Sashi Narahari: [0:09]
Alright, let’s get into some serious business. I’m sure there’ll be time for pranks tonight. So my topic for the presentation today is the future of b2b software. I’m going to start with a bold prediction that we are betting on b2b cloud software will be dead by 2030. It kind of feels weird. We say that because that’s the business we are in. And then the answer first is the next generation of enterprise software will be autonomous software. Now, if you just think about it, right, doing predictions is kind of easy, because you can just go put it on a LinkedIn and nobody will remember you after like, 10 years. So you can say whatever you want. But being a tech company CEO, I’m going to be a bit vulnerable because our business has evolved to it’s one of the most disruptive category that somebody else can eat your lunch. I mean, if you look at companies like Tesla, which is disrupting automobile industry, very capital intensive business in a decade, Airbnb is disrupting the vacation rentals, hotel business in a decade. Our business is software, and all you need is developers to build code. So it is probably the most vulnerable business and makes me feel the most insecure as a CEO of the company. So for that reason, our roles have evolved to you have to predict where the puck is going, and you have to make some big bets. So the autonomous software bet we are making is we are literally talking about hundreds of millions of dollars, and either you eat somebody else’s lunch, or they will eat you in just three to five years. So let me double click a bit on how we came about making this decision and this big bet. So the best way to think about features always looks at history. And history gives you a lot of clues of what’s to come. So I’m going to rewind a bit to talk about the evolution of software. So if you go to the 1950s, we think of this as the age of paper typewriters and filing cabinets, right, this is probably how many of our parents and grandparents did business, you print a purchase order, print a sales order, and then you stack it in a file cabinet as a record. And then the next stage of that was the 1980s, which was the IBM mainframes. Many of your companies probably build software paid for it, it was the green screen terminals, it’s the same stuff. Whatever was paper became digital. And then the 2000s is where I started my career as an SAP developer is age of SAP, the biggest value proposition was instead of each of you needing to build your own software and incur millions of dollars of costs, you could just go license, a packet software, and then use it versus building. And then 2010s actually belong to Salesforce, we all know that the whole cloud revolution over the last decade is the age of cloud software, the value proposition was, instead of you buying servers, hardware operating system, and army of people to maintain and manage all of that is hosted on the cloud, and you get the software delivered over the cloud. But let’s take a simple example of a purchase order a business transaction that you create with a vendor. If you look at the evolution, we went from paper to mainframe to SAP on-prem, and let’s say somebody like Arriba, or a cooper or others, at a macro level, it is still the same information. We went from huge shift going from paper and filing cabinets to electronic systems of record. But the core business process of is still creating read updates delete of information records. When I started my career at SAP, I was kind of surprised that every transaction was v 01. Create sales order V zero to change sales order via 03 display sales order. And almost every business transaction was like that. It’s still a massive shift, because we went from paper to information, digital information records, but it is still the same create read update delete software. Now, let’s kind of summarise right, we went from paper mainframes on-premise cloud. And you’ll also notice that the disruption cycles are shrinking. What took 30 years went to 20 years and 10 years. So you can easily now extrapolate and say, maybe the next disruption cycle is seven years because of the shrinking but what next? What is the next after cloud software? I’m going to shift gears to a parallel technology economy that we are all part of. Maybe that will give us some clues. So the world of b2c applications, all those apps that you download and use from iPhone.
