Nitin Jayakrishnan of Freehand, chief executive officer, joins theCUBE Research hosts John Furrier and Dave Vellante to discuss applying agentic artificial intelligence, AI, to accounts payable, supply chain finance and enterprise decisioning. The conversation examines how context graphs and observability enable measurable return on investment in high-stakes finance functions.
Jayakrishnan explains Freehand’s approach to automating invoice payments and supply chain workflows. They describe how a context graph connects systems of record, documents, emails and external signals so agents act with accuracy. The discussion addresses domain depth, integration with existing enterprise resource planning, ERP, and the scaling of agentic workflows to distribute expert knowledge across large enterprises.
Key takeaways emphasize explainability, traceability and compliance when deploying agents in finance. Jayakrishnan attributes grounding agents in deterministic and probabilistic data to preventing costly errors and enabling auditable decisions. The hosts stress an ROI-first approach: start with high-impact finance workflows, build observability and Sarbanes-Oxley, SOX, ready controls, and use context graphs to compound operational knowledge into enterprise value.
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Register For theCUBE + NYSE Wired: Mixture of Experts Series
Please fill out the information below. You will recieve an email with a verification link confirming your registration. Click the link to automatically sign into the site.
You’re almost there!
We just sent you a verification email. Please click the verification button in the email. Once your email address is verified, you will have full access to all event content for theCUBE + NYSE Wired: Mixture of Experts Series.
I want my badge and interests to be visible to all attendees.
Checking this box will display your presense on the attendees list, view your profile and allow other attendees to contact you via 1-1 chat. Read the Privacy Policy. At any time, you can choose to disable this preference.
Select your Interests!
add
Upload your photo
Uploading..
OR
Connect via Twitter
Connect via Linkedin
EDIT PASSWORD
Share
Forgot Password
Almost there!
We just sent you a verification email. Please verify your account to gain access to
theCUBE + NYSE Wired: Mixture of Experts Series. If you don’t think you received an email check your
spam folder.
Sign in to theCUBE + NYSE Wired: Mixture of Experts Series.
In order to sign in, enter the email address you used to registered for the event. Once completed, you will receive an email with a verification link. Open this link to automatically sign into the site.
Sign in to gain access to theCUBE + NYSE Wired: Mixture of Experts Series
Please sign in with LinkedIn to continue to theCUBE + NYSE Wired: Mixture of Experts Series. Signing in with LinkedIn ensures a professional environment.
Are you sure you want to remove access rights for this user?
Details
Manage Access
email address
Community Invitation
Nitin Jayakrishnan, Freehand
Nitin Jayakrishnan of Freehand, chief executive officer, joins theCUBE Research hosts John Furrier and Dave Vellante to discuss applying agentic artificial intelligence, AI, to accounts payable, supply chain finance and enterprise decisioning. The conversation examines how context graphs and observability enable measurable return on investment in high-stakes finance functions.
Jayakrishnan explains Freehand’s approach to automating invoice payments and supply chain workflows. They describe how a context graph connects systems of record, documents, emails and external signals so agents act with accuracy. The discussion addresses domain depth, integration with existing enterprise resource planning, ERP, and the scaling of agentic workflows to distribute expert knowledge across large enterprises.
Key takeaways emphasize explainability, traceability and compliance when deploying agents in finance. Jayakrishnan attributes grounding agents in deterministic and probabilistic data to preventing costly errors and enabling auditable decisions. The hosts stress an ROI-first approach: start with high-impact finance workflows, build observability and Sarbanes-Oxley, SOX, ready controls, and use context graphs to compound operational knowledge into enterprise value.
>> Thank you. We want to bring in the leaders who are doing great work on AI and actually bringing in the future as we look at how AI is creating the value, but also how companies are capturing that value, measuring it from supply chains to finance to tokens. Everything's on the table. We're seeing massive growth. We've got a great guest here, Nitin Jayakrishnan, CEO and co -founder of Freehand. Great to see you. Tell us what you're working on. You guys have a unique approach. You're bringing AI intelligence into supply chain and finance, which actually probably will do a lot of great agentic work, hopefully, for folks. Give us a quick overview of what you're working on.
Nitin Jayakrishnan
>> John, always a pleasure. Thanks for having me. at Freehand, what we do is help some of the world's largest businesses, a lot of them listed on the stock exchange in New York, help them essentially think through what this transition from the old world order to what is today shifting very rapidly to a new world order is. I think for the last 30 odd years, three things were true. I think there was a technology paradigm, which was the shift to cloud. Two, I think, supply chains have largely operated in a fairly stable state of manufacturing in China, consumption in mature Western markets, and a fairly stable state supply chain. And then the labor market has also operated in a fairly stable state, which is that knowledge workers were becoming increasingly global, immigration policies held stable, et cetera. All three, I think, are now coming under question. And that's changing the way enterprises globally think about how supply chain and finance are operated. Freehand does one thing and does that thing very, very well, which is effectively we help the world's largest companies pay their invoices on time accurately to their suppliers globally. Except we do that agentically, which means that no large enterprise has to open an invoice ever again. And agents are able to check contracts, ensure compliance, ensure transaction data is in line with what you're being invoiced for and ensure that you don't pay a dollar more or a dollar less than what you should be.
