Section Insights
Trusting Fintech with Money
How can fintech companies gain trust from customers?
The speaker reflects on initial skepticism about fintech's ability to store money securely, but notes that trust can be built quickly, as evidenced by a significant deposit made shortly after launching their service.
- Trust in fintech can develop rapidly, even in uncertain times.
- Customer experiences can shift perceptions of fintech services.
- The future of fintech is unpredictable, but opportunities exist.
Automation in Banking
How does Mercury plan to balance automation and customer interaction?
Mercury aims to automate as many processes as possible, especially for lower loan amounts, while acknowledging that larger transactions may still require human interaction.
- Automation is a key focus for improving customer experience.
- Certain transactions will still necessitate personal communication.
- Mercury is evolving towards a more automated banking model.
The Role of AI in Business Growth
How is AI impacting Mercury's growth and operations?
AI has accelerated product development and hiring at Mercury, enabling faster innovation and a competitive edge in the fintech space.
- AI is a significant enabler of rapid product launches.
- The competitive landscape is intensifying as all players leverage AI.
- Hiring in engineering is crucial to support growth and innovation.
Understanding Startup Financials
What trends are observed in startup financials and customer retention?
Mercury's revenue is growing as more startups use their services, but the main challenge is customer churn due to business failures, particularly among new startups.
- Revenue growth is linked to increased customer usage.
- Churn primarily results from startup failures rather than service issues.
- Mercury is expanding its customer base beyond tech startups.
Evaluating the Startup Ecosystem
Are we in a bubble regarding startup valuations?
While AI is transformative, there are concerns about overvaluation in the startup ecosystem, with many companies unable to sustain high valuations.
- There is a risk of over-investment in early-stage startups.
- Not all startups can succeed despite high valuations.
- The demand for data center capacity remains strong, indicating ongoing growth potential.
Transcript
0:00 I was like, you know, this is going to be a slog. I was like, who's going to trust you know, this fintech to like store all their money? Within four days, I think we had someone sign up that I hadn't met that transferred a million dollars to their Mercury bank account. I think the future, especially at this time, is like really hard to predict, right? Like, I mean, why isn't there a business of zero? Right? Like, is there a CEO like that?
0:21 >> Like, go away. Leave me alone. Like, stop asking me questions. I mean, there's there's clearly going to be like lawsuits every direction at some point. Somebody's going to spend money that they regret even if it was like their fault, and they're going to like sue everybody involved and say an agent was involved. Like, you should have stopped it, right? Like, do you worry about that as a company? Is that a risk? >> Maybe, but I'm I'm guessing the same thing happened when internet banking happened.
0:45 >> Immad Akhund, CEO of Mercury, thank you so much for joining the Newcomer podcast. >> Excited to be here. >> Newcomer, you know, it's I think I'm safe to say it's like a you know, 3 million revenue business. So, we run a lot of it on on Mercury. During the Silicon Valley Bank crisis, we opened you know, a JP Morgan Chase account, you know, to hedge our bets, have both. I'm sure you had to deal with that. And honestly, I think I didn't quite appreciate how awesome Mercury was until I opened the old-school banking account.
1:15 I'm like, oh my god, like it's hard to get in. It's hard to move money. It's like not as aesthetic, you know, lots lots of different things. What I mean, we have some you know, news. Obviously, you have some new product announcements. But do you want to just start off with like, what was the core insight when you started Mercury in the first place? >> Yeah, it came together in lots of ways. It's actually my fourth company. So, I was like pretty and I sold my previous company and I was like, you know, I want to do something that I could do for 10, 20, 30 years.
1:45 And I don't wanted it to be like a company that gets more interesting over time. And you know, my core insights were you know, number one, I I like these tools that every single business needs are like really powerful places to build businesses, right? So, if you think about like Gusto or Stripe or Slack, right? Like those are like these core things that every business needs. but, you know, in 2017 when I kind of first started Mercury, banking was one of these things that every single business needs, but like the experience sucked, right? Like I would and I had this like, you know, because I'm an entrepreneur so for so long, I was like every single time I had to kind of sign up for a bank account, make a payment, try to get an API from a bank, like all these things it was just so hard to do.