Sashi Narahari: [4:48]
So these are the top b2c applications, Facebook, Google Maps, LinkedIn, Pinterest, Amazon, Netflix, and Uber. These are all part of our everyday lives. They were not part of our everyday lives and even a decade ago, we take it for granted today, let’s take a simple application like Google Maps, which probably many of you used coming into the conference this morning. But if you think about what is happening under the hood, it is a data-driven intelligent navigation system. It is taking into account the time of the day, the real-time traffic data, the history of what happened at that time in that location. And then it’s dynamically giving you route options to minimize your time to get to a location. But as a user, you don’t know all that stuff. But it’s basically data-driven intelligent navigation. Let’s switch gears and look at Amazon a little bit more advanced b2c application from an end user standpoint, and let’s go through the journey. A month ago, I was trying to buy a coffee maker, I wanted to get a coffee maker. So I started going back to the office, you know, what Stella did at gunpoint. So I was like, Okay, I gotta show up at the office. So I need some coffee to drive to work. So I started typing coffeemaker, and you’ll see the suggestions below, I noticed this thing at the bottom called single serve coffee maker. That’s interesting. I just needed coffee for myself. Let me check that out. Now, this is something that I did not think of, I only thought of coffee maker. But some the intelligence of the application suggested stuff for me. And then I looked at the results. You look at the second result here, the elite gourmet coffee maker, I click on that looks interesting, has 2600 reviews, reasonable price, I could buy it. But before I do that Amazon is also now giving me a bunch of suggestions to make the decision. Right back in the days, the suggestions came from your friends at the office, you should buy this, there was just three or five friends telling you what to do. Now the world at large, and the democratic power of the reviews is actually telling you through the Amazon algorithms, okay, so I find this option, which is a Black and Decker single serve coffee maker that has actually almost four times the number of reviews, good price looks sleeker, I’m ready to buy. And then before I buy, Amazon is giving me another option. What else can I buy? It’s kind of shocking that I usually have a breakfast blend that is stronger coffee, because I need to wake up in the morning, after a late night. And then an afternoon blend, it’s lighter. It’s just made it easy for me, and boom, I go, I make the purchase, you can do all of this in less than 10 minutes today. Right? You can make your entire decision of most of the categories of products in less than 10 minutes. And that is because, under the hood, there is the NLP algorithms, which Amazon calls them a 10 year of clustering algorithm, based on what you’re looking at. It’s making alternative suggestions. And there is another set of algorithms called collaborative filtering, which is also suggesting what else can you buy with that. But the unique thing about this is it is just made it very easy for you as an end user. Alright, so what is the common theme here? If you look at them, the lowest column here, this is data-driven, intelligent b2c applications for specific consumer needs, Facebook, personal data, Google traffic data, LinkedIn, professional data, and so forth. But if you now look at the explosion of all the b2c applications, they all go deep on one specific consumer need. It is really related to a set of data specific to that consumer need, and the democratic power of that millions or billions of records. And how does it bring it together to make your experience easy? Now, switching back to VDB, software applications, whether you are a user in some remote part of the world, literally in a village, let’s say, to sitting in the corporate office in New York, what you see on the left is the end user experience that they’re going through today, they all have mobile phones, smartphones, they’re all using this application on a daily basis. But when you go to work, you’re still doing the create, read, update, delete, in a grey screen in SAP. So this is two parallel economies, exact same users, they use already. It may be there could be some change management issues. But then does that make sense? And that is why we think in less than seven years or probably seven years, every b2b software category, we already cash procure to pay, e-commerce will be disrupted by data-driven intelligent applications. We are calling them autonomous software, maybe somebody else would give it another name. So what is the bet that hybrid is making? We are going we are investing heavily on building domain-specific autonomous software and we are calling it autonomous finance. That’s the category we are slowly shifting to where the autonomous finance is the data-driven finance applications. Of course, we have made a lot of progress in autonomous receivables for already cash. We have over 700 customers. We launched autonomous Treasury two and a half years back we have about 50 clients It’s an autonomous accounting is one of our new favorites, which we launched last year.
Sashi Narahari: [10:06]
So what are the building blocks of an autonomous finance application? The core foundation? Is the domain-specific big data. If you think about the inspiration from BDC, the layer after that is the intelligence layer, how do you work on the data through a massive set of data with a set of algorithms, and then deliver a unique experience to users, which I think it’s a long way to go. Right. Most of us as b2b software companies, we have long ways to go when it comes to delivering an experience. And we’re working on it. Now, this is just a technical slide further double click.
