Gen AI Series: Gen AI in the Enterprise with iCertis, a conversation with Sudarshan Chitre, SVP Product

Enterprise SaaS solutions are rapidly adopting Gen AI technology to solve business problems. Learn about iCertis, a leader in Contract Lifecycle Management and how they are deploying Gen AI technology to offer a number of CoPilots to allow enterprises to gain intelligent insights into their contracts and to perform risk assessments.

Sudarshan Chitre brings over 25 years of experience in developing Data & AI products, having worked with leading technology firms including Microsoft, Amazon, and Oracle, as well as several innovative startups. Throughout his career, Sudarshan has transitioned from roles in software development to executive leadership, including that of CEO. He remains deeply engaged in the evolving landscape of technology as an avid learner and an advocate for its practical applications. He holds co-inventorship of more than 10 patents in the fields of databases, data synchronization, and IoT.

Currently, he leads the AI Product and Engineering team at Icertis. Sudarshan is academically credentialed with a BS and MS in Computer Science and an MBA from the Haas School of Business, UC Berkeley, and Columbia Business School.

Brought to you by TiE Seattle
Hosts: Shirish Nadkarni and Gowri Shankar
Producers: Minee Verma and Eesha Jain
YouTube Channel: https://www.youtube.com/@fromstartuptoexitpodcast

Transcript

SPEAKER_00 0:01

Welcome to the Startup to Exit Podcast, where we'll bring you work class entrepreneurs and VCs to share their heartbreak success stories and paper. This podcast has been brought to you by TIE Seattle. Thai is a global nonprofit that focuses on foster entrepreneurship. TIE Seattle offers a great program including the both articles or creation. We encourage you to become a Thai member so you can gain access to this great podcast. To become a member, please visit www.seattle.ti.org.

SPEAKER_02 0:43

Hello everybody. Welcome to another episode of From Startup to Exit. Uh, a really exciting one. This is the continuation of our Gen AI series. And uh this is the fourth one that we are going to uh uh record and release for you. Uh a very exciting guest with uh from a very exciting Seattle company. I will let Sherish uh do the introduction. But prior to that, uh thank you all for supporting us and continuing to subscribe. Please let all your friends know. We are available on all platforms. It's available. Uh uh my name is Gauri Shankar. For those who haven't who are the first-time listeners, uh, I with my co-host Sri Shatkarni, I serve along with him on the board of Thai Seattle. Uh Sherish is a serial entrepreneur, much like me, but he is also has the distinction of being a serial author. He has two books. Uh, I hope you buy both his books, the latest being from uh winner take all, his first one being from Startup to Exit. Thank you to Tai Seattle for producing the show and uh being a big supporter of us. Over to you, Shirish, to introduce our guest and uh take it away.

SPEAKER_01 2:04

Thank you, Gauri. Um very pleased to have our guest uh today, Sudarshan Chitray from iService. Uh Sudarshan is the SVP of product for iService, a contract lifecycle management company based out of Bellevue or Seattle, Washington. Uh Sudarshan brings over 25 years of experience in developing data and AI products. He's worked for companies like Microsoft, Amazon, and Oracle, as well as many innovative startups. Uh he actually holds 10 patents uh in the fields of databases, data, and IoT. Currently, he leads the AI product and engineering team at iCervice. So welcome, Sudarshan.

SPEAKER_03 2:45

Thank you. Thank you for having me. Very excited to be here today.

SPEAKER_01 2:49

Great. So before we get started, let's talk a little bit about your career. Uh, you've had quite a varied career uh working on the engineering side at Microsoft and Amazon and a little bit at uh at Oracle. Uh then you had your own startup. Uh after that, you um you know were a CEO and then you took on engineering and then product role at iService. So uh be helpful if you could introduce yourself and talk about your career and how you ended up at iService.

SPEAKER_03 3:20

Oh, definitely, definitely. It's been uh it's been fun. It's been over about very class. I would I think of it in in three parts. You know, my first part of my career is very much uh, you know, it's as an engineer, started as a software developer. Uh came back to Galvie to work for Microsoft in '97, spent the first one third of my career just learning the how of it. Uh building software, work in Windows, file systems, databases, a lot of the spirit systems, uh built some of the early uh Azure services at Microsoft and uh you know as a natural progression kind of grew to leading the engineering teams and and and product teams. I think the second second uh uh part of the career was like kind of moving out of the out from an IC engineer to kind of leading teams and moving to other companies, getting a multi-company experience, and uh and the and the last one third, last few years has been in smaller uh uh companies, uh more startups. I did a couple of the ones where it was two person starting from two person startup, some post Series A, and now with iCertis, uh uh you know, we are uh I would say we we are a startup from the way we work, but we're you know we're pretty pretty large for a startup. Uh we're not public yet, but uh uh you know uh we are we are uh getting there. Um how I got to iCertis, I want I um I worked in AI uh in my previous startup, which which uh uh was also a Seattle-based startup, uh Ken Sai, which which uh we uh had an exit. Uh we uh worked in AI, so I spent uh several years uh applying AI in healthcare, and so after that uh phase was over, I was looking for uh uh staying within AI and with with what's happening with Gen AI, you know, you can't sit on the sidelines. So uh when I saw the opportunity to work uh at iCertis, I was I was uh really excited to join forces.

SPEAKER_01 5:21

That's great. So at uh iCertis, you manage both uh product and engineering. Can you tell talk a little bit about your role there?

