"Right after we graduated EF, we started building the earliest layer of accounts receivable automation, which was collections automation. These are different modules within the same AR stack that help companies send invoices faster, collect on those invoices quickly, give clients professional branded portals to make payments, check invoice status, extract statements of accounts, and apply payments to incoming transactions. We built that first, and by 2022, the entire stack was in place." - Saurabh Chauhan, Co-founder and CEO of Peakflo


"We have been doing a proof of concept with a lot of our AR clients on a voice AI agent that can be embedded in the existing workflow. It coexists with automatic email reminders, SMSes, or WhatsApps, and somewhere in the workflow, you can insert a voice AI agent to make a call, for example, seven days past the due date depending on your preference. Just like a collection officer, it has full context about the client it's calling—it knows the outstanding invoices, the total amount due, the original due date, any open disputes, and even prior conversations like a promise to pay in two weeks. It tracks whether that's an upcoming or broken promise to pay." - Saurabh Chauhan, Co-founder and CEO of Peakflo


"Apply everywhere and figure out what suits you best. It's hard to be prescriptive when it comes to talent incubators because every cohort has different people, and a positive experience in one cohort doesn't necessarily translate into the next. Even if the ecosystem is supportive, it's ultimately about the co-founders you meet, and that carries the most weight. Test the waters and try to see where you can find the best match because that's what a talent incubator is ultimately about." - Saurabh Chauhan, Co-founder and CEO of Peakflo

Secure your privacy with Surfshark! Enter coupon code BRAVESEA for an extra 4 months at www.surfshark.com/BRAVESEA

Saurabh Chauhan, Co-founder and CEO of Peakflo, returns to BRAVE with Jeremy Au to reflect on their journey since first meeting at Entrepreneur First in 2020. They unpack how Saurabh identified pain points in finance ops during his time with Rocket Internet, how he structured his co-founder search, and how early customer interviews shaped Peakflo’s product roadmap. They explore why he rejected the social commerce hype, how Y Combinator reset his scale ambitions, and how Google’s AI Accelerator helped move Peakflo from traditional SaaS to agentic workflows. They also discuss startup fraud detection and how external stakeholders can cut through opacity.

02:40: Why Saurabh joined EF: He had identified cash flow and supplier payment issues in past startups and wanted to solve this. EF was attractive for its high technical founder density, which led to matching with Dmitry, a PhD in AI and former CDO at AirAsia.
10:02: Rejected social commerce despite hype: Saurabh entered EF with two ideas—cutting customer acquisition costs and finance automation. Interviews with 30 to 40 operators in social commerce showed no real CAC compression and surfaced product quality and platform leakage issues.
23:47: Peakflo started with accounts receivable automation: They built modules like collections, dispute management, customer portals, and payment reconciliation. It took over a year to fully build the AR stack from late 2020 to 2022.
25:40: Accounts payable followed based on demand: Customers didn’t want to use two systems. Peakflo added AP features like invoice capture, PO matching, and supplier payments, which took another year and launched by late 2023.
28:30: Google AI Accelerator pushed the move toward agentic workflows: Peakflo now builds AI agents to perform human tasks like logging into client portals, submitting invoices, and reconciling ERP systems.
30:26: Voice AI agents now handle collection calls: The agent knows invoice details, dispute history, and broken promises to pay. It engages clients like a collection officer, takes notes, and feeds updates into follow-up workflows.
36:32: Agentic workflows are the future of finance ops: Saurabh sees Peakflo evolving into a workflow engine powered by AI agents across back-office functions like lead qualification, collections, and month-end closing.

Jeremy Au (01:06)

Hey, Saurabh, good to see you! Good. So, we're just laughing because the last episode that we did together was episode 16, which was all the way back in 2020. Isn't that crazy?

Saurabh Chauhan (01:07)

Hey, hey man, how's it going?

It is. I just learned that you're on episode number 600 and something.

Jeremy Au (01:21)

it's a crazy number. Yeah, I mean, you know, time flies since we started doing this podcast. For those who want to check out your episode, check out episode 16, back when you were McKinsey consultant, Rocket Internet General Manager, and both of us were at roughly around, we're doing the Entrepreneur First program. And then I went into VC and then eventually, you know, start up CEO. And then you actually launched a program and company out of EF, Entrepreneur First with Dmitry as well. Yep.

The company picked for today. So, I thought it would be a great way to do a sequel episode about between the past five years.

Saurabh Chauhan (01:55)

Yeah, for sure, for sure. Thanks for inviting me back!