Sashi Narahari: [10:43]
Today, we have about 4.7 trillion worth of data to operate on. And the algorithms itself intelligence could range from simple heuristics to RPA for certain automation of steps, do AI and natural language processing, which is more like what you saw in the Amazon free text and suggestions based on that. And then the experience layer will always have the traditional enterprise Web UI. Think of it as the web-based UI, most of you probably use it hierarchies as well, it needs to be mobile UI as well. You heard of the Frieda digital assistant voice and touch interaction. And we are working on another IP, which we launched in our treasury applications called ConnectED workspaces. What we learned was a lot of users switch from the enterprise applications to excel, they’ll export the data, think of it as the SAP FBI, five in transaction and do a lot of work. What If analysis, modeling, and go back. So with the connected workspaces is a punch out from the application. So you never have to go to Excel. It’s been a great experiment for us in Treasury. And now we’re going to bring that into the other solution areas as well. So to summarise, we have today about 13 Plus AI use cases for audit cash, you don’t have to understand the details. But if you’re interested, you can work from credit to collections, and we are continuing to invest more and more this is fairly complex, and advanced r&d stuff that we are doing. I’m going to double click on few of the products so that you can get a feel for what does it mean from an end user? Or a manager’s standpoint? What does this mean? So in cash application, we went from in 2016, and OCR template based capture two, I think the last time we did the analytics, our cross-client, item level hit rate is 91. Person. All this is because the machine learning algorithms are getting more data more context sensitivity, and how do they get better with time, you can kind of do the parallel economy to the b2c, your shopping experience with Amazon, seven years back to now it’s becoming more and more accurate. But of course, never be 100%. Right? So how do we deliver that? Foundation is big data. You have remittance information, payment, open and close information, emails, everything related to cash application and payment. And how do you operate on that using intelligence algorithms from Remington’s prediction to auto-suggesting reference numbers? So that the exception management experience for the end users is mod, which I think as I said, a lot more work to do, compared to where we need to be. Let’s look at collections. In the collections, the core layer of big data is hundreds or 1000s of customer data points payment history, dispute history, purchase-to-pay history, and how do we leverage that the most important algorithm is predicting the payment date of every invoice. And then also segmenting the customers correctly. Because if you think about most collections, you take your one-time slice of data and you say, Okay, those are my good paying customers bad paying customers high-risk customers. But that’s just a static decision at any given point of time. How do you change that to dynamic segmentation, where you don’t define what the three categories are, but the system proposes, and then you also then drive the collection behavior of what’s your prioritize, work list, whom to contact on a given day, if you have a team of 10 collectors, and then also recommending, how to contact. Now the key here is that it’s the recommendation engine, a user can always override when I was doing the Amazon shopping, I don’t have to exactly do what the system is telling me. But it’s helping me accelerate my decision. So this is one example of the user interface, which is the enterprise Web UI. Mostly the modern version of traditional enterprise software should get better. But we are also experimenting with what we call as the Frieda UI, which is touch-enabled and voice enabled. And then I think we have a lot of clients who are live but I think the change management will take time, everybody will be ready at their own time. So you have both the choices to switch back and forth. And then to summarise, I mean going back to the why we are making this big bet we are continuing to make this bet otherwise we think will be irrelevant. We have two patents. We have nine patents pending we’ve done in lab. So we kind of going on an overdrive of r&d investment. Because we think autonomous software is the future. Switching gears to a different solution suite is autonomous treasury. In the autonomous treasury, if you look at the current state of the world, on the right, what you’re seeing is two options, you log into the bank websites, look at your cash position, information at any given point of time typically starts in the morning. And then you can wrap up the lesson by 10 am. And then you can capture it in Excel. Or if you have a TCMS system, it has the bank interfaces with API, but then big picture, it is still a create, read, update, delete of cash flow records, is just telling you the fact at any given point of time, how much money do you have as a cash position, and then you’re making the borrowing and investment decisions? I think the opportunity in Treasury is how can you transform the Treasury function? Again, this is long-term thinking, from a back office transaction monitoring by using the bank’s cash management software, too many of you probably have CMS systems, which is automating the reporting and cash positions to a role where you could be a low-risk daily castrator. Another daily trader is not a great term because it started off as gambling money. But that’s why it’s low risk. The concept is the same on a daily basis on a real-time basis. Can you make investments and borrowing decisions for your company? Based on the policy of the company, some companies might have a very low-risk tolerance to some might have a little bit more. But that’s exactly what hedge funds do or day traders? Do. They have different viewpoints of how to trade stocks. But to be able to do that you need the data-driven intelligence software. So you have the context. So what are those layers look like? The core foundation layer of Big Data is the historical inflows and outflows across all of your cash flow categories, AR AP payroll time and expense, and so forth. You also have the plant cash flows, those open purchase orders, open sales orders, things that are not cash yet, but they will be based on your terms. And then also pulling in the third party data, like FX rates can interest rates, and I think we are going through this as we speak, it’s fairly volatile of what we’re having, and how can you bring that together and deliver Cash Forecasting and forest risk exposure, so that from an end user standpoint, you have a daily cash position for the next 90 days. And then using that you have investment and hedging options. So think about that complete paradigm shift of a treasury function of where it is today. And then how it can completely transform to a new stage.