SPEAKER_03 5:31

Yeah, so uh so I'm SVB for AI and and what we're doing with iSertis is we're gonna since it's an uh iCertis has had AI products for a while. So it's it's uh 2018 Italy was the first AI product that was launched in uh at iCertis. Uh but with Gen AI, uh there is uh we're looking at things in a new way. How can we look at the problems and apply Gen AI in a in a uh in a completely different way of what does this technology bring and how do we apply it to to the problems we're trying to solve? And it's it is to do that we're gonna carve out an organization which which includes product and engineering and and uh uh separately from from the platform teams. And so I head the uh combined product and engineering or I report to our CTO uh and co-founder um uh Monish. Um uh and and so we are running it is as a unit uh so we can run fast and we can build more AI uh Gen AI based products to the market sooner.

SPEAKER_01 6:35

Okay, great. So let's start with some basics about uh you know what iService does. You know, I remember um when I used to work for large companies like Microsoft and then uh Blackberry, you know, contract management was essentially uh exchanging documents over email, contracts over email, you know, deadlining them, you know, negotiating the contracts with your vendor or your partner. And then uh, you know, I would literally uh file it away in my file system, and I'm sure you know the legal department would have a document management system or some kind of a filing system to file it. That was the extent of contract management.

SPEAKER_04 7:20

Yeah.

SPEAKER_01 7:21

Um so um tell us a little bit about what iService does and why do we need why do we need a solution like iService? I understand it costs you know for large enterprises can be in the millions of dollars. That's a pretty expensive purchase. You know, what does it at a basic level, at a basic level, what does iService do?

SPEAKER_03 7:40

Yeah, yeah, good question. And I think I think the the way I think about it in a simple sense is as a business, you know, we you're either buying something, you're selling something, you're hiring employees, or even if you're you know uh signing a lease at the at the place that where you work, everything that you do is uh is uh is in contract with uh with another party. It's an agreement, it's uh it's you have decided to have a set of rules that will govern your relation, your business relation. And so contract is really capturing that intent of how are we going to do business together? And so um yes, it is in a document, yes, there's a lot of legalese in it, but but really what it serves is it really gives you the rules of engagement in terms of how are we going to conduct the business when we work together, when we buy from you, when we sell to you, when are you gonna deliver? Are there are there uh guarantees, are there penalties, are there gonna be volume discounts, rebates? And so it can start very simple, but it can get get complex and nuanced, especially in larger, larger enterprises. So uh uh what iCERT is does is it looks at this. How can we help uh businesses uh do two things? One is how can we help them memorialize this intent that they are looking to do business with all of these parties? And this can be thousands of contracts a month that that enterprises can do. It can go to tens of thousands or hundreds of thousands of contracts. So we're talking about a large number of these agreements between two parties. So one of the things I status does is what is the life cycle of this? Sometimes it's many years, and there are amendments and so on. So if you think of it from the intention that started when two parties shook hands, so then we memorialize it, what's the process of that? Like you send back and forth documents, you do redlining, negotiating until the point you get to a signature. But then post-signature, also are we conducting our business that is in line with what we what the intent was? And so if you think of the whole life cycle, it that's where ISAT's software comes in. It it allows you to draft a contract, negotiate a contract, sign a contract, uh manage obligations, you know, uh uh to uh look at look at the business operations that are happening and see if they're in line with the contract, if you have further amendments, you know, take care of those and understand the relation between all of these documents and and data points. So it is really looking at that holistically, and that's the CLM uh space. And like like you rightly pointed out, all of this uh is fairly new, it didn't exist, and iSight is one of the companies that that helped shape how to think about uh CLM in a structured way. So, you know, um it used to be set of documents stored in a file system and forgotten after they assigned, and unless you have to go. But with this, with with as we are moving towards um uh more, I mean obviously it's all digital, but the expectation is with new changes and regulations and other factors coming in, the expectation for you to stay uh compliant is also a big one. So it's not just about just about uh doing this contract and negotiating on the pricing, but uh can you look step back and look at all of your contracts together? Does the system allow you to do that? And uh by asking questions like what is my exposure, supply uh supplier side exposure, what is my risks? Are there pockets where my suppliers are concentrated? And so if there is a supply chain disruption, am I more um likely to get impacted? There are a lot of these holistic things that become very interesting from a business uh continuity and and risk management perspective as well. So uh all of this makes it important to have a really good contract management system at the core of your business. If you're not able to sign contracts, you will not be able to run your business very quickly. And so uh, although it's kind of one of those back office uh you know things, it's it's it's not what everyone talks about every day. Uh but uh it is really one of those things that uh allows your business to run smoothly and you will know, just like security, you will know if it's not working well.

SPEAKER_01 11:50

So is there a uh workflow uh component uh to um uh to iService where you know the different parties are you know working together, negotiating the contract that manages the entire workflow between the elements?

SPEAKER_03 12:05

Yeah, absolutely. From from the drafting, then from the approval chain, and then also uh which you know, in large enterprises, different people are responsible for different types of uh contracts, uh, but also different people are responsible for different parts of the contract. So sometimes your expertise and this this SLA piece has to go to this team, you know, this this risk uh and insurance piece has to go to this other set of teams. So uh there is uh workflow managing all of that, who needs to approve when something changed and a red line came back, who needs to go look at it? Are these standard variations of our clause? For example, like if you if some clause came back and said the other party said, like we need to re-reword it, are there standard alternatives that from the library of the enterprise that are already signed off on that you can pick from, or does this change need an escalation and a sign-off from an approval? So there's a lot that goes goes into streamlining that workflow that happens within the software system.