2:30 so that was definitely like a core part of it. I just thought huge market, every single business needs it, it's so painful, and no one likes it, no no entrepreneur likes it. And and yeah, actually, if you think about it, we launched in April 2019, and at the time I was like, you know, this is going to be a slog. Because I was like, who's going to trust you know, this fintech to like store all their money?
2:50 and there's a there's a really large component of trust with Mercury. So, but yeah, within days, I think, we had someone sign up that I hadn't met that transferred a million dollars to their Mercury bank account. And I was like >> networks did you really lean on in the beginning to get >> Honestly, it was really organic. I mean, I had a really broad set of investors. I had about 60 investors in my seed round. and it was partly that I was like, "Hey, if I get a broad set of investors and then recent orbits was an investor." we had Elad Gil, a bunch of other kind of well-known entities that would kind of engender trust in us. And I think that did really kick it off, right? Like all those people kind of tweeted about it and did all this thing and it got the snowball rolling, but then after that it was very kind of organic and word of mouth. I mean, even today, you know, more than 60% of our customers are organic and word of mouth. So, like, roll into this news that we have at the >> Yeah, yeah, I was going to ask you a direct question about it, but yes.
3:46 What is What is the milestone? Where are we today? >> I mean, like the yeah, crazy thing is actually >> you know, cut to alert alert flashing blinking. >> We just had about 1,300 people that work at Mercury in one place in New Orleans a couple of weeks ago and I actually literally the day before we hit this milestone where we hit a billion dollars in revenue per year. So I was just talking about how like, you know, in 7 and a half years we've kind of gone from literally zero to a billion in revenue.
4:14 >> And this is if you take what a one month span and >> yeah, I think last month times it by 12 kind of thing. And yeah, we announced I believe 650 million revenue basically at the exact same time a year ago. So you we've grown 50% actually I'd say the growth has kind of accelerated in the last year really driven by kind of AI. Yeah, we're a lot of AI companies use Mercury. I I think there's been acceleration in entrepreneurship because, you know, AI has enabled kind of anyone to be an entrepreneur, especially online online entrepreneur.
4:47 >> Do you make more money the more people interact with their banking information or is it Yeah, is it sort of the usage increase or is more customers seeking the features you have on offer? >> so it's both. Like we make money on kind of five different ways, but yeah, it's like if you hold more deposits with us, we make more money. If you use our cards more, so we have a corporate credit card that's kind of our second biggest revenue stream.
5:10 We have FX, we have kind of treasury. More recently we launched books about a month ago and that's a subscription plan. Actually next week along with this podcast we're also launching payroll which will also be a subscription plan. So you know, it's kind of a plethora of different kind of models. >> Start doing business here and then >> Yeah. >> sort of expand. I'm curious on payroll specifically, so many of the payroll companies target specific size companies. I imagine you have a range of >> Yeah.
5:37 >> business sizes. Like how do you think about that? >> You know, most people start Mercury in the same way you started it. They start a business, they're like, I need a bank account and they look at things and they're like I like I can you know walk to my bank branch at JPMC or Chase or whatever or I can just go online and you know do it. Actually someone just like posted on X that they use Instinct to sign up for my bank account. So yeah, we just make it super easy to get started and you know especially at the start like you just want to get on with things right?
6:10 Like you as an entrepreneur are not like wow, I need to like do all this work to set up like my business. You want to go start building your product. so we make banking really easy and yeah, 75-ish percent of our customers are within 6 months of incorporating so we're probably their first bank account. and this is how we think about kind of all the other products we're building. So yeah, when we think about payroll at some point you're going to be in your bank and like yeah, actually often people start off by paying contractors and they're paying them straight from their bank account and you know, we make that really easy. And then eventually they're like oh, actually I want to bring on this contractor as an employee or yeah, I want to hire someone or I want to pay myself as a W-2 employee. So it's like a very natural thing for a business.