John Furrier
>> You know, Nitin, this is a great, great approach. One of the things I love about this market, besides all the great AI tech, the deep tech is phenomenal. Of course, we love that. We'll get into some of those complexities that are being abstracted away. But the surprise to me has been the rise of the finance vertical. with vibe coding and all the agents and the tooling that's emerging, the finance area, the departments, the workflows are highly impacted. And that's where the money is, right? So when you look at cashflow, you look at compliance, how do I deliver something? These were old IT complex workflows, relegated to old systems, but now AI has brought intelligence to that layer. And there's a lot of domain expertise and then another theme of AI. So when you match domain expertise with kind of this complex IT-like infrastructure, that is basically now running business operations. It's a whole nother ballgame. So share your vision on this because you're hitting a nerve here around, hey, what was once an IT project takes months, maybe years to scope out, can be rolled out. And there's a methodology. Talk about how you're tackling that problem because you guys are hitting that right square in the middle.
Nitin Jayakrishnan
>> Absolutely. AI is impacting almost every part of the enterprise, from front office to back office, from support centers and call centers to finance workflows. And you're absolutely right that this combination of deep domain expertise with intelligence, with AI, is changing the nature of how a lot of this work is happening. I will tell you, though, that finance is not one function. Finance is every function. If you think about the way cash conversion cycles work or the way contracts are structured, the way supplier relationships are managed, the way marketing spend is managed, the way payroll is managed, each of these aspects of the business, managed by very different functions within the business, ultimately rubber hits the road with finance when invoices need to be paid or revenue needs to be recognized on the receivables end. And so what's effectively happening is you're seeing two kinds of AI bets happening. You're seeing horizontal AI bets that are effectively saying, hey, finance is finance for companies of all sizes, for companies across all industries. And within these large companies, for categories of spend and categories of revenue, finance is finance. And so you're seeing AI -native ERP solutions, AI -native finance workflow solutions that are coming up in the market. On the other hand, you're actually seeing AI solutions with domain depth, which are effectively saying, hey, accounts payable for certain categories is different from accounts payable for other categories. The way you treat logistics and raw materials is different from the way you treat contract labor, is different from the way you treat MRO spend or data center spend, et cetera. And effectively is different from how you would look at FP &A workflows or how you would look at some of these other aspects of finance itself. And so what Freehand essentially is taking a bet to say is that domain depth combined with AI's breadth is ultimately what will deliver ROI in the enterprise. Because let's not kid ourselves about the bubble that some of us live in in Silicon Valley. The real world, in New York, in Minnesota, in Chicago, in Atlanta, in the rest of the world, where real large industrial enterprises live and breathe, And indeed, still, the thing that matters more than anything else is dollar ROI of any investment that's made. And when that is realizable is when you combine deep domain expertise with the ability to leverage new software paradigms like AI to deliver value to shareholders and customers.
John Furrier
>> That's a great point. And one of the things I love about the CloudWave as you mentioned earlier, is that that was horizontally scalable. But now with SaaS. Okay, with SaaS, you had the app. When you look at these large-scale companies, whether they're a business that's transitioning, some are more ahead than others, they all have the big systems, ERPs and CRMs. Those are systems of record. But the system of intelligence doesn't have to necessarily replace every type of system of record. So that's going to be something we talk a lot about. In fact, I was talking to the CTO of Workday. They're like, yeah, we have data, but there's other data everywhere. but we're still a system of record, but the system of intelligence is the AI layer. So this new bet that you're making, which I think is right, is that that horizontal AI layer has to touch the data. So you mentioned invoices, compliance. There's a lot of nuance. So context really becomes the key. This is something that you guys are very much leaning into. Knowledge, grasp, whatever you want to call it. Context-aware data. How do you look at that? Because it's not obvious for the folks that were doing the systems of record because they built their own GUIs on their systems and you have to come in and stay with that form put into the database but now you can abstract that away and say I thank you for doing that maybe I have some more data over here another database time series database graph database now you could rethink the interface and then you got agentics doing the work so you have this new trend of okay I need the data. I don't need the UI. I just need the context. Talk about how you look at that, because this is what people are trying to architect, the systems of intelligence layer that will be real-time, domain-specific, context-aware, and accurate.