SPEAKER_01 13:04

I see, and and then you have a library of of standard language for different clauses that you can that that lawyers can insert and so forth.

SPEAKER_03 13:13

And yeah, yeah, and that that makes it like easier for uh these negotiations because you have a library to draw from versus coming up with new new clauses on the file. I see.

SPEAKER_01 13:25

So now once a contract is uh is signed, uh what is the process? What how does the CLM system uh manage uh the process so that everybody knows what the obligations are and you'll track in against those obligations?

SPEAKER_03 13:41

Yeah, and I'll I'll I'll here's what I'll kind of um maybe maybe bring in, you know, uh selfishly uh what I think is the big opportunity here for for Gen AI and NAI and what I'm super excited about right now. Um I think there there are traditionally there have been lots of applications that extract oblig, for example, obligations from contract and then manage them. Um and and and then and we have that as well. And we have uh obligation management workflows where you extract, okay, these are all these obligations that we have signed up for, it's part of this contract. Now it's signed, let's manage them. What are the deadlines? What are our deliverables? Are we on track or not? But I think when you where the apartment, there is a lot that is missed today. And I think I think because it's it's not easy to always stay in the intent of the contract, because you need an interpreter in the between many times, because a lot of these things are am I allowed to do this? Like let's say you're ordering something as an incremental or uh purchase order you're creating, or or or you're you're making a decision at the at the procurement side, or at and the sales team is trying to make a decision, trying to uh sell something and maybe bundle something. Is this in line with the contract we have in place? And and that many times is hard to answer because that requires some judgment, that requires some some interpretation and understanding. And I think what is exciting is now with the large language models, being able to better understand that, now you can unlock that value, like under really understand the nuances that are not only in one contract, but you take what's in one contract and you combine with what else is happening in your enterprise, and then you can really do those some of those things where your post-signature activities can be improved and can be more efficient and in line with what's in the contracts.

SPEAKER_01 15:34

Right.

SPEAKER_03 15:35

Right.

SPEAKER_01 15:35

But essentially, if I were to summarize what a contract lifecycle management uh you know software does or service, you know, there's a work, there's a workflow element to manage the creation of the contract, uh, there's a library of standard clauses that can be inserted and managed and so forth. And then there's a post-contract piece, which is managing the obligations. Is that does that cover all the elements of CLM or am I missing something?

SPEAKER_03 16:03

Yeah, it's managing obligations, it's is is managing compliance uh regulations, um uh and so yeah, so looking at across the board uh risk exposure. So you can you can do analytics on top of all of your contracts uh together. So so you're right, you you covered those areas, like it's it's the pre-signature and post-signature as the two boards.

SPEAKER_01 16:24

Right, right. So um I understand that uh you know um Gartner has created uh oh its own category, you know. Before you didn't even have a category called contract lifecycle management, and there's now that category, and you're considered the the leader in that in that category. So, what is it that differentiates you versus uh other players in this market?

SPEAKER_03 16:50

I I think uh looking at this uh more holistically, I think I think what iCetis has always followed is is what is the value to the business at the end of the day? Where are the areas where you can move the needle? Like the contract is is is managing and building the contract is one aspect of it, but we've always kind of focused on on the end business value that that having a more efficient contract life cycle management system brings. And so our strength is like we've approached this as a platform where it is not gonna be we're gonna have out-of-the-box solution that does everything for you and in a silo, but it's been built from day one as an as a platform that's able to cater to different verticals with the same product. So I think one of the big differentiators is our design principles from the very beginning have been our same products uh supports a finance, uh supports a bank, supports a provider, a payer, uh, you know, it's uh supports a retailer. Uh it's able to manage contracts of all of these different, very different contracts, very different verticals, because the way the product is built is it is it is uh very much as a platform that allows for different kinds of contracts to be brought in. And and uh through our SI partnerships, we are able to then configure these things specific to each vertical and and really finish the last mile. Um, because there are also integrations, we have lots of integrations with other systems within the enterprise. Um SAP, Dynamics, Salesforce, because it just by itself, the contract is just giving you one piece of the person. So when you when you connect this with the with the data model that understands you know purchase orders or uh uh coming in from somewhere else, for example, now it's become stronger in terms of answering questions. So I would say one of the main things that we have, there's several things that differentiate us. One is our platform mindset that we we want to be able to do a lot of things uh by able to be extensible and and configurable. Our SI uh and ISV partnerships that then allow us to go build end-to-end uh high-value solutions to it. Uh, and then I think we have been innovating at a very high pace. I mean, with Gen AI in the last six months, we've we've uh we were among the first ones to build co-pilots to market, and we have a very uh rich roadmap right now on lots of exciting new things in Gen AI. So I think again, you know, you have to keep continuing to innovate, so it's never done. So uh, but I feel like those things I think uh has given us the edge.

SPEAKER_01 19:33

Got it. Okay, let me turn it over to Gavri. Uh, we'd love to explore more the AI and the Gen AI components that you're working on that you're responsible for. Yeah, yeah, yeah.

