what you mean so it could be let's say um again in ops heavy companies right if you have certain processes which required which have required let's say hundreds and thousands of back office operations people can you figure out a way
00:36:39
Speaker
to do a proof of concept that lets you do all of that with only, let's say, 10% of human layer, if not, if not, less. now Even if that were to go south, it's it's more work, but it's not a depleted consumer experience. It's not going to affect your top line.
00:36:58
Speaker
right So you know most CXOs will have to you know kind of plan it in a way wherein if anything goes unexpected or anything goes in a direction they did not anticipate, it still does not hit them in terms of top line.
00:37:13
Speaker
Okay. Is this driven by cost-saving as a motivator? yeah Yeah, I think ultimately um those are the two metrics most people care about, right? Which is how much revenue and how much profit. And cost saving ah is not the delta cost saving you're looking at.
00:37:31
Speaker
We are looking at something which could really, really bring it down. So for example, if you think about it, um let's say contact centers, right?
00:37:42
Speaker
If you think about it, PPO would still charge a significant cost. um However, if you can build that agentic layer, let's say talking it from a bank's perspective, right train it on your custom data, right which is um all the different systems you have, integrate that deeply.
00:38:05
Speaker
From there on, you would have superior quality and your cost comes crashing down. And that is why I think, um you know, i what I foresee is most people taking a build, operate, transfer kind of ah approach, which is where, you know, IT services, the so-called IT services like CodeBiasa, right, would play a much, much larger game, right, which is help these customers, right,
00:38:33
Speaker
build these modern AI stack systems, agentic workflows, run it for you know a few quarters, and eventually the customer takes the IP in-house.
00:38:46
Speaker
And you know that is that is going to be a very big shift because I think in Indian or even global context, right the last 12 years, 10 years has been all about SaaS and how powerful the platforms have have gotten.
00:39:01
Speaker
But I think with where rob the modern stack today is, right where everything is open source, where you know you have local tools so that proprietary workflow that proprietary tech is not proprietary anymore that's pretty much out there for most companies to give themselves a chance to build on that and the real power will now move to implementation so it moves from i would say the power of platform to power of implementation which is how well can you context contextualize that given product or platform
00:39:39
Speaker
into your ecosystem, how well it can connect with different workflows, different systems that you have, and can you keep the IP in-house. And I think that is when that is why you know most people do today are very, very bullish even on the traditional IT service providers even today, right? Because the role changes, but the pie size keeps on increasing.
00:40:03
Speaker
um And with you know this hypothesis, right which is that SaaS companies potentially what can be, a large part of that can be built and you know kind of kept in-house, that's going to increase the buy for IT services even even bigger.
00:40:20
Speaker
right And the role becomes much more strategic. So what you're saying is, this is like the classic build versus buy dilemma.
00:40:33
Speaker
Typically large organizations like, let's say, a Walmart would probably not buy an off-the-shelf retail POS device for the checkout counters. They would probably build something in-house versus a ah small a store would buy something off-the-shelf for the retail POS. Absolutely.
00:40:54
Speaker
And there are um I'll give you another example. And, you know, ah so for a large oil and gas company, ah Code Vyasa has built them, ah you know, a customized procure to pay.
00:41:09
Speaker
Now that oil and gas company, a fairly large public sector company, right has been using a large says SaaS window for that. But the problems innate to that SaaS were ah the high amount of annual recurring cost. right Beyond that, one-size-fits-all kind of of approach, right which is that given those SaaS companies have to accommodate for so many different user personas,
00:41:36
Speaker
they are almost much, much more powerful and complicated than what a simple user would want, right, in an enterprise context. right And we were able to build that in about a couple of quarters that that project is live and that's that's there.
00:41:54
Speaker
And eventually when I talked to the CIO, in fact, we were planning phase two and multiple phases for that. They are now planning to replace the large ERP which they have with a custom tool that Codeway Asa can build for them.
00:42:08
Speaker
So if you think about an SAP or Oracle kind of ERP. Absolutely. It's SAP only that they're trying to replace. They've started with ah the P2P, which is the entire vendor management system.