6:49 Yeah, exactly and same with books right? Like we already have all your transactions and we can already automatically categorize them and eventually you're like hey, actually I need a P&L so I can you know do taxes and things like that and we want to we want to make that like a super smooth kind of transition. so most of these tools we're thinking about as additions. so we've already you know, got bill pay in, invoicing in and credit cards >> Another product I'm loyal to another startup, you know, Gusto does payroll.
7:17 That was sort of our other early business building tool. What's what's the pitch to switch I guess? I mean and in particular will you offer, you know, they chased me down and say oh, Massachusetts is mad that you haven't paid, you know, filed this thing. Why do you do that? >> yeah, we want the payroll product to be complete so it includes benefits, includes kind of compliance alerts and you know, we'll try to do as much of that as possible for you and obviously going to it's AI powered so you can you know really easily set up your employees you know get alerts on things and I think the big benefit is if you're using us for banking it's all integrated in one place you know that we'll make sure the money's in the right account so that your payroll's funded and you know when you set up the employee you can set them up for corporate credit cards or reimbursements at the same time so you're fully integrated it's one place you know we'll make it a little cheaper since it's kind of a bundle product you know eventually I kind of see this bundle as like gets you know all of these things kind of interact with each other right so the money movement interacts there's kind of the data piece so like if you're using us for payroll and for books you know you don't have to go integrate these things they'll just make match each other all together and the last piece is kind of this AI piece where if you're using AI agents or using Claude or chat GPT you know it's one place and you can do kind of these broad interactions so >> I'm definitely excited to spend a lot of energy talking about that I feel like I'm in the moment right now of downloading statements and you know dumping them into Claude it's like having a much more automated system around that would be really interesting before we get into the AI of it all I mean you know always sort of the background context of Mercury is it is a bank that that is not a bank you know you obviously have banking partners but you're making progress right to becoming a bank like what is your view where are you along that journey and what what would change if you were a bank and how do you think about that >> Yeah so we applied for a bank charter last year we've got conditional approval from OCC and FDIC so sometime next year we're going to be live and serving customers through Mercury Bank NA and the idea is you know eventually all of our customers would be on Mercury Bank kind of think about it as kind of two main benefits from a customer perspective you know, number one you know, we'll have this kind of direct regulatory relationship. So, we'll have a lot more in our kind of control and we can kind of innovate much more freely when it comes to kind of adding new payment products and kind of growing with our customers. we'll definitely have a bunch of features we'll be able to enable. So, right now we don't have Zelle, which is quite a limiting thing especially on Mercury Personal.
10:01 so, we'll have Zelle. our lending products are pretty limited right now. We do working capital and obviously the corporate credit card, but we'll be able to do much broader set of lending once we're once we're a bank. >> I know I did not to make it all about my business, but there was a second where we're like, "Oh, maybe we should have some working capital." >> Yeah. >> And then you guys, you know, were like, "Oh, you know, you have to tell us about your business or whatever." I mean, I guess that will that always be the case or there is a degree to which you see sort of my business. You can you can sort of understand the money flows or will does that always involve sort of a conversation?
10:34 >> there will definitely be elements of it that we'll automate. I mean, our corporate credit card, there's no there's no conversation and we're taking some time there. >> Mercury attracts to most like not you know, my job is to be social, so it's not like we can't get on the phone with somebody, but it's funny, you know, you've attracted these customers are like, "Oh, I should be able to just do it." >> one of the reasons I started Mercury is I don't ever want to talk to people. I mean, it's so annoying if you can't just do things.
10:58 but yeah, we'll automate as much as possible. I mean, at the end of the day, I think if you're expecting like $5 million or something, you probably have to talk to someone. That's a point that's like a diligence step. but we'll automate as much as possible especially on the kind of lower end of loans. >> Will you still have banking partners or or you'll you'll just be the bank at that point? >> there will still be elements of banking partners, but it'll be more on you know, I don't know, if you're doing something with stable coins or like we're making a international wire and we're using a correspondent bank. So, they'll be kind of it won't be all the core deposits on the whole.