Jeremy Au (01:58)

Yeah, so I think maybe you have like for those who don't know yet, could you introduce yourself?

Saurabh Chauhan (02:03)

Sure, of course. So, my name is Saurabh. I am the co-founder and CEO of Peakflo. Peakflo is a finance ops automation software. So, we work typically with enterprises to automate their collections and payables workflows. Yeah. And yeah, I guess a bit about my background. My background is, you know, I started my career with consulting, then, was the general manager for a few Rocket enterprise, Rocket Internet ventures in Sri Lanka and Singapore. And then, yeah, finally decided to start my own startup

in the finance operations automation space. And that is what I've been doing for the last four years, which is Peakflo. Yeah, I think a few updates since the last time we spoke would be, I think the most two or three most prominent ones would be, we did start Peakflo in 21. And then in 2022, we finally raised a seed round at my Combinator. So, you had this insane list of investors, institutional investors that came on board.

Yeah, I think since then, it's been about two and a half years since we raised our seed and we're done with our zero to one journey. We're starting a one to 10 journey now, and yeah, just looking to keep the business growing and yeah, literally excited about what comes next in this journey.

Jeremy Au (03:11)

Yeah, so lots of things to cover. Obviously, for those who want to hear about your prior experience in a kind of like being a GM for Rocket Internet and those early lessons, they can check out episode 16. Who knows what the quality of the episode is, but it's available on Spotify and Apple Podcasts. But I'm just kind of curious, just, you know, I think what was interesting is that, you know, while you were an entrepreneur first, you know, we went through the experience, which is kind of like a, you call it a venture accelerator. There was a bit of a matching where you found

and match with your co-founder in B3. And then also there's an ideation phase as well. So, maybe you share your experience. Why did you want to join Entrepreneur First back then?

Saurabh Chauhan (03:47)

Yeah, sure. I think I had spent about two years in consulting, five years building startups, and after seven years of getting a taste of two things that were pretty distinct from each other, but still had the same amount of, I would say, exponential learning, because you're getting exposed to so many problem statements in consulting, and then you're getting more or less slapped in the face with the same problem statements at

startups but with a lot lower resources at your disposal. You quickly realize that there are a lot of gaps. And I think one of the problem statements that I personally identified was the challenges that I saw in my five years of building startups, which is just getting the money in and getting my suppliers paid on time. And I think those things were, I think problem statements that

I didn't see good solutions, at least in the APAC region for, and decided that, you know, that was something that I wanted to work on. So, the way EF figured is I'm a business background. So, like my expertise is in go-to-market strategy, sales, ops, and all that stuff. And I needed a really strong technical leader. And for me, just moving towards startup incubators, like Entrepreneur First was the next logical move because it would allow me to find a person with a technical background who would be as committed

towards building something that really makes a difference. So, I think that's how EF figured in the plan. I applied at a bunch of places, but EF seemed to be the best fit from a business founder perspective because they had a very high talent density back then for a lot of technical founders, deep tech, PhD folks would show up there. In fact, my co-founder, Dmitry himself is PhD in artificial intelligence.

So, I think that really did work out well for us.

Jeremy Au (05:29)

So,

you know, I think what's interesting is that looking back at that point of time, obviously, I was in a similar spot and I weighed Entrepreneur First and Antler at that time. Entrepreneur First has since left the Singapore market. But yes, I think at that point of time, Entrepreneur First had a lot more technical talent than Antler and a lot more people like us.

Saurabh Chauhan (05:49)

Basically, it's like looking at yourself in the mirror. Everybody seemed to be from Rocket.

Jeremy Au (05:53)

And then I was like, because I went to accepted by both programs and I was very much like, well, don't know, it's like dating, right? It's like you want to see people that you want to match with, right? And so, I was like, okay, you know what? I'm not here to date myself. Okay, well, thank you, very kind. But I think that's why I chose Entrepreneur First at that point in time. But I think what's interesting is that when you chose Entrepreneur First, I think when you look at the whole cohort, I think there's so many people where there are different stages. And it was this accelerated that was like,

Saurabh Chauhan (06:05)

You're not that bad, man.

Jeremy Au (06:22)

finding a co-founder, who you work with, and the second part of course is ideating and coming to a consensus on what problems you want to solve together, right? So, could you share a little bit more about that first experience and after that we'll dive into the second.

Saurabh Chauhan (06:37)

Sure, I think in terms of, yeah, so I would say it it was like speed dating, but also on steroids. So, if you recall from EF, we used to literally make teams and then within a week, either things are working and you've set some predefined criteria that you want to achieve to quickly validate the problem statement you're working on as a group, as a unit. And then if you meet those criteria in a week or maybe two weeks, then you say, okay, yeah, I'm going to build with this guy.