00:42:21
Speaker
And they have been able to, they are like absolutely liking it given how customized, how easy to use. And in the long term, how cost friendly it is. In the longer term, I i use the phrase. And then they'll eventually do it for an SAP.
00:42:34
Speaker
and And that thesis will play out so well across customer sizes. Right. So, um you know, probably if you have 10 Salesforce licenses or Zoho licenses, it might not make sense to build it in-house. But the moment you talk to a mid-sized consumer company, which uses thousands of CRM licenses,
00:42:52
Speaker
right very very soon you will see that they would want to build it keep the ip in-house and the economies of scale really make a lot of sense in the longer run and i think that if if it turns out true uh which is strongly what i feel will really change the game uh for both the sites be it saas players be it it service providers like us Okay, I'm going to play the devil's advocate here. So, ah you know, on the one hand, you i hear you that ah it is, ah thanks to AI, it is a lot easier to build custom software. Why force yourself to learn Salesforce when you can build your own Salesforce, which is exactly working the way you're people behave. ah You don't have to change their behavior for them to adopt a Salesforce. um
00:43:42
Speaker
But the other way of looking at it is because cost of building software and SaaS is going down so much, you might have ah thousands of SaaS tools which are very, very industry specific. So you could have ah somebody who's building a CRM just for the oil and gas industry or somebody who's building a procurement tool just for those 20, 30, 50 customers in the oil and gas industry. ah Very, very ah very niche yeah yeah yeah very vertical SaaS tools. ah
00:44:15
Speaker
Why do you feel the future is not thousands of SaaS tools, but the future is ah everyone building their ah in-house tools?
00:44:27
Speaker
I think a great question and you know part of this is what I've discovered talking to last year, which is one of the things which is not which has not been spoken about till now is the core IP staying the vendor and the vendor locks them in for multi-year contracts.
00:44:43
Speaker
right So it's not just about the cost, it's about can you keep your IP in house? Is there a risk that possesses that you're posing if you keep all of the data in in the vendors controlled?
00:44:56
Speaker
And secondly, if you think about it, the whole model of having thousands of SaaS would undercut the fundamental hypothesis that there could be multiple players because the market might be deep for 10 multi-billion dollars to be built.
00:45:10
Speaker
But the moment you have thousands of 30 million dollar companies, right it suddenly is not so exciting. So I'm not saying everything falls apart, everything moves to id services.
00:45:22
Speaker
It's about what percentage stays with the large shark SaaS companies. And I think a great um example, the parallels I keep drawing in my you know kind of hypothesis formation is look at D2C. Over the last six years, seven years, right?
00:45:40
Speaker
D2C companies have completely abstracted the large FMCGs, right? Initially, they had IP of dealership. They had the dealer networks. They had the brand which takes which took years and decades to build.
00:45:54
Speaker
But today if you have an Instagram, if you have a Shopify, you can next day you know pretty much compete with ah any product. You could get your niche and you could compete.
00:46:04
Speaker
You could build a personal brand. you could come you could You could have a deeper rapport with your customer. You could have a better customer support. So all the things that were IP to large companies have been now, you know, pretty much, you know, every city would have multiple D2C companies, right? Similarly, that might just turn out in the case of SaaS vendors versus now modern AI democratizing, um you know, the SaaS platforms.
00:46:35
Speaker
Okay. Okay. Interesting. oh What is the... ah you know, what's typically the decision-making buying process of these kind of large enterprises and how do you really crack these? You must have learned about cracking enterprise sales. I'd i'd love to get some of those insights from you. Right.
00:46:58
Speaker
Right, right.
Cracking Enterprise Sales and Competing with Giants
00:46:59
Speaker
I think um this is fundamentally the most exciting thing for me, right? So like I said, before starting CodeOS, I was a GTM person. And that has also evolved, Akshay, drastically. So for example, in initial, I would say, first 20 sales were, you know, you calling up your CTO friends and, you know, kind of telling them that, hey, this is one use case I can do.