Nitin Jayakrishnan
>> Absolutely. The paradigm that we're moving into is that software systems effectively were built for the user. AI platforms are the user. And so when you think about agentic technologies like Freehand, we're effectively saying, there used to be a time with a whole bunch of our customers when it was a team of humans in Manila, in India, in different parts of the world that used to log into systems of record and check for data, then use their tribal knowledge of what they knew to be true, which oftentimes was either outdated by a few hours to a few days, sometimes a few years, or oftentimes was siloed information about finance without understanding of procurement or siloed information of logistics without understanding raw material flows, etc. And was then taking decisions and then acting upon those decisions, which in turn had P&L implications of how you paid your suppliers, how you recognize costs and revenue. We are just essentially saying, hey, if an agentic system had access to your system of record data, but also had access to email communications and then also had access to documents, PDF documents, your contracts, your invoices, your bills of lading, customs documents, et cetera, And then also had access to real -time port congestion data, had access to trade lanes and how shipments were being delivered globally, etc. What Freehand is able to do is essentially create a context graph by connecting all of these dots that are lane-specific, that are category -specific, that are company -specific. But some of it, about 25 % of this information is also global, is not specific to any category or specific industry itself. And so by connecting all of this data, we're effectively able to take into account all the information that's required in real time to take a particular decision. And then with certainty, consistency, and accuracy, act upon that decision and then ensure that that action and decision is actually auditable right back to the source information of that one document or that one email that was effectively used to then approve or decide on a particular transaction. And this compounds. The idea of a context graph is that it compounds with every transaction, with every customer, and with the scale that you're operating, that not even the world's largest companies by themselves have enough context to be able to automate at the scale at which a platform like Freehand does.
John Furrier
>> Yeah, I love this conversation because one of the hottest things, thanks to companies like Palantir, who have made the word more modern and mainstream, is ontologies are kicked around. The ontology has been around for a long time. However, when you start thinking about ontologies and relationships, companies are looking to build their corporate brain, as you just highlighted. left hand doesn't know what the right hand is doing. There's no coordination. Talk about the importance of the context graph. Because if you can get it done right, what are the outcomes? What happens next? Because a lot of people are starting to realize that graphs are compatible with how AI thinks. You've got recursion, computer science, it all fits together. And it doesn't mean that you're going to have to get rid of all the other databases. So you can connect in the time series database over there, relational unstructured data, go to a data lake over there. But the graphs aren't mutually exclusive. You guys are thinking about it differently. And talk about how performant it is. What does it all mean? How does this, because you're basically building a brain that knows everything.
Nitin Jayakrishnan
>> Yeah, great question. I think you're right. Ontology has been around for a while. But ontology largely took into consideration data. It didn't take into consideration decisions. Decisions were not considered data or decision traces of the why behind the what was not considered data. And I think the difference between context graphs and ontologies of the past is that context graphs essentially take into consideration the why, not only the what. Effectively, I would sort of categorize a lot of enterprise knowledge into three buckets. Data that is stored in systems of record. Data that is not stored in systems of record, that are either stuck in people's heads, or stuck in PDF documents, in email exchanges, on Slack and Microsoft Teams conversations, etc. And then the why, that almost we did not have either the vocabulary for or the technology to capture till very recently. What context graphs do is it's a super graph, so to speak, of what was a subset, which was called ontology, which takes into consideration all of these three buckets. So let me contextualize this for you in Freehand customer use cases to effectively approve a single invoice. Let's say you are Procter & Gamble, or you're Dunkin' Donuts, or you're Meta or any of the world's largest companies, you have thousands of suppliers globally. You're sourcing all kinds of components, products and services from different countries and suppliers globally. To approve a single invoice of $200, you need to know what was the contract? What did I contract to buy? At what price? At what terms? At what service level? You need to understand the transaction itself. And these are transactions that are happening every day. And so what exactly have I bought? How did I consume it? How was this service or product delivered? Was it actually delivered? Did you meet the service level or not? Et cetera. There is going to be some manager in Lithuania or in the UK who's approved one particular service that had breached the service level according to the contract. as I understand, there was an unforeseen event. And so I'm actually approving this and going to be understanding of it when you actually invoice me. Except this invoice turns up to finance six months later. And then I need to understand the invoice itself. And oftentimes the invoice has got 40 different transactions across six months. And then I need to understand all of these things in the context of that one invoice and then take a simple decision, a seemingly simple decision. Should I approve this number, which is $200? Should I pay only $180? Or actually, should I push back and say, you should be charging me $220, but you only invoiced me for $200 because I'm accruing $220 in my books? That one decision has a whole bunch of decisions that have been taken by the enterprise, by different teams before this. And if you don't understand the why behind the decision, even if you just understand the what, you're not going to be able to take actions against that decision. That is the context.
John Furrier
>> I love that point because what you just talked about, many things, but the one thing that jumped out at me was the undocumented data that's in these workflows. It could be a nurse. It could be someone on the front lines. They learn what the nuances are for data. Subjective information, that's a decision-based thing. So as action becomes the benchmark, not just speeds and feeds, that intelligence is beautiful. Beautiful. So talk about how agentic comes in, because we're seeing agents pick up these deterministic workflows, but taking into account data that could be disaggregated, disparate, siloed, brought together with the graph. So take us through the agentic piece and how that plays out, how you see that evolving, because you can take a non -deterministic thing, someone's knowledge, and make it deterministic to answer the why versus being stuck with the syntax of the app. which is maybe built on the system of record. Again, we're seeing people complain all the time. Yeah, I do things this way in the data entry because the app makes me do that. That's gone. AI is the infrastructure now.