And if you don't hit those criteria that are mutually defined and you say, okay, fine. It was great working with you for the last two weeks. And I'm going to pair up with someone else. In my case, I took a lot of time to like, just make that first pair. So, I think, I maybe the first couple of weeks when people had already gone through maybe two or three pairs, I actually invested that time in like meeting tons of people. I think by the end of it, I had like a spreadsheet with the names of like 50 odd batch mates in terms of what were the, I would say, complimentary skill sets and complimentary backgrounds that they had to mine.

And it was all about sort of making sure that was a good fit. And those were done in the form of like these one, one-to-one coffee chats, like really understanding what motivated them, why were they in this? Because, you know, when you have such a big group, a lot of people can be in it for fairly different reasons. Some are just like out of a long-term role at a big company and they're like, I just want to relax and see how things are in the market before I make a big decision. Some would be, I've been thinking about the strong statement, but then might feel a bit disillusioned when they actually realized that you

know, climbing the startup hill isn't, yeah, it's pretty, pretty hard, right? So, there were a lot of experiences and lot of, you know, discussions that happened with people. But I think the reason why it really, really clicked with my co-founder Dmitry was I think in terms of the idea, he was completely on board with solving something that was super, super vertical and super, super specific to an industry or a pain point that necessarily didn't have to be like super sexy. So like,

nobody wakes up in the morning and says, you know what? I'm going to do financial operations automation. It isn't like your very sexy space to begin with. So, I think he was totally on board with solving maybe even a boarding problem, but that was intense enough for the right ICP or our target market. And I think that really helped. And then second is obviously, he has super credible technical expertise and

I think it was a no brainer given his background from as a PhD, CDO of AirAsia and also the fact that he had worked in startups. So, I think that was a really good fit wherein we could sort of meet each other somewhere in the middle and figure out in terms of, you know, having already having shared values and also having like common consensus on what we wanted to work on. So, yeah, I think just like every other team, we set certain key metrics. Like I'm going to make 50 customer calls over the next two weeks.

And we're going to validate whether this pain point is even worth solving for. Like I think, that was our idea validation journey, wherein I literally reached out to ex-CFOs, current CFOs from my own startup network and said, I just want to talk to you about what challenges you faced in collections and payables over the last two years, three years, four years. And we would just have conversations with them where we're just trying to validate whether it's an area that is painful enough, what are the solutions or substitutes that they're already using to solve for it.

And if there are some gaps that they still see as, you know, industry leaders of the finance function, so I think those things work. We met those predefined criteria and we decided, okay, this is the space and we want to put our head down and just, you know, start building. To our credit, I think, especially Dmitry's credit, we literally put our head down and for the first, I think, six to 12 months, we were just building and building and building and iterating on the, I would say the product.

Jeremy Au (10:15)

And I think what was interesting, obviously, is that I think everybody in the cohort recognized that Dmitry was a very strong technical leader. And I mean, I too also had a spreadsheet. And Dmitry was also very high on that list. So, I think that was an interesting piece. And there's always that joke about MIT. It's like everybody's coming to MIT looking for a CTO. So, I guess what's interesting, of course, is less about fact that recognizing they're strong, but the fact that you're able to ideate

and find the idea and also that it's on board. And I think obviously, this idea is a little bit more compressed because it's within a program format rather than, you know, the dragged out process where you had the idea first and you meet a co-founder. But I'm just kind of curious, like, how did that process of coming on the idea, how did you land on the idea? Did you walk into the program thinking this is something you wanted to do or was it something that as you're debating and brainstorming, how did it come about?

Saurabh Chauhan (11:08)

Yeah, I think they were like, I walked into the program with like two main challenges that I had seen and while building e-commerce for the last five years, first at the RAS and then at Vanity. And the two problem statements were customer acquisition costs are ridiculously high. Is there any way in which we can push down CAC and my finance operations headcount in the RAS was like 25 % of my team, makes no sense. Is there any way in which we can streamline these processes without having investing into, I don't know, the technologies that

time, which used to be ridiculously expensive. And I'm talking about SAP level intensive investments. So these were the two problem statements that I did walk in with. I think I wasn't particularly inclined towards it. I wasn't 100 % about either of the two. So we did tons and tons of calls, as you would do with different CFOs, I think, for the CAC