00:47:25
Speaker
and you kind of doing even a POC for them and then eventually landing the contract. The next few customers was largely about building a storyline around your specialization.
00:47:39
Speaker
right So you kind of research very deeply into any customer segment, be it you know the size of the company or let's say on a x-axis, the industry.
00:47:49
Speaker
And you backtrack what are the use cases that will make sense. And while the initial first 20 customers were direct access, like just like I can call you, I can call Ashish that, hey, this is what I am doing. Can you, would you be interested?
00:48:03
Speaker
The next 70, 80 customers, right, become more about building a GTM machine, rate right? You have to kind of be very, very sharp in terms of your research on them.
00:48:16
Speaker
backtrack what value you can deliver to them and keep all of that very, very clean, very non salesy. Right. And then you continue kind of building that GTM engine. So that's, that's on the outbound side. I think from a decision decision making perspective, it's, it's very, very different. Like you would imagine across enterprises, mid market and upstarts and upstarts, you know, you have almost,
00:48:42
Speaker
one week of sales cycle. If you are able to build a case, you're speaking to the founder directly and you know there would invariably be common connects or an investor might have put you in touch.
00:48:53
Speaker
And if it has to close, it will close very, very soon. In case of mid-market, the structure increases. You might have, let's say, you know a few VPs across VP product, VP engineering to kind of make sense of Then there come would be you know kind of of an audit or a due diligence on your company on those levels.
00:49:19
Speaker
And then the deal gets signed. So this could take anywhere from, i would say, four to five weeks starting and could take two months. right That's the sales cycle.
00:49:29
Speaker
The moment you come to enterprises, you realize that there is only as much you can do in this quarter, right which is you kind of take a very, very patient approach, which is because you realize that procurement at a Fortune 500 company would take a month's time to negotiate. um Again, a negotiation could pretty much be done like, let's say Akshay and I, if we were to negotiate, we would either negotiate in a day or not be able to do it, but will not take a month's time. But that's how the nature of operations are, right? And that's why you have to be very, very sharp in terms of
00:50:04
Speaker
uh being a sales rep across these customer segments which is understanding what is the nature of uh the play here and just just trying to expedite as much as possible but a large watching 500 company could could take you know four to six months how do you even become a part of their consideration set because you when you go to a large fortune 500 company you would be competing with these publicly publicly listed companies like say an infosys or a wipro or like say a global logic ah how ah how do you really become a part of their pool of vendors that they're evaluating
00:50:45
Speaker
Absolutely. So ah great question. So the example which I gave, which is the oil and gas company for which we made, um apparently the last two vendors there were PwC and God Vyasa.
00:50:56
Speaker
And we were able to seal the deal. um you know Goes without saying a PwC has all the credentials, but how do you build a case for yourself is how can you at every stage of t decision and making go above and beyond?
00:51:11
Speaker
right So if the customer wants you to submit a proposal outlining the nuances of their ecosystem and how the new product will land and merge with all of that, instead of just doing that,
00:51:26
Speaker
Can you create a meaningful proof of concept? ah okay Can you go down to the customer's office with your engineering team and demonstrate a similar product that you have done?
00:51:39
Speaker
Could be even for a large bank, not even within the same industry. but something which emulates their level of nuances. So it's pretty much very simple to understand, hard to execute, right? But very simple to and understand, which is at every step, can you outrun, can you, you know, go above and beyond what a partner at VWC would do and his team. Simple. Okay. Right. And like, yeah.
00:52:06
Speaker
I mean, like the typical PwC partner would have learned how to make decks, whereas you have learned how to make and MVPs and you instead of going with a deck, you would invest in an MVP and show them, hey, this is what it could look like, you know, like some sort of a rough experience. Okay, very interesting. And um as a matter of fact, I know it might sound absurd. We have PwC as our customer.
00:52:32
Speaker
yeah so We work with PwC also and I have been a you know couple of PwC partners across different divisions. right are One of them is a close friend, a mentor, right who tells me his learnings running this for the last two decades.