Nitin Jayakrishnan
>> Absolutely. the important piece across all of this is that you ask the world's largest enterprises, they will tell you, hey, we're waiting for a world where our data is clean, our data is perfect, before we can embark on our AI journey. What we're actually seeing is that the enterprises that are leapfrogging and delivering substantial ROI to both shareholders and customers leveraging AI are those that are not waiting for it, but leveraging AI and agentic workflows to solve for bad data. So they're putting sort of the cart ahead of the horse, so to speak, and using the cart to then drive the horse to then subsequently compound this loop. I'll tell you how this works in practice. Today, if you essentially look at the finance workflow of paying invoices, it's a pretty high stakes game. You're essentially talking about, you know, if AI hallucinates when you're asking it a question to one of the LLMs, that's a cute meme on Instagram. great. But when AI hallucinates, when you're paying millions of invoices for a Fortune 500 company, that's a $20 million chargeback, and that's not funny. And so you don't really have an option but to architect these systems for tremendous scale, but also for consistent accuracy and auditability. And so what is important is that the agents are able to actually ground themselves in not only the reality of deterministic data within the enterprise, but also in the reality of non -deterministic probabilistic data. Now, within this business, you're essentially going to see six different records of what the weight of a single iPhone package is. Your procurement system is going to say the weight is X. The ERP is going to say the weight is Y. The bill of lading will say that the weight is Z. Now, how do you determine if you were now an agentic system? How do you determine which weight do I essentially take? But while we may get stuck with this in software terms, does the business stop? There is a Tom or a Jim or a John somewhere in the business who knows what's right and what's wrong. and he uses that information that's in his head to move the ball forward, deliver this product, pay a supplier, and then build a $300 billion behemoth. What agents today do is they're essentially able to take that information from that human in the loop and then ensure that that human does not have to stay in the loop so that knowledge workers can be unshackled from this keep the lights on kind of work and focus on more strategic topics and don't have to spend their time chasing invoices and contracts and would rather spend time growing the business and being accretive to the P&L.
John Furrier
>> You know, that's awesome. One thing to validate what you just said is that there's two areas that I'm seeing this following statement play out. Robotics and agents in the enterprise. And that is, I can't be wrong. So robotics, if you're a self -driving autonomous car or you're in manufacturing, they can't get the wrong answer because safety is huge. They don't want people to get killed. There's issues there, right? You can't be wrong. And the other one on agents is workflows. They have determined they can't be wrong. The money's on the line, as you pointed out in that example. You can quantify it. There is now quantifiable evidence on the outcomes that you can't get it wrong. You can't have a bad answer because they can actually quantify it. So that's consistent. The other thing you brought up I think is important to talk through is that a lot of the back end has to have that kind of cloud -native observability, traceability. That's complex computer science platform issues because you have to have explainability. If you're going to do the AI, you better prove why that happened. So talk about that dynamic of the explainability, the tracing. There's a lot of content talking about agents. All great, but not a lot of people are talking about, hey, you actually kind of have to close the loop and prove that it actually worked.
John Furrier
>> Absolutely.
John Furrier
>> There's data involved. this is a concept we've seen before.
Nitin Jayakrishnan
>> Yeah, absolutely. You're right. On one hand, AI observability, AI governance is a computer science topic. And there's a lot of great work happening both in the academic labs, but also in industry to push the ball forward on how we derive observability from or borrow observability from the cloud era, and then drive forward observability and compliance and governance in the AI era. This includes traceability back into the source of information, explainability in terms of ensuring that every individual decision that's being taken, and usually we say approve an invoice or take a decision on an invoice. That's not one decision. It's a swarm of agents that are essentially taking hundreds of decisions along the way. And so you want to be able to explain all of those individual decisions and then composite, in a composite fashion, bubble that up to the business user in a way that they're able to understand, etc. So, there is the computer science topic of this, but I don't want to underplay the compliance angle of this. Because the reality of this is that there is an entire body of work that's now going into SOX compliance for agents, insurance, E&O insurance for agents. Because today, if it is not humans that are taking decisions and agents that are taking decisions, how do you ensure that E&O covers agentic workflows as well? How do you ensure that SOX compliance from an internal audit standpoint by a Big Four auditor also covers agentic audits as well, etc.? So there is an increasing body of work as increasingly in call centers, in contact centers, in back office operations, in shared services centers globally. As agents increasingly are taking more and more of these decisions, I think the compliance angle of it, which isn't a computer science topic, is becoming more important.
John Furrier
>> It's a huge point because agents are touching resources, storage, network, memory, kind of like microservices were during the cloud wave. Great point. And that has to be table stakes. Nitin, so I want to get your thoughts on what you guys are doing now. What's next for you? Obviously, you've got your finger on the pulse. You're in a great area. What's on your agenda? What's your focus?