conversation, it was also done with a lot of CMOs and Chief Business Officers and all. And yeah, I think at the end of the day, we thought that this was a better fit. But yeah, I think we were pretty close to also considering a venture in the social commerce space at one point in time, because personally, like in e-commerce, like, you know, you're always talking about that magic LTV to CAC sort of ratio, and you're trying to figure out how to like optimize your lifetime value for every customer required while minimizing the customer acquisition cost and

back in 2020, social commerce was a space that was exploding. And we thought that was also a pretty reasonable space to deep dive into. But just we couldn't make the numbers work. We did lots of validation calls. We realized that based on the feedback we got from about 30, 40 odd conversations we had with either people who were doing social commerce companies or had very deep insights into how these social commerce platforms work, we weren't really able to sort of

validate the hypothesis we had, even though it was a way better funded space. Like if you recall from 2020, almost every social commerce startup seemed to be exploring and getting funded, but we just couldn't agree on whether it made sense from a scalable startup perspective. Whereas, in terms of the FinOps automation, I did have that conviction that the space was

important. There wasn't any particularly like, it wasn't really a hot space back then. But I think what we were really sure about is like, this is a deep enough pain point that even though it's on the sidelines, industry leaders might not be very vocal, but it is a pain that is felt by the CFOs and the finance teams. So, I think that was pretty much about how we sort of drifted into this ideas, did multiple calls and then realized that, okay, this is worth pursuing. We were not able to validate the hypothesis that

the social commerce route indeed contracted the customer acquisition cost as was the initial hypothesis for the tons of companies that went into that space. We just did not see it in the numbers. So, we spoke to people who were actually working at very fresh and young social commerce companies. We spoke to folks that were on the ground working with those sellers and reported that guys, I mean,

there's no compression in customer acquisition costs. If anything, you might face other additional issues like low quality of products being supplied through these sellers that are sourcing the customers or potentially the sellers moving those customers away from the platform. So, there are other internal issues that you would face in the pursuit of trying to compress that customer acquisition cost and thereby optimizing your lifetime value to CAC ratio. So, that's not going to... We just didn't see it happening.

Not to say that it's not a successful model. There are still some amazing companies in this space like Nicheo that seem to be doing very, well. Yeah, it just seemed as it did not appear to be a home run as we had initially thought.

Jeremy Au (14:51)

I think it's interesting because I think there so many founders who have gone on the growth side and I think we've seen a lot of that. Social commerce was also super hot in 2021, 2020. Like the fact that you chose the contrarian bet, I think it's proven to be the better bet in retrospect. But I'm just saying that I have to give you kudos because, you know, it was definitely contrarian to the rest of the group, I would say.

Saurabh Chauhan (14:59)

It was.

Jeremy Au (15:16)

So what's interesting is that obviously you've stepped out of EF, you eventually went to YC and so forth. I'm just kind of curious that how would you have contrasted your EF experience versus your YC experience from your perspective?

Saurabh Chauhan (15:31)

Yeah, it's worlds apart, right? I think EF did a fantastic job in terms of being how should I say, a great space to meet a co-founder. I think Y Combinator just like completely changed our perception about what is possible. When you see amazing companies like, I don't know, Stripe and Airbnb, and you have those founders actually come in and you get to speak with them in person and they

talk about mostly off the record about their experiences, their challenges, take your questions. Yeah, it's completely different because then you're no longer thinking about like our perception in EF used to be, hey, we just want to get to the first milestone and like get to the first hundred thousand dollars in revenue, get to the first million dollars in revenue. after going through YCN and speaking to, you know Brian Chesky or to the, you know, co-founders of Stripe or to co-founders of, I don't know.

DoorDash and all those other large VICE startups, the perception became, okay, how do you get to $100 million in record revenue in a super profitable manner? And how do you, like the in-depth perspective was hyper growth. And because you're surrounded by such like, you know, high energy, high potential group partners at VICE, but also these alumni's, you basically move into a stage or in an environment where everything is possible. And it's no longer just

aspirational, but it's almost like if you just work hard enough and you do the right set of things, you can scale to that number. So I think that was a complete mindset shift between EF and YC that I personally experienced and also my co-founder experience. You know, that would be the two main differences between the program. Not to say that one program is necessarily better than the other because they have completely different objectives. YC is a startup accelerator period.

Whereas EF is essentially a talent incubator where it is just trying to pair up people with complementary talents and allow them to like, you know, explore their mutually agreeable ideas.