SPEAKER_02 19:42

Great. Thank you, Shriesh. Uh hi, Sudashan. Uh, very excited to uh excited to talk to you. Um, I mean, you used two words when you started this conversation. Uh, intent and interpreter connect my mind through language. And uh the memorialization of the intent of two parties uh uh in a language that's interpreted after the memorialization. That seems like your definition of a contract. And it may be modern today, that but it's time immemorial that uh as humans have traded, they've always used some form to keep this thing alive. Now, uh uh AI uh is dependent on large language models as we have been exposed to today. Looks like your company and what you're doing was tailor-made for the way AI has come to come to us or come to everybody, right? Uh, because you have language, you have intent, and you need interpretation. Seems like the you you are you're like sort of you almost have that vision, like get it done. So the first is uh when you guys decided, hey, this AI thing has to be integral, that uh and iSert is decided that you you need a leader just for AI. What was the what was the sort of turning point and what was the pivot that uh said uh the aha moment saying, hey, we are ahead of everybody and we could change essentially commerce in in a company, right? The way they do business, which otherwise, as Shirish pointed out, was uh retrieving file system. That's you know, you there's a problem, you go retrieve it from a filing system all the way from big filing boxes to your email uh inboxes. So, how did you guys think of AI given the landscape it was it came to the world as we know it now?

SPEAKER_03 21:49

Yeah, now it's a lot of it is before my time. So I I will I will be very careful in in in you know uh saying it on behalf of others, but I'll share it in the way I have seen it. I I've heard it from others as I came into joined the company. So ICITIS has been looking at contract intelligence for a while. Contract intelligence is a term ICITES has used for a long time. It isn't gen AI, it's just intelligence, contract intelligence. And a big part of what uh dealing with documents is to understand what's in the document. And and a lot of the content was in just PDF documents. This is not even word necessarily. So as a result, ICTS has always uh invested in technology to extract information from a document. And previous to this, it was a lot of uh NLP, a lot of um um extraction of entities uh from you know OCR, PDF, tables, stuff. So before 18 months, there was still a lot of AI applied to just getting words out of a PDF document and in in the most accurate way possible. Uh and that it itself was was recognized that that was a need, was recognized by by not just ISIC, but that's that was that was something A was applied to. I think you asked the question on on when was the aha aha moment or was there an aha moment? And I I I think I think it's we all kind of saw this all together. Like this was one of the first times when something like they say, you know, it was a 20 years in the work and an overnight success um at the same time. And I think I think things just came together where people looked at what that this is now really approachable to everybody. This technology, I mean, having worked in AI in the previous company, remember, we have had to explain what is machine learning. We had to explain before we could even sell anything, we would have to spend two meetings just explaining what this technology is. We don't have to do that today. We we you when we go in, our customers push us, oh, you can't do this yet. Okay, when is that coming in your roadmap? When is that coming? Because the the proliferation of and uh of the understanding of what capabilities of this new technology is there is is completely different. We've never seen this kind of a pace at which technology has been made that the market has been educated on a new technology so fast. And so I think part of it was that moment that's saying we gotta move fast, we have to, this can't be the pace as usual. This has to be a different differently run. And I think lots of companies are recognizing that. And I think at ISURT is also that this is different. This the the potential here is way different. It's a 10x more potential of what we can do, and like you rightly pointed out, we're all about language. You know, legal uh documents are are all about language, interpreting what what are what is our application, what's our penalty, what should what we can and cannot do. So I think last year around summer is when we were on we released our initial copilots. Uh uh, we did that very quickly, very fast, after it was made available in on and we have built on Asure. So you know, uh I just have released Asure Open AI, you know, we were one of the first ones to kind of we were working with them even before it in pre-release form already, and and we're able to kind of release this co-pilots, and then we got a lot of validation when we have had an amazing traction on these products in the last six months. I mean, this has been the fastest going product line at iCertis uh uh in its in its journey of 15 years, and so that's been the validation that there is this is different and that it has to be uh, like you said, focused on separately. We've got to have a separate art, we've got to move faster and have a more more quicker feedback loop as well on the market.

SPEAKER_02 25:45

Sure. So um as you uh went into AI, at least from a out from just from our viewers' perspective, right? Gen A has been initially ChatGPT, now co-pilot from Microsoft, and others are also doing something of uh similar nature, right? Uh other language models are releasing their own versions of co-pilot or their own versions of chat GPT. So so when you look at the enterprise, enterprises have rhythm, right? They're working on some rhythm, and change has been hard in enterprises. Technologies have come along, but they are quite happy with what they're doing and takes a while to get change in. Why uh first for for our viewers, just kind of uh take a simple use case of how you view your co-pilot working in enterprise, and then if you could expand on why enterprises are calling you as opposed to your salespeople calling them, saying, hey, this is our next product. What's the what's the change in the velocity of intake of your product? So, first is if you could just define so we all have the same understanding.

SPEAKER_03 26:54

Yeah, yeah. I'll give in a couple of examples. Maybe maybe one uh on on how this technology can be can be beneficial to an enterprise, right? So the way we if you think of look at a document today, um, just any contract, and there's let's say there's a service legal agreement or an SLA language or a clause in a document, uh, typically when you have to go look at look at that and say, you know, is this good? Is this is this, you know, can I you're reviewing this document. Today you can just go to a large language model and say, can you suggest me an alternate language for this clause? Right in this document.

SPEAKER_04 27:33

Yes.