00:52:49
Speaker
right And I think it's very interesting to pick their brains in terms of how they perceive sales cycle versus how upstarts like you know ou ourselves, how we look at things.
00:53:02
Speaker
And there's no absolute right, there's no absolute wrong, right but it's it's about a difference in approach and how you can kind of try to make the best of both the sites.
00:53:13
Speaker
So that makes it really interesting. yeah Okay. ah You are now running a 600... headcount business you you're what like early 30s i'm guessing yeah yeah um and this is no external fund so far right absolutely not uh completely homegrown ah never um you know kind of raised uh money from from outside and uh yeah you know this is phenomenal like uh hitting this kind of scale what what kind of arr are you at currently
00:53:45
Speaker
Yeah, so we are in that 10 to 15 million ARR. and How did you do it, man? No external funding, 10 to 15 million ARR within a couple of years. What's the... no I mean, most people who started DevShop don't hit this kind of scale so fast.
00:54:04
Speaker
Yeah, yeah just just just one small thing. It did not take us two years. We are in our fifth year. We completed... four and a years. Yeah, I mean, like minor details, right? Nonetheless, it is a phenomenal achievement. and most people who start dev shops don't hit this kind of scale. How did you do it?
00:54:23
Speaker
I think when I reflect, one of the things is that I am not a services guy. Like I told you, right, I've always worked in product companies. So my approach of doing this was just not even know if there is a services playbook.
00:54:38
Speaker
Just always first principles focus on first getting an initial 20 customers. Then, like I said, you graduate to a phase where you are trying to build early team across GTM, operations, engineering, and then you try to create. So at our stage right now, the last thing we would want to do is hustle on a deal.
00:54:57
Speaker
It's about, can you have repeatability um to, you know, for us to get past that 20, 25 million revenue mark, right? And for that, No amount of my personal hustle can can do it.
00:55:10
Speaker
um And if I have to hustle, right, i have to hustle to create the best team on on marketing, on sales, on finance, and engineering, of course, right? So now it's a very, very different game versus what it was even a couple of years ago.
00:55:27
Speaker
And I'm not talking about because of the landscape change. I'm talking about the organizational maturity, right? Which is how do you kind of structure and scale? Because I think They often call this a dead zone, right? 5 to 10 million. We've been just past that, which is when you get 5 million, you have hustled your way through to to reach here, right? But getting past a 10 million is a very, very different ballgame, right? So we have just about done that. And I think now we realize that the game changes pretty much every, I would say, 5 to 8 million mark after every 5 million, after every 8 million.
00:56:03
Speaker
How does it change? Like like i've I've never seen that kind of scale, right? So how did you as a leader change when you you were in the 1 to 5 million and now you're in the, like say, the 10 to 20 million ah
Scaling Challenges and Organizational Changes
00:56:16
Speaker
bucket? How did you change in terms of how you manage, what you spend your time doing, stuff like that? So um I think initially 1 5 million is about
00:56:28
Speaker
trying to hustle your way on every deal. When I say your way, which is literally, you know, me trying to... You're saying like sales, very sales focused. It could be even before that, which is GTM, which is outbound, which would be on, let's say, the agreement slash negotiation phase with a customer.
00:56:49
Speaker
And there comes in engineering. So i would say that engineering execution does not change. It changes from a customer segment to another customer segment, which is As an engineer, if I'm working on with Uber as compared to if I'm working with, let's say, a manpower group, that's going to change.
00:57:11
Speaker
Because there are two different of customers or code users, so that's different. But the game changes all the more when you're trying to win Uber, win an Uber, win a manpower group.
00:57:24
Speaker
do outbound to them, you know, to, to present your capabilities in, in the form of sales, negotiate contracts, uh, procurement rates, those kinds of things. That's a very, very different game.
00:57:35
Speaker
Uh, when you are of one to 5 million and five to 15 million. Uh, how, how, how is it different? Like when you're doing the same, uh, like contract negotiations, uh, so I tell you a macro answer to that would be that at a one to 5 million, uh, You are trying to master your own self in terms of doing all of that.