Nitin Jayakrishnan
>> Absolutely. Today, we announced a $75 million fundraise led by Battery Ventures, NewRoad Ventures, NewBuild Venture Capital, Nexus Venture Partners, and a great host of investors. A lot of that capital is coming on the back of us, coming out of stealth four months back. We've been building Freehand now for two and a half years, and we've had stellar growth. In the two by two of the world's largest enterprises on one extreme and the most high stakes workflows on the other, we decided to go after the top right, which is finance workflows in the Fortune 500. And we've seen phenomenal reception. It seems to be a pain point that has been pervasive. A lot of enterprises were almost held hostage to these BPO contracts and outsourcing contracts where they've been in these contracts for 20, 30 years. They don't have a better alternative. They're struggling with their current relationships. and they couldn't bring this in-house because there isn't organizational capacity or bandwidth to be able to take on these workflows anymore. And now Freehand provides them with a third alternative. Don't bring it in-house. Don't keep it outsourced. Hire an AI team that can essentially run this process at scale. And so for invoice payments for some of the world's most complex categories, if you are a business that is a billion dollars in revenue or above, Freehand will take over your supply chain finance or accounts payable processes and automate it with AI at scale. But the real value of this to your earlier point is that you're building this organization's brain for its supply chain. The context graph, the supply chain context graph that we're building is essentially the brain of their supply chain. And we're leveraging that brain to take one decision, which is should I pay my supplier or not? And eventually, we will leverage the same brain to take hundreds of decisions across their supply chain. Who should I buy from? At what price should I buy? Which suppliers do I engage with? How do I leverage my suppliers to deliver customer value? And eventually, how to run my supply chain at scale?
John Furrier
>> Well, Nitin, it's a great bet. I love the horizontal data layers, infrastructure, and you got the domain expertise. The graph is the backbone of the brain. You can connect to other systems. Everything kind of clicks in place. That's systems of intelligence, you're bringing intelligence into the organization. very focused execution. Congratulations on the funding. Thanks for coming on our Mixture of Experts series. You're now in our AI brain. Thanks for coming on theCUBE here at the NYSE Wired.
Nitin Jayakrishnan
>> Yeah, I appreciate it, John.
John Furrier
>> Always a pleasure having you. Thanks for coming on. I'm John Furrier, the host of theCUBE here at the NYSE Wired program. Of course, we're here in New York connecting Silicon Valley from our Palo Alto studio to New York. Thanks for watching. Thank you.
>> Thank you. We want to bring in the leaders who are doing great work on AI and actually bringing in the future as we look at how AI is creating the value, but also how companies are capturing that value, measuring it from supply chains to finance to tokens. Everything's on the table. We're seeing massive growth. We've got a great guest here, Nitin Jayakrishnan, CEO and co -founder of Freehand. Great to see you. Tell us what you're working on. You guys have a unique approach. You're bringing AI intelligence into supply chain and finance, which actually probably will do a lot of great agentic work, hopefully, for folks. Give us a quick overview of what you're working on.
Nitin Jayakrishnan
>> John, always a pleasure. Thanks for having me. at Freehand, what we do is help some of the world's largest businesses, a lot of them listed on the stock exchange in New York, help them essentially think through what this transition from the old world order to what is today shifting very rapidly to a new world order is. I think for the last 30 odd years, three things were true. I think there was a technology paradigm, which was the shift to cloud. Two, I think, supply chains have largely operated in a fairly stable state of manufacturing in China, consumption in mature Western markets, and a fairly stable state supply chain. And then the labor market has also operated in a fairly stable state, which is that knowledge workers were becoming increasingly global, immigration policies held stable, et cetera. All three, I think, are now coming under question. And that's changing the way enterprises globally think about how supply chain and finance are operated. Freehand does one thing and does that thing very, very well, which is effectively we help the world's largest companies pay their invoices on time accurately to their suppliers globally. Except we do that agentically, which means that no large enterprise has to open an invoice ever again. And agents are able to check contracts, ensure compliance, ensure transaction data is in line with what you're being invoiced for and ensure that you don't pay a dollar more or a dollar less than what you should be.
John Furrier
>> You know, Nitin, this is a great, great approach. One of the things I love about this market, besides all the great AI tech, the deep tech is phenomenal. Of course, we love that. We'll get into some of those complexities that are being abstracted away. But the surprise to me has been the rise of the finance vertical. with vibe coding and all the agents and the tooling that's emerging, the finance area, the departments, the workflows are highly impacted. And that's where the money is, right? So when you look at cashflow, you look at compliance, how do I deliver something? These were old IT complex workflows, relegated to old systems, but now AI has brought intelligence to that layer. And there's a lot of domain expertise and then another theme of AI. So when you match domain expertise with kind of this complex IT-like infrastructure, that is basically now running business operations. It's a whole nother ballgame. So share your vision on this because you're hitting a nerve here around, hey, what was once an IT project takes months, maybe years to scope out, can be rolled out. And there's a methodology. Talk about how you're tackling that problem because you guys are hitting that right square in the middle.