Jeremy Au (17:29)

Yeah, I'm just kind of curious because, you know, EF left Singapore and obviously it shrank and then Antler has grown and so so forth, all those different dimensions. But I'm just kind of curious from your perspective is, you know, how would you recommend talent incubators change or do differently? Because I think you know, we've seen all of them. We've had friends also ask us about it. You are a successful graduate of a program of a talent incubator. I'm just kind of curious what you think.

Saurabh Chauhan (17:55)

Yeah, think, so if you think about it from the why probably they left the market, I think the economics obviously did not work out for them. Like nobody's gonna abandon a market unless the numbers don't make sense. I would say like, I don't know, if I had to just ballpark 20 startups graduating per cohort and them investing a ballpark of $100,000 per startup, then.

They are making with every single six month cohort, they're probably making an investment of about $2 million into these startups. So obviously, like simple VC math, they're looking at some sort of a markup of about like, three X, four X on that number. And if you have subsequent cohorts that do not have that markup materialize even let's say four years down the line or three years down the line, then it becomes hard to justify. I wouldn't say there is something, there was something particularly

wrong with EF strategy, but I think probably if I really had to take a short or take a guess why that bet or the economics didn't pan out is because EF was super focused on deep tech startups, which tend to have a much higher failure rate and a much longer sort of payback or markup period like in VC parlance, right? Like the markups would happen probably longer down the line because there's a lot of validation that you have to do and lot of like...

milestones that unfortunately deep tech startups have to jump through. Whereas, the other incubator that you mentioned like Antler was pretty much going for asset light, high software startup ideas, which are again, much higher, let's say survival rate and much faster markup cycles in terms of subsequent rounds of funding. I think just that small difference in their initial selection process

which startups graduate from that cohort, I think probably was the reason behind why the economics didn't work out. But from a cohort, like the sort of people that we interacted with, top notch quality of, especially my experience with technical founders, really, really smart folks, right? So that was great. The team was super supportive. Like I think if you recall, Bernie from our batch, like who was our GM and like

was also helping a lot of folks and all the mentors that we had. I think in general it was a very good experience, but if somebody is really figuring out why one talent incubator works versus the other, it could be that they probably have different preferences in terms of which startups they pick. And I think that probably was what did not pan out well for EF. That's my assumption again. I might be completely off here.

Jeremy Au (22:03)

So, you know, I agree with you that from a team perspective, I would say that at that point of time, the EF team, I think felt more supportive than the Antler side. And I think the economics actually were interesting where I think EF actually gave, I say a similar quantum, but actually took up less ownership than Antler did, which gave a similar quantum, but took a higher

ownership when you did the proper calculations. I would say that as a result, adding on to your point here is that, at your first exactly, like you said, chose companies that will take longer to demonstrate that markup rate, but also took less ownership. So I think that was a double whammy. Of course, I think that pricing attracted me as a sophisticated person who did analyze the numbers. And I think

it also attracted maybe a stronger group of business co-founders from what I saw at that point of time. But I think it didn't make up for that other set. So I think it's gonna be interesting. think there's continues to be a space for talent incubators. Before we wrap up this chapter, talk more about peak flow. Any advice you have for people who are thinking about talent incubators, whether there's on-deck or hacker houses or antler, any advice you have for them?

Saurabh Chauhan (23:17)

Apply everywhere and figure out what suits you best. It's really hard to be prescriptive when it comes to talent incubators because every cohort has different people. And just because somebody had a positive experience in the previous cohort doesn't necessarily translate into a positive experience with the next cohort. I think even if the ecosystem is super supportive, at the end of the day, it's about the co-founders that you meet over there. And that just weighs a lot. Probably, you know, test the waters and...

try to see where you can find the best match because that's what talent incubator is ultimately about.

Jeremy Au (23:48)

Actually,

I think that's a great phrase and I think that I'm going to give the same advice as maybe be more promiscuous.

Saurabh Chauhan (23:54)

That's probably one area where it's permitted.

Jeremy Au (23:56)

Yeah,

exactly. Because, you know, it's like every single cohort is going to say like, we are the best cohort, blah, blah. But at end of the day, you don't know. And you might as well parallel track and apply to all of them simultaneously. Because some of them are going to reject you. Some of them will set you. And some of them, when you go in, you're going to realize it's not a good fit anyway, or there's nobody's good match. But as long as one out of the five or 10 talent incubators throws up that one good co-founder, then there's a success for the whole search, right? That's true.