00:57:57
Speaker
When you get past 5 million, you want your VP sales, director sales to be able to do that consistently. Okay. You want to be able to step back and not be the hero of the pitch, but see if your team can be the hero of the pitch to win that client. Yeah, I think when you're in that 1 to 5 million mark, you're trying to be the best version of yourself, right? Which is to kind of hustle way on everything.
00:58:25
Speaker
When you kind of get past that mark, there comes problems of repeatability. The hustle you are able to do as a founder, can you have a team that can replicate across 20 instances in parallel?
00:58:41
Speaker
And for that comes, I think, the hard part of recruitment plus training plus you know having that dynamics where people want to really give it their all every single day.
00:58:54
Speaker
And that's a, I'll not say harder game or an easier game, that's a different game. So and then you have to have systems, processes that as ah as a founder let you identify the shortcomings, the areas of improvements in in the fastest way possible.
00:59:15
Speaker
And that is where systems, processes, dashboards, reviews, those kind of retraining, those kind of things come come into the picture. How do you build ah a team of people who want to give it their all every single day? like Like that sounds incredibly hard to do. Yeah, yeah no I think, no, it is. And um I think what I realized is, first of all, there's never a straight answer to that.
00:59:43
Speaker
So there's no playbook that can give you 100% results all the time. The question I'm assuming is that how do you maximize your chances?
00:59:53
Speaker
I think which is always about, you know, finding people who really want to build a career. So you would have for different roles. So for example, in finance, you have great folks, right?
01:00:06
Speaker
But you would want people with a lot more stable persona track record there because that's the nature of finance ops versus let's say in more aggressive dynamic ah roles, let's say in STR or in account executive.
01:00:26
Speaker
where you need people who might not be from, let's say, a PwC or a McKinsey, but have done, let's say, you know this and this, and more importantly, are really um into learning and building a network for themselves.
01:00:44
Speaker
So that's one example. If I were to answer that for CSM, I would say a people with a lot more empathy, because they're dealing not just with the customer, they're also dealing with our engineering team.
01:00:58
Speaker
And the dynamics could be not so transparent in CSM, right? So you need somebody who's not as volatile as probably in the AET and a lot more, I would say, of ah you know, kind of stoic on on that front. Okay. So Shubham, let me end with... the this, like you're a fairly young founder yourself and um you probably have a lot more empathy for somebody who's just starting their career today. What
Advice for Newcomers: Hard Work, Mentorship, and Risks
01:01:25
Speaker
advice would you like to give that person?
01:01:28
Speaker
Oh, uh, anybody who's starting the career, not necessarily a company, right? yes Yeah. Yeah. Yeah. I think, um, uh, The number one thing is you can't you know overstate the the ah value of hard work. I think you have so many resources at your disposal, but it's about are people willing to zero in um you know and really kind of make the most of everything which is available and most importantly, their time. So I think that would be the number one skill.
01:01:58
Speaker
uh you know i've seen so many of my like the ceos i've worked with working absolutely around the clock like so so hard i think that's the most uh unstated less stated kind of a thing that's one second would be i think having having mentors really really help um in the starting i e you know separately when we were talking i told about how Ashish has has helped me not just in the isn't that tenure when I was at Crownit but even beyond, right? How of you know so many other mentors of mine, right? Sandeep and so many others have been helpful throughout.
01:02:36
Speaker
ah That goes a long way. it It almost gives you the insight that you would come up with an year down the line. So kind of are able to almost see the future because somebody who has done that, right?
01:02:52
Speaker
And I till date, I you know reach out to so many people. and and And the funny part and the best part is that people are very generous. if they If they look at you as somebody who's working hard and is somebody who is coming up with structured questions, they'll respect you, they'll help you.
01:03:10
Speaker
And the third would be, I think, risk taking. I think um beyond the first two things, you have to be risk taking. Of course, you could calculate. But there is no way you can kind of make it big without kind of taking that bold move.
01:03:26
Speaker
So, yeah. Thank you so much for your time, Shubham. It was a real pleasure. Awesome. Actually, it was lovely to chat. Thank you so much.