Nitin Jayakrishnan
>> Absolutely. AI is impacting almost every part of the enterprise, from front office to back office, from support centers and call centers to finance workflows. And you're absolutely right that this combination of deep domain expertise with intelligence, with AI, is changing the nature of how a lot of this work is happening. I will tell you, though, that finance is not one function. Finance is every function. If you think about the way cash conversion cycles work or the way contracts are structured, the way supplier relationships are managed, the way marketing spend is managed, the way payroll is managed, each of these aspects of the business, managed by very different functions within the business, ultimately rubber hits the road with finance when invoices need to be paid or revenue needs to be recognized on the receivables end. And so what's effectively happening is you're seeing two kinds of AI bets happening. You're seeing horizontal AI bets that are effectively saying, hey, finance is finance for companies of all sizes, for companies across all industries. And within these large companies, for categories of spend and categories of revenue, finance is finance. And so you're seeing AI -native ERP solutions, AI -native finance workflow solutions that are coming up in the market. On the other hand, you're actually seeing AI solutions with domain depth, which are effectively saying, hey, accounts payable for certain categories is different from accounts payable for other categories. The way you treat logistics and raw materials is different from the way you treat contract labor, is different from the way you treat MRO spend or data center spend, et cetera. And effectively is different from how you would look at FP &A workflows or how you would look at some of these other aspects of finance itself. And so what Freehand essentially is taking a bet to say is that domain depth combined with AI's breadth is ultimately what will deliver ROI in the enterprise. Because let's not kid ourselves about the bubble that some of us live in in Silicon Valley. The real world, in New York, in Minnesota, in Chicago, in Atlanta, in the rest of the world, where real large industrial enterprises live and breathe, And indeed, still, the thing that matters more than anything else is dollar ROI of any investment that's made. And when that is realizable is when you combine deep domain expertise with the ability to leverage new software paradigms like AI to deliver value to shareholders and customers.
John Furrier
>> That's a great point. And one of the things I love about the CloudWave as you mentioned earlier, is that that was horizontally scalable. But now with SaaS. Okay, with SaaS, you had the app. When you look at these large-scale companies, whether they're a business that's transitioning, some are more ahead than others, they all have the big systems, ERPs and CRMs. Those are systems of record. But the system of intelligence doesn't have to necessarily replace every type of system of record. So that's going to be something we talk a lot about. In fact, I was talking to the CTO of Workday. They're like, yeah, we have data, but there's other data everywhere. but we're still a system of record, but the system of intelligence is the AI layer. So this new bet that you're making, which I think is right, is that that horizontal AI layer has to touch the data. So you mentioned invoices, compliance. There's a lot of nuance. So context really becomes the key. This is something that you guys are very much leaning into. Knowledge, grasp, whatever you want to call it. Context-aware data. How do you look at that? Because it's not obvious for the folks that were doing the systems of record because they built their own GUIs on their systems and you have to come in and stay with that form put into the database but now you can abstract that away and say I thank you for doing that maybe I have some more data over here another database time series database graph database now you could rethink the interface and then you got agentics doing the work so you have this new trend of okay I need the data. I don't need the UI. I just need the context. Talk about how you look at that, because this is what people are trying to architect, the systems of intelligence layer that will be real-time, domain-specific, context-aware, and accurate.
Nitin Jayakrishnan
>> Absolutely. The paradigm that we're moving into is that software systems effectively were built for the user. AI platforms are the user. And so when you think about agentic technologies like Freehand, we're effectively saying, there used to be a time with a whole bunch of our customers when it was a team of humans in Manila, in India, in different parts of the world that used to log into systems of record and check for data, then use their tribal knowledge of what they knew to be true, which oftentimes was either outdated by a few hours to a few days, sometimes a few years, or oftentimes was siloed information about finance without understanding of procurement or siloed information of logistics without understanding raw material flows, etc. And was then taking decisions and then acting upon those decisions, which in turn had P&L implications of how you paid your suppliers, how you recognize costs and revenue. We are just essentially saying, hey, if an agentic system had access to your system of record data, but also had access to email communications and then also had access to documents, PDF documents, your contracts, your invoices, your bills of lading, customs documents, et cetera, And then also had access to real -time port congestion data, had access to trade lanes and how shipments were being delivered globally, etc. What Freehand is able to do is essentially create a context graph by connecting all of these dots that are lane-specific, that are category -specific, that are company -specific. But some of it, about 25 % of this information is also global, is not specific to any category or specific industry itself. And so by connecting all of this data, we're effectively able to take into account all the information that's required in real time to take a particular decision. And then with certainty, consistency, and accuracy, act upon that decision and then ensure that that action and decision is actually auditable right back to the source information of that one document or that one email that was effectively used to then approve or decide on a particular transaction. And this compounds. The idea of a context graph is that it compounds with every transaction, with every customer, and with the scale that you're operating, that not even the world's largest companies by themselves have enough context to be able to automate at the scale at which a platform like Freehand does.
John Furrier
>> Yeah, I love this conversation because one of the hottest things, thanks to companies like Palantir, who have made the word more modern and mainstream, is ontologies are kicked around. The ontology has been around for a long time. However, when you start thinking about ontologies and relationships, companies are looking to build their corporate brain, as you just highlighted. left hand doesn't know what the right hand is doing. There's no coordination. Talk about the importance of the context graph. Because if you can get it done right, what are the outcomes? What happens next? Because a lot of people are starting to realize that graphs are compatible with how AI thinks. You've got recursion, computer science, it all fits together. And it doesn't mean that you're going to have to get rid of all the other databases. So you can connect in the time series database over there, relational unstructured data, go to a data lake over there. But the graphs aren't mutually exclusive. You guys are thinking about it differently. And talk about how performant it is. What does it all mean? How does this, because you're basically building a brain that knows everything.