And also I think a lot of these talent incubators, also tie their funding based on if you find a co-founder within that program as well. So you dispersing as long as you quit early, I think people are going to be like no harm, no foul. You are going to go somewhere else, right? So I think that would be my advice on this. Talking about peak flow as well, I think what's been interesting is that over the past four to five years, you've built out both the accounts receivable and accounts payable. Do want to explain a little bit more about

I guess which module came up first and why, and how you build out the second module.

Saurabh Chauhan (24:54)

Sure.

So yeah, think based on probably we may have touched on it in my last podcast, maybe not actually. But essentially based on my experience at Veraz and Vanity, which are two distinct rocket ventures. At one, I had a lot of pain points when it came to just paying suppliers on time. And at the other, was about collections. We had to collect our commissions from salons and both were...

Jeremy Au (25:00)

Definitely not because...

Saurabh Chauhan (25:18)

independently, the biggest challenges that we faced because they essentially impacted either our cash flows or our relationships with suppliers. Both are pretty essential. And in a world where we had all these different tools, like we got a great tool for our sales team, great tool for customer service, great tool for pretty much every team. We were still stuck with an ERP system for the finance team and a lot of the stuff was being done manually. So I think the way peak flow

ended up being was an accounts receivable and payable automation tool. The ER tool came first. So right after we graduated EF, we started sort of building the very first, earliest layer of accounts receivable automation, which was collections automation. And then we built out multiple modules on top like dispute management, cash application, payment collection, customer portals, so on and so forth. So these are essentially different modules within the same accounts receivable automation stack that essentially help companies send invoices faster.

collect on those invoices quickly, give their clients professional branded portals where the clients could potentially go in, make payments, look at the status of their invoices, extract statement of accounts, and also would be able to apply those payments pretty quickly to their incoming payments that they're getting. So that's essentially the end-to-end AR automation stack. So we built that out first, and that took us, like I said this all very quickly, but that literally took us like more than a year to...

put in place. So we started building that out late 2020. And then I think it was only by 2022 when we had this entire stack in place. I think AP came second simply because when we entered Y Combinator, like I mentioned earlier, like, you know, it just completely changed our mindset about like what a successful company could do and how big it could be. And we remember this conversation we had with the founder, I believe.

If i'm not screwing up his name, Parker Conrad. And he... Rippling, yeah. Rippling. And then before that, think it was Gusto or... Zenefits, yeah. And then he started rippling. And I think one of the phrases that he used very frequently in his talks was compounding product, compounding growth from building a very sort of deep product. And that was...

about how he was able to sort of become the one stop solution for the entire HR stack by essentially solving all the different pain points that the HR team had to sort of deal with and was responsible for. And we were like, okay, so let's now sort of go into the next most logical pain point of helping you collect faster. What is the next step? Well, let's help you pay more efficiently. And that's how the AP conversation started. So we started with purchase requisition to...

automatically capturing the supply invoices that you're getting, automatically matching those supply invoices to these purchase orders and then helping with payment processing and disbursement. So I think these were some of the modules that we started building out. And then again, it took us like a good one plus year to also build out the entire AP stack. And it was only until by late of 2023 when we had a, I would say a full fledged AR plus AP product in place. So yeah, that was essentially the journey.

Our existing customers actually were validating the pain points pretty clearly. So I think a lot of our ER clients would often come up to us and say, Hey, do you guys have an AP solution? Cause I'm looking for something in the market and I don't want to pay for two different softwares. Technically I would prefer if it's just one. That the first set of customers that we actually upsold the AP solution to were all our existing clients, because it was the most logical thing for us to just say that, hey, we already have a relationship with you guys for the last two years or one year plus, and you've been using our system.

Here's the AP solution and let's roll this out. So we rolled it. So those are the first set of clients we rolled it out. Now, obviously it's a standalone system and we've got tons of clients who would only sort of go for AP versus AR even today. But yeah, I think that was the genesis of AR and AP. The system has definitely, I think in the more recent months, we actually also ended up post our journey at Vycombinator. We ended up going through Google AI Accelerator in September of 2024, which is like about six, seven months back.