11:32 >> How much are you waiting with baited breath versus at this point is sort of a sure thing that you will become a bank? >> there's definitely work we have to do, but I would say at this point it's mostly execution rather than like us kind of waiting. Like we have this conditional approval and actually there's a, you know, with the approval there's a timeline that kicks off, so you're supposed to get live in a certain time period. So, there's definitely steps, right? That's not a it's not a given like we have to kind of go build it out and there's more kind of examination that happens along the way.
12:02 but it's much more like, "Hey, execute, you know, go execute these things and you will kind of get it." >> What do What is your view on I don't know, the banking as a service space right now. I mean, you've been sort of the actual like service provider or technology layer, you know, obviously with banking as a service, there's technology component, but there you know, there there have been a number of banks that have tried to be the back-end infrastructure.
12:23 You're obviously going to be try going to be full stack eventually. Like is there Yeah, why not just have a service separate banking as a service provider? >> We when we launched, we were nine people. I think realistically when you're small, like the only way you can provide, yeah, depends on what product, but the only way you can provide a lot of financial services as a fintech, is by partnering with a with a bank or banking as a service provider now.
12:50 and I really think, you know, becoming a bank was like pretty much impossible. You'd have to have raised like 30 million plus. >> Mhm. >> so, I think it's a real enabler. I think at some point, if you do something like we do, which is very much kind of a core banking service, it just makes sense to get a bank charter eventually. I think the two scenarios where you wouldn't is if you kind of stay relatively small, I think you'd continue working with with sponsor banks or if this is not a core part of your business. If this is kind of a, you know, if banking is a side like if you're most mostly a vertical SaaS player, but you want to provide a few payment or lending or banking banking features, at that point you kind of continue using a bass service.
13:32 >> All right, goodbye boring banking question. No, but I mean artificial intelligence and obviously there's this insane moment where I don't know. I'm going to Claude and ChatGPT and you know, for everything trying to understand what's going on in my life where any new situation like what what's your view on this? So obviously getting context on banking is so vital. I mean in my personal life, I'm canceling subscriptions left and right based on sort of audits from from Claude. What what do you see is sort of the vision in terms of yeah, connecting large language models and other AI to our banking services.
14:10 You know, even a year ago like this seemed like fantasy, right? Like like, you know, could you have like your agent talking to like people were talking about it, but it was very much like not being done. Whereas now it's kind of becoming reality. So I kind of think about it as like kind of three pieces. Number one is yeah, how do the agents access the data? Number two is how do you have controls over it? And number three is like what is the kind of suite you provide that the agents can do things with? And we're, you know, trying to innovate on all three of these. So you know, with number one, we obviously have an MCP, a CLI and an API and you know, depending on like what your setup is, you will integrate your agent into all of that. Separately we also have our own agent called command.
14:57 So yeah, if you want to use the Mercury app, you can directly kind of talk to us and we will generally speaking that's a little bit more fully featured. You can >> To get like insights about what's happening with my business? >> Well, with command you can do almost anything. So yeah, definitely insights. You can say like, hey, last three months show me my AWS bills and how's that changed? And I'll give you a nice graph and things like that.
15:18 >> I just say send money to >> Yeah, you You say send it say send money or like especially bulk actions. >> or I can just through text and then I'll fire. >> text but it'll you know the the final action and actually we just launched MCP actions literally today. so the final action still requires a click to like approve it which is also what MCP actions allows. So MCP actions is like basically you know if you're using Cloud Co-work or ChatGPT you can now say hey using you know you know send money using Mercury and they'll pop up a thing that says okay this is set up you want to approve it kind of thing. so that's kind of ties to the second part like you need a set of controls that makes sense. So you know we think about it as especially for bigger stuff like have a human in the loop to actually like click to send the money but it's all kind of set up for you. either send the money or do whatever kind of right action.
16:10 and for other elements so you know you can set up a gentle cards on Mercury so you can say like hey actually you know I'm fine with my agent spending $50 a day. >> And you do that through like a fake credit card a simulation. >> card yeah. It's not fake. >> It's real real real just digital it's not fake. So but basically it has like a limit of how much it can spend. >> Yeah exactly. and you can kind of customize all of that.