Nitin Jayakrishnan
>> Yeah, great question. I think you're right. Ontology has been around for a while. But ontology largely took into consideration data. It didn't take into consideration decisions. Decisions were not considered data or decision traces of the why behind the what was not considered data. And I think the difference between context graphs and ontologies of the past is that context graphs essentially take into consideration the why, not only the what. Effectively, I would sort of categorize a lot of enterprise knowledge into three buckets. Data that is stored in systems of record. Data that is not stored in systems of record, that are either stuck in people's heads, or stuck in PDF documents, in email exchanges, on Slack and Microsoft Teams conversations, etc. And then the why, that almost we did not have either the vocabulary for or the technology to capture till very recently. What context graphs do is it's a super graph, so to speak, of what was a subset, which was called ontology, which takes into consideration all of these three buckets. So let me contextualize this for you in Freehand customer use cases to effectively approve a single invoice. Let's say you are Procter & Gamble, or you're Dunkin' Donuts, or you're Meta or any of the world's largest companies, you have thousands of suppliers globally. You're sourcing all kinds of components, products and services from different countries and suppliers globally. To approve a single invoice of $200, you need to know what was the contract? What did I contract to buy? At what price? At what terms? At what service level? You need to understand the transaction itself. And these are transactions that are happening every day. And so what exactly have I bought? How did I consume it? How was this service or product delivered? Was it actually delivered? Did you meet the service level or not? Et cetera. There is going to be some manager in Lithuania or in the UK who's approved one particular service that had breached the service level according to the contract. as I understand, there was an unforeseen event. And so I'm actually approving this and going to be understanding of it when you actually invoice me. Except this invoice turns up to finance six months later. And then I need to understand the invoice itself. And oftentimes the invoice has got 40 different transactions across six months. And then I need to understand all of these things in the context of that one invoice and then take a simple decision, a seemingly simple decision. Should I approve this number, which is $200? Should I pay only $180? Or actually, should I push back and say, you should be charging me $220, but you only invoiced me for $200 because I'm accruing $220 in my books? That one decision has a whole bunch of decisions that have been taken by the enterprise, by different teams before this. And if you don't understand the why behind the decision, even if you just understand the what, you're not going to be able to take actions against that decision. That is the context.
John Furrier
>> I love that point because what you just talked about, many things, but the one thing that jumped out at me was the undocumented data that's in these workflows. It could be a nurse. It could be someone on the front lines. They learn what the nuances are for data. Subjective information, that's a decision-based thing. So as action becomes the benchmark, not just speeds and feeds, that intelligence is beautiful. Beautiful. So talk about how agentic comes in, because we're seeing agents pick up these deterministic workflows, but taking into account data that could be disaggregated, disparate, siloed, brought together with the graph. So take us through the agentic piece and how that plays out, how you see that evolving, because you can take a non -deterministic thing, someone's knowledge, and make it deterministic to answer the why versus being stuck with the syntax of the app. which is maybe built on the system of record. Again, we're seeing people complain all the time. Yeah, I do things this way in the data entry because the app makes me do that. That's gone. AI is the infrastructure now.
Nitin Jayakrishnan
>> Absolutely. the important piece across all of this is that you ask the world's largest enterprises, they will tell you, hey, we're waiting for a world where our data is clean, our data is perfect, before we can embark on our AI journey. What we're actually seeing is that the enterprises that are leapfrogging and delivering substantial ROI to both shareholders and customers leveraging AI are those that are not waiting for it, but leveraging AI and agentic workflows to solve for bad data. So they're putting sort of the cart ahead of the horse, so to speak, and using the cart to then drive the horse to then subsequently compound this loop. I'll tell you how this works in practice. Today, if you essentially look at the finance workflow of paying invoices, it's a pretty high stakes game. You're essentially talking about, you know, if AI hallucinates when you're asking it a question to one of the LLMs, that's a cute meme on Instagram. great. But when AI hallucinates, when you're paying millions of invoices for a Fortune 500 company, that's a $20 million chargeback, and that's not funny. And so you don't really have an option but to architect these systems for tremendous scale, but also for consistent accuracy and auditability. And so what is important is that the agents are able to actually ground themselves in not only the reality of deterministic data within the enterprise, but also in the reality of non -deterministic probabilistic data. Now, within this business, you're essentially going to see six different records of what the weight of a single iPhone package is. Your procurement system is going to say the weight is X. The ERP is going to say the weight is Y. The bill of lading will say that the weight is Z. Now, how do you determine if you were now an agentic system? How do you determine which weight do I essentially take? But while we may get stuck with this in software terms, does the business stop? There is a Tom or a Jim or a John somewhere in the business who knows what's right and what's wrong. and he uses that information that's in his head to move the ball forward, deliver this product, pay a supplier, and then build a $300 billion behemoth. What agents today do is they're essentially able to take that information from that human in the loop and then ensure that that human does not have to stay in the loop so that knowledge workers can be unshackled from this keep the lights on kind of work and focus on more strategic topics and don't have to spend their time chasing invoices and contracts and would rather spend time growing the business and being accretive to the P&L.