SPEAKER_03 27:33

And that in itself is is just 10x like a technology advancement. But now we all like, oh, obviously, that's not a big deal. We have we have like already we are all saying that, right? That itself is amazing. But the ability for for you to just generate that in the so so if you if you think of it, that uh that what is happening is the scope is the document, and and uh the technology is is is a uh uh just a basic LLM against a document. The way we're seeing this mature, uh, the is is the next level question. A better question to ask than this would be can you suggest me an SLA language for this document, an alternate one, based on the past performance of this supplier. So you're you're doing a negotiation with supplier. Now you're asking an interesting question that you really wanted to ask because you didn't want the same language for everybody. Maybe some supplier is really bad in delivery, they've missed three deliveries in the last six months, and you may want to have a tighter penalty or a bigger penalty and a tighter SLA for them because you know you've learned. And this information is not in the document. This information is in the end in the enterprise outside the context of the margins of your of your contract, right? Uh, what would be an even better question to ask is can can you uh I got an invoice from it from someone, and can you look at this invoice, compare it with the purchase order that this is for? I don't know what which purchase order this is for. Can you go check if the goods were delivered on time and at quality as written in the contract? And can you also make sure that I am getting all the volume discounts and rebates and applying all the penalties for this supplier and tell me if this number is right in the earnings? This is a huge business process today. Yes, and all of this now can be done with technology, just with because now we bring we understand what's in the contract, we have unlocked that, we take data points through adapters to different parts of the enterprise, and we are bringing that together and asking that that that higher level question. So, so to this isn't one example of what is now possible. So you can really have a connected, intelligent enterprise that is actually allowing you, giving you the answer that you really want to ask the to the question you really want to ask versus the question that you had to ask, because that was the limit of the time. So, this is one example of how we see where the potential of this is, and this is not too far away. We are actually building this stuff right now. So these things is now within within reach. And so that's to answer your question, like this is this is what excites us. This is this is an ex one example of them. There are several of these scenarios where the other one is on regulation side, where a new regulation comes in, and you have a very short time now to go figure out across your uh supplier base uh are you meeting these regulations or not? So uh happy to happy to give you a little bit more on it. But that's that's one example.

SPEAKER_02 30:28

Those are good those are good those are great examples. So uh as an enterprise or dops, right? They are getting they are hearing about uh genai from every possible uh technology vendor that they have. Yeah, or even anybody they deal with, there's there's this right. But you sit in the core of uh three things that you mentioned. One is uh my procurement, second is uh execution of that procurement, third is payment. You just in your example connected these two. Now, um your partner in this case, OpenAI and Microsoft, will also have its own dynamics and others doing it. How do enterprises look at you and saying, hey, I would have to use your co-pilot versus say Microsoft's co-pilot versus so on and so forth? How do they resolve that conflict uh internally, or how do you guys help them resolve it?

SPEAKER_03 31:26

Yeah, yeah, a good question. And and and to be honest, everyone's learning. This is the everyone's, and I'll share with you some of the things we are observing. That you know, we are we're working very closely with some of the largest enterprises. So we are really able to get a front seat at some of these transformations that are happening in these enterprises. And I think it starts with first the acknowledgement that there is a lot of potential here, and this technology can really help uh businesses get more efficient and make smarter decisions. So then the question is okay, how do I bring okay? I I have this and I have this, I have Microsoft and I have OpenA and I have 20 other uh potential LLMs. Uh, not only that, I'm also uh I'm investing in building my own LLM. Many enterprises are also looking at you know taking their data and having you know training and building LLMs and experimenting with that as well. So uh the way we we have seen this is again, you have to approach it with an open architecture, and which is where the platform approach for us works out. So we're not saying you have to do everything within our product. So what we are saying is there is a piece that that is not at the table today, which is the contract we are bringing to the table. Now, now, how do you take that piece of intelligence you combine with the other pieces of intelligence that other co-pilots and other things are helping you with? And how do these things come together to solve a business problem at a higher higher level? And that is where we are seeing a lot of uh alignment and traction because then you're solving, and this is where our SI partnerships help us, where you're solving really a top-level problem where where you're bringing one piece of the puzzle to it sometimes, and sometimes you're you're envisioning it completely, but you're providing one piece of it, and through partnerships you're getting that more specifically on co-pilot proliferation is an is something that we we went through this in the very like in I think in six or eight months time, the industry went through the realization that we can't really have a copilot for every single application. And I think even yeah, there's consolidation of having now a common copilot. That uh now again, whose common co-pilot is the common co-pilot? So there's further consolidation. But I think where it's probably gonna head to is that it's gonna attach to the knowledge worker more than it's attaching to an application where you know it's your assistant. And it maybe it draws intelligence from various applications and various intelligent systems in the back end, but to you it's a common, common interface. And I think I think that it's gonna surface in your in your app office applications, it's gonna surface in your uh knowledge worker applications, but it's gonna have plugins and it's gonna have uh connectors that that all the other enterprise applications and applications like ours will feed into and provide their value. So again, you know, this is all evolving, but we feel like it's going to get closer to the knowledge worker, at least that's my my view as a more of an assistant. You know what my calendar looks like, you know what what I talked to you about yesterday, and so my context today as I will expect it to be maintained. If I'm talking about invoice or whether I'm talking about a contract or I'm talking about travel uh for my business travel, like understanding across that will be becomes interesting.