16:35 So yeah that's I think the second bit controls which I think matter a lot obviously with financial like much more than maybe some other types of kind of services you want to connect with AI. and then third part which I think is also really important is like you know especially if you're using several different things like all of these get epiphany key right like your connections don't work or the money yeah the data doesn't sync between everything etc. So the more you do on Mercury the I think the better it'll get. So you know we really want to make it so you go to Mercury or you go to your agent and say like hey I want to send this person a job offer and you know it'll send them a a job offer or get them a contract signed it'll set them up on payroll and benefits set them up on corporate card like you can kind of do all of these kind of actions just by this like very simple kind of prompt.
17:22 But that's only possible if you kind of have like a much fuller stack of services under the hood. >> On the controls bucket, do you does is there like a number where you're like you can't automate like an agent can't send a million dollars or is there a hard number on how much you limit an agent sending? Is it a percentage of a business or like how do you think about as sort of the banking service provider?
17:47 >> I mean I kind of think about it like self-driving cars, right? So at this at this point we're much more on the like hands on the wheel and you better be looking properly kind of thing. So right now for any at least with MCB actions I don't command like any any sort of like money movement. >> Even if it's like five bucks I can't >> Right now you still have to click on it. Unless you're using the card, the card has you know you can pre-authorize like a set of it.
18:10 >> Oh okay, but yeah is there a limit on like how big the card can be? >> I think it's up to you what you said. I don't think we have I mean there is like some we have limits per customer but I don't think there's a separate limit for agent or human. And yes but that's very much in your control like you're choosing to issue the card so you're kind of taking some responsibility over it.
18:33 But I think this will change over time like I kind of think about like right now we're treating agents as different from humans but you know kind of like a Waymo is treating the AI as the same as a human. I think over time we'll get to the level where yeah we're like hey I just hired a finance ops person and like this is the level of like controls I'm willing to give it and like you know maybe you still like a lot of businesses still have people that have like multiple approvals for like sending bills, right? Especially above like certain size so I think there'll still be that expectation but over time you kind of give it more trust as it does more for you.
19:10 >> There's a little bit more of a permanence to a human than an agent. Or it's like if you have this like, "Oh, we trust this agent this this amount, but it can get swapped, you know, it's sort of like how it operates could get swapped out." >> then like you have evals, like it's like, you know, you have you have systems that are like, "Okay, yes, you know, maybe the the models are changing and getting better, but then you you have controls of like, okay, how do you test the next model?" So, it's I mean I don't I don't think of this is like that different from humans, right? Like a like you hire a human and maybe they kind of over time they're they're they may not get better at the job or the job gets too hard for them and things like that. So, yeah, you kind of build this trust over time and have the controls in place and test them over time as well.
19:54 >> What extent are you betting on this business of one human? Or it's like, "Okay, we want to create sort of enable businesses where it's just, you know, a one-person or a small team with mostly agents." And And do you see that as like a big business to serve going forward? >> I think the future, especially at this time, is like really hard to predict, right? Like I mean, why isn't there a business of zero? Right? Like is there a scenario like that?
20:20 >> or you're going to leave me alone? Like stop asking me questions. >> So, I think it's it's an interesting element of like, okay, you know, where is the human useful and where where is the AI useful? I think I think it's like a hard time to predict it. I mean, you can definitely say like today you can definitely be a business of one and have a much bigger business or like, you know, I think it was like almost unheard of that a business of 10 would have had like 10 million or 50 million in revenue and like this we're definitely seeing startups that are like small with much bigger revenue per employee than we've ever seen.
20:54 So, yeah, in general I'm like into like, "Let's enable the future." And let's see let's see where it falls out. I suspect there's always I don't know, maybe I'm like old school. I I think there'll be an element of human creativity that like AI won't replace anytime in the next kind of let's say 5-10 years. and so there'll always be humans involved. I also think like we're in a competitive world, right? So just like when I look at Mercury, like we're moving like crazy fast right now. Like, you know, we've launched like six really big products in the last year or something like that. I would say much faster than we were kind of 2 years ago.