John Furrier
>> You know, that's awesome. One thing to validate what you just said is that there's two areas that I'm seeing this following statement play out. Robotics and agents in the enterprise. And that is, I can't be wrong. So robotics, if you're a self -driving autonomous car or you're in manufacturing, they can't get the wrong answer because safety is huge. They don't want people to get killed. There's issues there, right? You can't be wrong. And the other one on agents is workflows. They have determined they can't be wrong. The money's on the line, as you pointed out in that example. You can quantify it. There is now quantifiable evidence on the outcomes that you can't get it wrong. You can't have a bad answer because they can actually quantify it. So that's consistent. The other thing you brought up I think is important to talk through is that a lot of the back end has to have that kind of cloud -native observability, traceability. That's complex computer science platform issues because you have to have explainability. If you're going to do the AI, you better prove why that happened. So talk about that dynamic of the explainability, the tracing. There's a lot of content talking about agents. All great, but not a lot of people are talking about, hey, you actually kind of have to close the loop and prove that it actually worked.
John Furrier
>> Absolutely.
John Furrier
>> There's data involved. this is a concept we've seen before.
Nitin Jayakrishnan
>> Yeah, absolutely. You're right. On one hand, AI observability, AI governance is a computer science topic. And there's a lot of great work happening both in the academic labs, but also in industry to push the ball forward on how we derive observability from or borrow observability from the cloud era, and then drive forward observability and compliance and governance in the AI era. This includes traceability back into the source of information, explainability in terms of ensuring that every individual decision that's being taken, and usually we say approve an invoice or take a decision on an invoice. That's not one decision. It's a swarm of agents that are essentially taking hundreds of decisions along the way. And so you want to be able to explain all of those individual decisions and then composite, in a composite fashion, bubble that up to the business user in a way that they're able to understand, etc. So, there is the computer science topic of this, but I don't want to underplay the compliance angle of this. Because the reality of this is that there is an entire body of work that's now going into SOX compliance for agents, insurance, E&O insurance for agents. Because today, if it is not humans that are taking decisions and agents that are taking decisions, how do you ensure that E&O covers agentic workflows as well? How do you ensure that SOX compliance from an internal audit standpoint by a Big Four auditor also covers agentic audits as well, etc.? So there is an increasing body of work as increasingly in call centers, in contact centers, in back office operations, in shared services centers globally. As agents increasingly are taking more and more of these decisions, I think the compliance angle of it, which isn't a computer science topic, is becoming more important.
John Furrier
>> It's a huge point because agents are touching resources, storage, network, memory, kind of like microservices were during the cloud wave. Great point. And that has to be table stakes. Nitin, so I want to get your thoughts on what you guys are doing now. What's next for you? Obviously, you've got your finger on the pulse. You're in a great area. What's on your agenda? What's your focus?
Nitin Jayakrishnan
>> Absolutely. Today, we announced a $75 million fundraise led by Battery Ventures, NewRoad Ventures, NewBuild Venture Capital, Nexus Venture Partners, and a great host of investors. A lot of that capital is coming on the back of us, coming out of stealth four months back. We've been building Freehand now for two and a half years, and we've had stellar growth. In the two by two of the world's largest enterprises on one extreme and the most high stakes workflows on the other, we decided to go after the top right, which is finance workflows in the Fortune 500. And we've seen phenomenal reception. It seems to be a pain point that has been pervasive. A lot of enterprises were almost held hostage to these BPO contracts and outsourcing contracts where they've been in these contracts for 20, 30 years. They don't have a better alternative. They're struggling with their current relationships. and they couldn't bring this in-house because there isn't organizational capacity or bandwidth to be able to take on these workflows anymore. And now Freehand provides them with a third alternative. Don't bring it in-house. Don't keep it outsourced. Hire an AI team that can essentially run this process at scale. And so for invoice payments for some of the world's most complex categories, if you are a business that is a billion dollars in revenue or above, Freehand will take over your supply chain finance or accounts payable processes and automate it with AI at scale. But the real value of this to your earlier point is that you're building this organization's brain for its supply chain. The context graph, the supply chain context graph that we're building is essentially the brain of their supply chain. And we're leveraging that brain to take one decision, which is should I pay my supplier or not? And eventually, we will leverage the same brain to take hundreds of decisions across their supply chain. Who should I buy from? At what price should I buy? Which suppliers do I engage with? How do I leverage my suppliers to deliver customer value? And eventually, how to run my supply chain at scale?
John Furrier
>> Well, Nitin, it's a great bet. I love the horizontal data layers, infrastructure, and you got the domain expertise. The graph is the backbone of the brain. You can connect to other systems. Everything kind of clicks in place. That's systems of intelligence, you're bringing intelligence into the organization. very focused execution. Congratulations on the funding. Thanks for coming on our Mixture of Experts series. You're now in our AI brain. Thanks for coming on theCUBE here at the NYSE Wired.
Nitin Jayakrishnan
>> Yeah, I appreciate it, John.
John Furrier
>> Always a pleasure having you. Thanks for coming on. I'm John Furrier, the host of theCUBE here at the NYSE Wired program. Of course, we're here in New York connecting Silicon Valley from our Palo Alto studio to New York. Thanks for watching. Thank you.
John Furrier