SPEAKER_02 34:46

So but seems like you have a core knowledge of how uh you know the intent of the enterprise is captured, at least on in their contracts, right? You you by by sheer position, you have that knowledge. Others have other knowledges, but not the core knowledge, and this is the only place where commerce is valued. The other places add to the value, but it is the only place where the value is affected or even multiplied, right? So in many ways, uh, as a leader of iCertis for AI, you you have both you have both the problems, right? You have you got to build your own AI engine, but you also have to work with every other AI engine uh that comes into the enterprise. How are you um looking at that saying? Because you got you got to build your own while letting others feed off you or feeding into you. One of two, you know, it's always so. How as a leader are you sort of uh looking at that problem within the enterprise?

SPEAKER_03 35:52

Yeah, I think it's a huge opportunity. To be honest, I I I feel like there is there is I don't think one company is going to be able to solve the enterprise's intelligence and smart decision-making needs, and so but you got you the way ICERTIS is positioned, as you rightly pointed out, as because we are we we are we're looking at the intent, like this is the reason for all of the other things that are happening, all the operations and transactions that are happening in the enterprise, at least from a commerce perspective, like the the nucleus of it or the brains of it, at least the intent is in here. And so we we do see uh us as the conduit to business value. And so the way we are approaching this is really we are thinking of it as an ecosystem where we can and we are we are doing this at two levels. One is at the at an LLM adapter level, like more concretely in our architecture, where we support all LLMs within our copiles. So you so you could have an and and and all meaning uh also together in the sense of of of hybrid uh application. So we are embracing the the uh hybrid LLM world at the level of the level of models, but then at the level of plugging into the workflow at the right time. So if you need some information, let's say, for example, you want information that comes from even outside the enterprise, uh, like uh uh credit ratings of some some companies or some financial stability numbers that are maybe provided independently. Uh, we are embracing a plugin framework which is a lot of the co-pulser as well, where that can be plugged into into our copilot. So we can then, in the context of the same question that I was talking about, like uh can you suggest me an SLA language for this supplier, given the uh known financial risk of this supplier? And the known financial risk of the supplier comes from outside the enterprise, maybe from a service or subscription that you have you have to that is a plug-in into the iCervice copilot, for example. So those are the ways I think I I I I feel this is exciting that there is that we get to be in this center to be able to, but it is going to be uh it is going to only succeed if we have an ecosystem and we make it easy to plug in this intelligence because it's not going to be one model or one co-pilot that gives you all the answers. It's it's going to rely on this other data points from from within the enterprise and from outside the enterprise as well. But that's how that's how I I have I am kind of uh aligning our roadmap and what we are building to be able to kind of be there uh to kind of nurture that ecosystem around our our platform.

SPEAKER_02 38:37

So it looks like your role of your positioning of open architecture and embracing the ecosystem is helping you in the enterprise, instead of being saying, Yeah, yeah, it's mine and you play in mine, you're you're saying let's all play together. And eventually enterprises are going to demand that of you if you don't. But yeah, it looks like you're taking that as a fundamental philosophy of your of building your products.

SPEAKER_03 39:01

Yes, absolutely. And and and also in in in conversations with our with our current customers and prospects, you know, we are getting inputs and validation of like this is the the way to kind of think about think about uh really bringing value. At the end of the day, we want to bring value to our customers, and and if they have a model that they have invested in that that they want to leverage, then it makes sense for us to be to think about how can we leverage the investments they have already made in our while from bringing in new value from our process.

SPEAKER_02 39:36

Let's work with that shift the question to a slightly different. So you're a leader who's got both engineering and AI, so that means you're dealing with very smart people. On the one hand, you're in the trenches fighting the war for talent, on the other hand, you could be uh just closing the trenches because you don't need that many people as your AI gets better and you can processes get better. Look, uh, tell us from as a just as a people leader running an organization, how are you balancing this where you have to attract people at the same time, you don't need as much as you had before, simply because some of it is better done in the new world order.

SPEAKER_03 40:25

Um good question. And I think this is this is this is a very uh top of mind question for all of us working in this space, right? Because you know, the the the there is an there are multiple sides to this to this opportunity, and and and there are there's an impact of of of of these efficiencies as well. For me personally, on product and engineering side, to be very honest, I I don't think that there is going that that applies in some sense that you're gonna name less less people. I don't I actually feel like the this unlocks so much more potential to do that we have actually looking, we are expanding to get more people to to build more things, to be able to do more more things from and from that's the skill set is changing. That I that I would say. I think the uh there is definitely a shift in the the skill set of the folks you need. Uh and and uh uh with a lot of these uh especially in the last uh you know 18 months with this new technologies, the the skill set has moved up up a stack up up the stack a little bit. So so there is there is some shift in the kind of people you need. I don't think there is yet a shift in the requiring less people. Because I think the demand just catches up with the expectation catches up. Like what was expected to be done in six months is expected to be done in three weeks. And and I think that's what ends up ends up happening. And so I think I think that's one part of it. I I feel so that so there's a change in skill set, but there's also a change in expectation of how quickly can this technology help me, and how quickly can you release a product and build something and that's so if you have your GitHub co-pilot, but you're also also then expected to that as a result now you everyone's a 10x developer. Uh and and so that I don't know it's a good or bad thing, but but there is a I there is at least on the product engineering side. But I I do feel like there is there is definitely an human aspect on the overall if you step back and and look at the labor force overall. I I think a lot of the uh lot of the ROI of of Gen AI based applications is is coming from from the the uh the overall labor spending. Right? So I think I think there is there is a bigger macro discussion here around how does that impact and how does that how do how what do you what do you do in terms of you know addressing that uh and and and uh being conscious of that is the first part.