Sashi Narahari: [17:39]
My next topic is a favorite. This is where I’m in. I kind of say like first grade some of the second-grade learning, sometimes it’s good to think about questioning, right? So what happened, what is happening in the world of accounting, and we’ll talk about autonomous accounting. But let’s look back in time again, history tells you a lot about how professions have evolved. It all started with handwritten accounting journals, you go to a retail store, buy binders, and you have a column called credit, and a column called debit. And you put a description and you start recording your business transactions. What happened next? The next was actually a major innovation in the 1960s. And NCR machine. Has anybody heard of this machine? Accounting machines? Raise of hands? Okay, we have few. Wow, interesting. Let me play a video very interesting. Let’s look at this machine. Before. This is a machine that was a combination of a typewriter and calculator.
Machine: [18:33]
I think most anyone would agree that it’s better. It’s better in every type of accounting. It’s better for getting out your telephone, gas, and water and electric bill is the most complete machine for the job to be done. Built to give all the right answers all the time.
Sashi Narahari: [18:54]
Right, that is a machine that was invented by NCR with disrupted accounting back then. Because all they did was it’s almost like when you go to a restaurant, now you have the point of sale system. They don’t do every click possible. Just few and they just calculate. So there’s a typewriter and a calculator put together to make the accounting transaction postings easier. Right. It’s a unique machine. But then let’s look at the evolution today. I think whether it is SAP or Oracle, it is still a create read update service delete system, you’re still making journal entries manually, it’s just that it’s going into a database with the front end UI. So how can we change that? So let’s think about autonomous accounting the art of the possible. So your domain-specific Big Data is GL transactions and sub GL transactions. So the good thing with ARP is now you have this information easily accessible, you can get the GL and sub GL for your entire set of company codes. Once you have that, the AI and machine learning algorithms and we are piloting this with few clients is the first version of the anomaly detection is looking for missing and open transactions based on three variables. Given a transaction that got posted, did it get posted to the right month? The posting period by looking at the history of transactions? Did it post to the right GL account? And was it the right amount, which is a little bit harder because there is amortization and other complexities? And then if the system can just classify every transaction, every GL and sub GL, relevant one as red, green, or yellow, green means this looks fine. Red means maybe more the advanced technical accounting stuff, this really, really needs a review, and yellow is maybe a review. So what does it look like? So the simple anomalies are those 20 people, you could just workload back to the users, I see this in our business too, like we have somebody from our AP department, this is an invoice from Amazon. And they just code it as office expense, because they think we are buying something from Amazon. But it’s actually, we also use Amazon AWS for our infrastructure. So that’s actually cost of goods sold, it needs to go into a completely different jail category. But we’re looking at the vendor, they’re just coding it that way. So that could just be simply workload back to the AP user, versus accounting teams catching that, because it’s simple stuff. The more complex transactions can just be a collaborative workflow between accounting and the business users itself. So this is an example of a UI, the enterprise Web UI where you on a daily basis, you get a work list. Depending on your role in the company, it could be a business user accounting user, and the anomaly detection module is tagging every transaction that you need to look at. And then you can migrate from a reactive bookkeeping more, where business transactions are being posted manually, via journal entry systems could be ERP or third party. And then you always have this constant pressure of closing the books in as soon as possible. The CXOs of most companies say I can’t believe why you can close the books in a day, this should be so simple, right? To a continuous closing mode, where on a daily basis near real time, you have an internal business audit function, I’m using the term internal business audit, for lack of a better term. But how can a machine continuously look at every transaction and tag them so that the accounting department can evolve to a one to three-day close? That’s the art of the possible. Alright, so I’m going to summarise the key to this is domain specific. And I’ll explain why it is important. So let’s think about big data as a term. AI is a term, these are actually hyped terms. Everybody uses it, everybody claims it. But let’s see what’s happening out there. So from your perspective, many of you have probably heard these vendors, you have RPA vendors saying Intelligent Automation. And then you have the cloud AI majors like Google, Amazon, Microsoft, talking about AI. And then you have the ERP solutions, like SAP Oracle’s also document AI, and all these other vendors, any of the big data vendors like snowflake, Amazon Redshift, it can get really confusing, then you have advanced AI stuff happening with deep learning algorithms, the next generation of AI algorithms. So that’s could be quite confusing, because everything is AI, I don’t know. And double-clicking a bit more. I checked this like two weeks ago, there were 12,300 startups, AI startups in the world. $67 billion has been invested just last year, this never happened in the history of time. So even if there are 99 person failures, the remaining one person will literally power the economy, from an internet economy to intelligent economy in less than seven years. This is just coming. It just happened in the b2c world. And think about all the apps you have the b2b world is literally going to change in less than seven years. And that’s why we’re making this big bet. Now, all it is great, a lot of technology terms, what does this mean for you?