And I think what was super revolutionary about our experience during the Google AI Accelerator was putting in the foundation that allows us to move away from being a SaaS-only product towards an agentic workflow product, which means you've got independent agents rather than, let's say, humans that are actually using the software to achieve the outcomes that it needs to deliver. So agents making calls for collections.

or agents doing reconciliation or your month end closing by literally operating your ERP and just doing closing one by one and reconciling your payments or agents doing invoice delivery to your customer portals like we have tons of clients who are working with large enterprises and those large enterprises would say, hey, if you want to get paid, I need you to submit this invoice into this really unique portal that I have,

which no one else does. And when you've got 50 enterprise clients like that, it doesn't matter what sort of internal automation you have for your ERP processes. At the end of the day, that invoice is created, you still have to log into 50 distinct portals and deliver it. And that's what a human is still doing. But with an agent, it invoice delivery bot, you basically have that bot just like a human pick up the invoice from your ERP log into each of these independent client portals, which you can't build an API integration because it's not scalable.

and drag and drop those invoices into the portal at scale concurrently without having to take a hit on your DSO. Because one of the things that actually delays payments, especially from large clients is when you're not really jumping through all the hoops and not doing the invoice delivery properly, because they're going to reject it. And that's just going to be a delay in your day sales outstanding or your average collection time. It doesn't matter how many collection calls you make. So I think we quickly realized that after doing the Google AI Accelerator that

you know, applying agentic workflows into areas that cannot be automated through conventional SaaS is really the future. So that's, I would say how the product has really, really evolved in the last six, seven months.

Jeremy Au (31:25)

No, it's fantastic. And do you have any examples of what something agentic would look like?

Saurabh Chauhan (31:32)

absolutely. In fact, we have been doing a proof of concept with a lot of our AR clients on a voice AI agent that would literally can be embedded in the existing workflow. So it sort of coexists with automatic email reminders and automatic SMSes or even WhatsApps, depending on what is the size of the client that you're collecting from. And somewhere in the workflow, you can just insert

a voice AI agent to make a call, let's say seven days past due date or 14 days past due date, depending on what is your preference. And it would just like your collection officer would have all the context about the client that it's calling. So it would know what are the outstanding invoices? What is the outstanding amount? What was the due date that it should have been paid by? If there are open disputes, it would already be aware of those disputes. If the conversation has already happened with the client and the client has responded, I promise to pay two weeks from now, it would know

that that's the upcoming promise to pay or that's the broken promise to pay if it's in the past. It would have all the contacts and just like a, you know, a collection officer would typically call the accounts payable department of their client to get paid. This agent would basically call up the AP officer of the client and say, hey, this is the status. Could you help me understand what is the promise to pay date? By when would you clear this and take notes from the call, just like a regular collection officer would.

And upload those notes in a report that can then be used to define next steps. So most of our AR clients would have these weekly meetings with their collections team and say, okay, what is the status on our outstanding receivables? And they would typically open up this report and all the collection officers would have to update every single line in the report for every invoice. And what is the promise to pay date? What happened when they made the call? What was the outcome? Was they able to connect? Was it not connected? What did the client say? So all that is very neatly and painstakingly

populated by collection officers and essentially can be done by an AI agent. So I think that's one of the first POCs that we've rolled out with our clients. I'd love to demo it.

Jeremy Au (33:29)

Yeah, if I have a chance. Why don't we demo it now? Yeah, sure.

Okay so we receive a call from this agent. Let's see how it goes. And then I'm put it on speaker mode. Okay yeah, uh who's this?

Saurabh Chauhan (33:39)

Hey Jeremy, this is Steven from ABC Corporation. I'm calling regarding your account. I noticed that you have five invoices with $6,673.80 outstanding, of which $4,466.55 is overdue, with the oldest invoice being 30 days past due.

Jeremy Au (33:54)

There's a lot of information for me to hold up right now. Can I pull up my laptop and... Sorry, who are you again?

Saurabh Chauhan (34:01)

Yeah, sure. No worries. Yeah, sure. I'm Stephen from ABC Corporation.

Jeremy Au (34:04)

Okay.

And are you the accounts payable, the accounts receivable team? Okay, great. So what do want me to do?

Saurabh Chauhan (34:09)

the accounts receivable team.

We value our relationship with you guys and we want to continue providing you with the best service. To do that we need to ensure timely payments. So when do you think you'll be able to make a payment on these invoices?

Jeremy Au (34:24)

Uhhhh I have to check to my manager maybe next month

Saurabh Chauhan (34:30)

Okay no

problem. Are you the right person to speak with? Okay, so you're thinking next month. Can you be more specific?

Jeremy Au (34:36)

I have to get approval from my manager first. Then after that, have to log in to system. And then after that,

Saurabh Chauhan (34:42)

To get the approval, can you give me a specific date you are targeting next month? Okay, well since it's more than a week away, I'll need to get approval from my manager for that payment date. I understand you need to get approval from your manager first. In the meantime, could you provide me with your email address so I can forward you the current statement of account? It has all the invoice details.

Jeremy Au (34:46)

I don't know. Thank you.

But

Yeah, can email me at the email you probably have which is finance.