SPEAKER_02 43:06

Yeah, yeah, I think that you uniquely sit uh in the space uh within the enterprise where you're able to both see what the contract or the memorialized contract says it has to perform at, and the savings that comes to increase value, right? It's it's one of those. But as you correctly pointed out, right now the savings is attributed to some labor force saving because you you can do it faster, better, cheaper, quicker, smarter, whatever the setup, right? That puts it in a very difficult scenario because on the one hand, you need more of them, but you need different more of them, not the same more of them. So you on the one hand, you gotta kind of either retrain them or retrench them. But uh it's a it's a uh because you have grown at a space, at least sitting in Seattle. It's incredible, right? If you look at the number of people ICERT is now employees versus even say I don't know six years ago, it's just like uh incredible, right? So while you are also constantly juggling this, that's why I find it very interesting the way you phrased it, saying we need more, but not the same more of the same, but but more of more. I guess it's a good good way to uh uh sort of phrase it. Uh and now you guys are spread all over the world. Is the uh talent that you're acquiring you go to where it is available? Because you already were uh you know in multiple geographies for you to build this product, but as it is it making you go to more geographies or are you able to find it? You're in two great geographies, right? You're in Seattle and you're already in India, where the two key hubs are. But other than that, how are you approaching this from a talent perspective?

SPEAKER_03 44:59

Yeah, uh good question. I think I think uh you know, last few years, and one of the outcomes is that it really doesn't matter where you are in the world. Um, apart from when you sleep, when you want to sleep uh at in the night or the day, uh it does it does matter. But we are primarily focused, uh we are split between um you know Bellevue, Seattle, uh uh as one of the hub, and and uh India, uh Bangalore, uh as the other other one. So I think that's primarily where we are we are growing um uh and and adding adding to our teams. Uh but we have hired folks from from where we are we don't have a specific target if that you know to answer your question more clearly. Uh but we are looking within US and within India. That's that's the two primary areas where anywhere within US, anywhere within India. Um that's that's where we are currently looking at. It is it is not that easy to find people who have experience shipping products, Gen AI products. I mean, that's just because you know that there is a small, there is not that big of a pool. So it is it is a lot harder, I would say, to to find find talent. Uh the the the more senior architects and senior engineering talent and product leaders who have that mindset or who have done uh those kind of products yet. So so this is this maybe in another couple years it'll be uh it'll be easier. But it we we we we have uh it's been much harder than before to to hire, I would say, right now.

SPEAKER_02 46:40

Luckily you are aligned with uh Microsoft and you're located in Seattle, so it's gonna it's gotta help you in some fashion. Yeah, you can now look back and say that was strategic, but it you know sort of sort of works. So that that's that's uh the that that helps a lot. And the DNA of the company has been Seattle all along, or greater Seattle, I should say, and Microsoft, of course. So it's it's it helps in attracting the right talent. Uh uh that's that's Chinish, back to you.

SPEAKER_01 47:16

Um so I just had a couple of closing questions regarding the co-pilots that you have. I noticed on your website that you have three different co-pilots. Uh, you know, um uh and and I was wondering why do you have why do you need to have different copilots? Should there be just one co-pilot that does everything and you know you can ask it all kinds of questions?

SPEAKER_03 47:38

Yeah, yeah, yeah. Great, great question. And and and as you know, you will see some of those those changes as well. And like I said, we were one of the first ones to kind of release copilots last year in summer, and and uh just to be very explicit about what they do at that it made sense to have uh co-pilots that are aligned to a specific uh value part. And I think so you'll see the discovered copilot, which helps you discover things within the document. There's an uh in uh interactive insights copilot, which is more interactive, give you insights within the document and summarizing and QA in an interactive fashion. And then there's a risk assessment copilot, which allows you to look at your company policies in this document in context of your company policy, and as soon as you open up the document, it tells you these are the six risks that this document already has based on what you usually look for. So even before you start looking for anything, it tells you this, and then you can click on it, tell you how to mitigate it. So, yes, you're you're right. In like in as we as we move forward, a lot of these different functions will probably be available through the common assistant or common role uh user role-based copilot that is more closer to the to a legal user, a business user may want to look for different things. A legal user looks for different things, a business user looks for different things. A CFO might want to log in and look for a different set of things uh within the same same set of documents, right? So, so yes, directionally, I think that that makes sense. I think the what ones we have right now are very clear to understand, and that helps helped us in being very clear about what they are for and how to use them. And that's why we've seen a lot of good traction then, and we look to adding more uh uh of uh uh these capabilities, I would say, whether whether we call them co-pilots or not, but these different capabilities that allow you to be more productive uh is is kind of directionally where where you look start to see us going.

SPEAKER_01 49:36

Do you also anticipate uh an architecture where uh your co-pilots are interacting with uh co-pilots from other applications? You know, like the example that you gave about a supplier and getting information about the risk profile, right? Um you know that solution that is providing the risk profile information may have its own co-pilot. Right. And so you so your copilot basically asks the question of the other co-pilot and gets a response back. And so do you anticipate that kind of an architecture in the future?