Sashi Narahari: [23:53]
Your business, folks, you’re not tech folks. So what we think is vendors with domain-specific AI will prevail because ultimately, you’re going to measure not based on how cool the technology is, you’re gonna measure based on the value it can drive. And we think all of us will be challenged to deliver value speed to value in less than six months. So otherwise, generic AI platform vendors will end up being a specific science experiment. I’m hearing this a lot I speak to some of you. That is because you still need to get hired data scientists, you need to still bring your subject matter experts sit with them, and figure out how to use this complex technology. Those would probably buy destined probably like couple of years. And that is why at HighRadius, we have our leadership principles. We introduced this leadership principle called maniacally focus on value to the customer because unless we force ourselves to disrupt ourselves, we will likely go out of business with the same business model. So what does that mean? Speed to value implement in two phases, phase one fast track in three months without the box features, and phase two optimises in another three months. Let’s take cash up as an example, let’s say the theoretical hit rate you can get is 85%. based on analysis simulation whatsoever, why can’t we go live in phase one, just with out of the box are already cash platform is 32 million lines of code. We have 1000 people in our r&d, can we just not turn it on with so much IP and see where we land and go live? This is how I’m challenging our consulting team. This is customer’s money at stake, the longer it takes to go live, the farther away they are from going back getting ready. I know there is a conflict because the customers want to get more requirements, the Fit gap analysis, but why not turn it on capture value and then do another phase, the first three months you have to do all the design debate, and then decide what to do. And that could be face to face. Okay, the phase two can be what is the optimize the Fit gap analysis, but only selection of features that deliver value. I do think that you need exact gardens committees on both sides to keep us all honest. So that we can focus on creating value for you. And being experts in whether it is already cash or Treasury or record to report and then the right governance committee to make sure and you don’t have to get the max possible value. Because I think of that as the long tail will spend end up spending so much energy and getting the last mile optimization that it’s better off to move on and solve another problem and do the value even we to capture. So this is a framework called Value calm framework, it has two dimensions, what you’re seeing on the x axis is low value, and high value, what you’re seeing on the y axis is enhancements and out of the box. The first is high value out of the box, do it in phase one, because you have the IP, you just need to turn it on, it’s high value. Number two is low value out of the box more with caution because we can sign up for hundreds of things and get distracted. So you have to be careful, it might be out of the box, whether now you have this laundry list of activation of features. Number three is high-value enhancement. These are the ones where you have to deliver enhancements, you have to be more thoughtful. So that’s why I do them in phase two. And then low-value enhancement. I use a phrase called walk over my dead body. But my marketing had changed it to avoid at all costs because it’s our customer’s money. And we’re not cannot suck ourselves into building all these enhancements, that is not creating any value. Alright, so in this is the transformation we announced on January 1, 2020. To all the departments a lot of hydrogens here, some are smiling, some are crying. So we got to do this and we had to force ourselves through this discomfort. So starting with product, every feature you build has to focus on value. Can you just look at every feature and not get carried away by the coolness? And then sales has to manage the customer expectations? Right. So you’re not overselling, you’re doing the ROI right thing, talking about phase one and phase two, and getting the right math. And then Consulting has to deliver speed to value that’s where the rubber meets the road. Phase one, phase two. And then it’s possible that after we go through these two phases, sometimes I’ve seen in the data-driven platforms, you lose some optimization because you’re more data. So you need to retain what you have, in some cases you need to do fix it up. So we created a value optimization services team, I think there is a booth here you can visit them. This is the evolution from the traditional customer success, or that software departments have to what we call as value optimization services. Because we are not CRUD software, we are a data driven intelligence software. alright with that closing, we think b2b SaaS will be dead. Autonomous finance is the next generation of software, but you will keep us honest, based on speed value. Thank you.