Saurabh Chauhan (35:06)

of account to finance eight. In the meantime, I'm going to need to get approval from my manager for that payment date sometime next month. Can I put you on hold for a moment while I reach out to them?

Jeremy Au (35:15)

I think we can end the call now. I mean, you know, this is part of the interesting to see this happening.

Saurabh Chauhan (35:21)

So I think, so basically the call summary, which basically gets fed into the report comes in and says, okay, customer needs to check with their manager before committing to a payment date. customer will check internally and call back. So this would essentially go into the collections report that, okay, the person that we spoke to needs to check internally before they can even give a payment date. And that would automatically schedule maybe a follow-up call in a week or so that, hey, were you able to check? Do you think you can give us a date

to commit? and that's basically what an actual collection officer might even face. So instead of getting a clear response like, I promised to pay X amount by Y date, they could just have to do a lot of chasers just to get a date to begin with. Yeah, so that's really where the tool shines because the one call that you receive, we could essentially do concurrent hundreds of calls at precisely, I don't know.

12:48 PM because that's the lunch hour where you can actually catch people, but they're just busy.

Jeremy Au (36:17)

Yeah, that was really interesting because a year ago, this technology was not commercially available. It was impossible to automate such a call. And now it's, I'll say, 95 % there. I mean, I think that the good things, of course, is that...

it hit the key points, which is when are you going to pay me? Two, was able to answer some questions and have some of conversation. And three, of course, is that zero work for the actual accounts receivability, which are normally quite a small fraction of the finance team of the company. So those are good things. Obviously, I think the things to work on would be obviously as it climbs through the human likeness of it. I think it's pretty close. Obviously, there's some parts where you're like, this person is

either very enthusiastic or AI. So I think that's going to be interesting, but the technology obviously is going to improve over time. So I'm sure this is only going to get better over time, not get worse over time, that's one. And then two, ofcourse, is making sure that it's part of the right ladder of escalation, right? Because I think at the start it's all emails, then it becomes letters, maybe it becomes calls, and then it becomes ultimately the calls after that.

So think there's an interesting funnel that's going to happen. Those are my quick sense, but yeah, I think it's really going to be the future for accounts receivable. I'm just kind of curious that when you look towards the future, any parting thoughts about what you see the future for peak flow is going to be?

Saurabh Chauhan (37:38)

I think it's going to be deep diving into probably deeper into one of the verticals that we're serving and really deploying these agentic workflows for a broad sense of back office operations. Potentially, if you've got some inbound queries coming in and you need to qualify those potential leads or you're trying to get certain queries answered upfront without raising a ticket for human involvement.

Or whether it is what you just experienced in terms of having a collection agent sort of doing calls at scale. I think those would be some of the more specific areas that we see ourselves deep diving into, but it would all fall under the umbrella of having AI agents that are performing specific back office operations or tasks and being able to allow them to do it just like humans would at scale, but part of a workflow. So as you correctly pointed out that having calling is part of a larger

Dunning workflow or collection workflow where you have those emails and you have those Dunning letters as you said, and then you have calls as maybe the third step or the fourth step. That's essentially how the entire workflow design would look like that. You know, X days past due date, then only make a call and then Y days past due date actually escalate the call and also call the manager of that company or something like that. Only if you hit certain triggers, those actions should be taken. So I think that's probably going to be the correct way of how these agentic workflows are going to work.

So I think that that really is the future. And I think this sort of ties in into how technology has evolved. I think as a consumer, you may have felt that in the last six months, your Google searches have really gone down, even though that was probably how you were interacting with the internet for the last, I don't know, two decades. And that really is, I would say it's a seismic shift if I was a Google shareholder, but yeah, nevertheless, it's a massive shift in how the consumer interacts. I think this is going to be the similar shift that is seen in

large businesses and I think deep flow along with a lot of companies is going to be at the forefront of bringing that about but more focused on the vertical that we are good at. So I think agentic workflows is really going to be our future.

Jeremy Au (39:45)

Awesome. On that note, thank you so much. I'd like to summarize the three big takeaways. The first of all is thanks so much for sharing about how you went through the experience in the talent incubator and how you went about, you know, kind of thinking through product market fit and finding a co-founder in an accelerated format. So that's one. Two, thanks so much for sharing about, I think, your own experience building flow about some of the product market fit, the expansions, building a deep company and how to experience a wide company to help make some and crystallize those insights.

And lastly, thanks for sharing about how you see you building AI agents into your workflow.

Saurabh Chauhan (40:18)

Thanks for having me here.