SPEAKER_03 50:10

Absolutely. We absolutely I I think I think like integrating a data layer and then integrating at the intelligence layer, and I think we see that that moving up because each of the systems is you're gonna be able to ask uh the system of a better question. Um, you know, uh, and that that although there isn't a very good interface today on that, I would say, like I hope that that that becomes more of a standard way to interact at the intelligence level. Because yes, you can send a natural language query today, but that's you know, again, working with natural language queries in that way. You know, if you have a big query and you're breaking it down into smaller and saying, okay, I'm gonna ask this part of the query to one intelligence system, the other one, and then come back and then answer. Those orchestration is happening, uh, but I think there isn't standards today, but I do anticipate that we will have to um you know do some of this just from a technology layer in terms of you know, how can I break down my question? Some one orchestrator will help me break down my question, and then each of these providers of intelligent systems will help me answer a high-level question themselves. And they may have further breakdown of that question into their providers and so on. So I I do anticipate that that kind of a world uh is how we would and we are already integrating with plugins, which is a very kind of an initial way to kind of think about this. Like you can have six plugins and you can have a question, and then you can say, Oh, I need to call this three of them, and and the the OpenAI is doing this, Microsoft's doing this. There's there's a lot of uh precedence for uh you know uh orchestration and agents of uh across these different capabilities. But uh to answer the question, I I think the lot needs to be done to make this more seamless, where you can build an intelligent system that also plugs into some other intelligent system in a seamless way. What is your API signature when you talk about intelligence, right? And and I think I I I hope that we there's some standard there that starts to become interesting where you can plug in intelligence together.

SPEAKER_01 52:11

Now, the um uh last question on Gen AI, which is uh um, you know, prompt engineering, as you know, becomes very, very important. Knowing that you can even ask a question like the one that you framed earlier, which is you know, suggest a uh a new SLA based on the risk profile of this of this vendor and how timely they have been in paying us, and and how timely they have been and uh you know uh sending us product and blah blah blah. It's you know you have to even know that you can ask that kind of a question to get the answer. Yeah. So how are you solving that problem?

SPEAKER_03 52:48

Uh I I think prompt prompting uh is a thing. I think I think uh and and this is the thing. Like most of these co-pilots actually will prompt you, uh you as the human to say, hey, ask me this. These are six questions you may want to ask.

SPEAKER_01 53:02

I see.

SPEAKER_03 53:03

Like a lot of these, these uh and and the way we are tackling this is uh understanding the user persona, and then we know that this user persona, when they log in, they are interested in reviewing these things, and so you can pre-configure uh uh frequently asked questions, frequently asked prompts, and you can very in the context kind of show these just like autocomplete of queries and things. You can show, okay, you just asked me for this document. I'm sure you probably next want to ask me about this thing, and so there could be two or three of the uh pre-configured FAQs, frequently asked questions on on for that pro user persona, and so so you are asking it. So, this is more of a from a QA perspective, uh uh what you may anticipate, and then you learn as so you're also training the human to use the system in a in a way because the way you write the prompt or matters, uh uh, and so teaching people how to write prompts automatically through your product is is is also an important part of the UX of the product, uh, so that they get better answers. And then you know you have you have uh you know um uh you know you go from there and then you have uh you know like and dislike buttons on the answers, and then you feed that back. So uh so that's at the user level. Uh uh I think that's what you were asking. Like, how do you how do you even know what to ask? Right, and I think that is that is how the product has to tell you that these are the things I'm able to do, and and you know, because otherwise you won't know what what to ask, you're right. Yeah, okay.

SPEAKER_01 54:37

Okay, one final question is um just uh trying to understand, you know, what is the future of the company? Uh, you know, uh last round you raised at a million dollar plus valuation. Uh the markets have become um you know much tougher uh as we speak. Um, Nasdaq has obviously done quite well in the last six months. Um do you anticipate an IPO happening soon for the company?

SPEAKER_03 55:07

Yeah, I I think it'll be a CEO question. Uh but uh I I think we are we are very much uh you know uh on the right track in terms of from a from a company fundamentals perspective, I feel like we're very strong. And we we are we are um you know also very focused on not taking that for granted. So there's a lot of uh work in terms of how does how how do we keep growing at the rate that we are which we're growing. Uh and then you know all these uh is our you know you'll still have a be a company before or after IPO. So I think from from our perspective in terms of product roadmap and and portfolio and how what we are building and our sales growth and all, I mean we look at that as some point that will happen, but but you know, our our strategies and our our our planning is is is on the fundamentals side, and so that that doesn't change uh what we have to do on a day-to-day basis. So but yeah, I think we're we're doing good.

SPEAKER_01 56:06

I I would say okay, very good, very good. That's that's great to hear. So thank you so much, uh Sudarshan. This was a great uh conversation. Uh it'd be very interesting for listeners to really understand how you are kind of innovating in using Genai in your product line. So we wish you all the best and uh hope to see uh an IPO soon, another big IPO soon in the Seattle area.

SPEAKER_03 56:29

Yeah, absolutely. Thank you so much. Thank you for inviting. It's been a pleasure to to meet you and and and speak with you. Thank you.

SPEAKER_04 56:36

Thank you very much.

SPEAKER_02 56:42

Thank you for listening to our podcast from Startup Exit brought to you by DI Seattle. Assisting in production today are Isha Jain and Mini Verba. Please subscribe to our podcast and rate our podcast wherever you listen to them. Hope you enjoyed it.