21:30 >> Because of AI coding >> Yeah, I think that's really been a big enabler. Yeah, 100%. But everyone else is also moving fast. Like we're just in the space so like everyone's moving fast and it's competitive and we're hiring more people than ever, right? We hired like 40-ish percent this year. We're planning to hire maybe slightly less, but like yeah, same number of people but maybe 35% or something, but we're still hiring pretty aggressively and actually like the biggest hires are in engineering because, you know, we're like we think the time is incredible and like these engineers are like, yes, super linearly enabled right now. Like you want to get as many of them as possible building cool stuff.
22:08 >> In terms of agents spending money, you know, you you sort of said it's like the fail-safe is like a person presses a button. It's like the agent can do all this. Like we should planning and I decide what it should probably be, but a human has to approve it. Do you do anything to stop customers from using agents, you know, to click buttons on their behalf, right? The real risk is like now we have all these computer use sort of agents.
22:31 >> I mean, if a person really wants to like I don't I'm not here to like, yeah, we're we're trying to protect the customers. It's like it's not we're not here to police the customer. >> Right. >> So if they're doing something if they really want to go out of the way, I'm sure they can like enable like fully automated payments. I really >> But don't There's going to be like I guess we're we're clearly I think I was talking to John Dean, the town CEO who's a former Plaid CTO. I mean, there's there's clearly going to be like lawsuits every direction at some point. Somebody's going to spend money that they regret even if it was like their fault and they're going to like sue everybody involved and say an agent was involved like you should have stopped it, right? Like do you worry about that as a company? Is that a risk?
23:12 >> Maybe, but I'm I'm guessing the same thing happened when internet banking happened or whatever, right? It was like before it was like, oh, you had to go into a bank branch and you had to authorize this payment and now you can like be hacked on the internet and yeah, there's laws in place. There's like things you can dispute. actually like the banking system has dealt with this a lot, right? There's like there's fraud and there's lots of fraud controls and things like that.
23:32 then there's like you doing something and then, you know, when when is it a merchant's fault? When it's when is it your your accountability? So there's a lot of actually rules already in place. So it depends on like the exact situation probably. a lot of the benefit of banking is a lot of stuff is reversible in a way that you know, when when people look yeah, talk about crypto or whatever, like one of the I think annoying things about crypto is none of this reversible. So, you know, if an if your agent goes rogue and sends a bunch of Bitcoin or whatever, there's just no way to get it back. whereas yeah, if your agent goes rogue and like goes and does big e-commerce shop, you can normally dispute it and I mean you might have to send the stuff back, but you'll get the money back, right? Like so the money is going somewhere and like, you know, what what is like what's this exact kind of circumstance that the agent's spending all this money?
24:24 I feel I feel like a lot of this stuff like ends up being like this theoretical like what if what if. I think for most people, they're not doing that many payments in their business that they needed to be like go agent go rip it and do like 500 a day kind of thing. Like they can approve it or >> with like Instinct where I'll send emails that people didn't expect. >> Yeah. >> And I mean actually even email, right?
24:44 But but if you send that out, it's like what once payments you know, are connected >> I've been using Instinct and used to buy a lot of things for me and it's I don't mind clicking that approve button, right? Like right the end it's like, okay, you know, like I just had my daughter and she went kind of fitting room. She wanted like all of The Beatles albums on vinyl. And I was like go find it and it's like found it. I didn't look at the website.
25:06 I was like okay, you know, I was like yeah, does this website have good reviews? I was like yes. I was like okay, pay for it. And then I just click the button at the end. It was fine. I think like actually all the other things that are annoying bit like clicking at the end and saying like yes, do it is not that bad and I think that's like probably the use case that most people will have with agents. I think there is another set of companies that are trying to automate like things at scale, right? Like the you know, maybe they're sending a ton of invoices, right? Like they have lots of customers. They want to like sum it all up, send it at the end. So, you know, we have an API where you can send a ton of invoices.
25:41 Same with payments. So, there'll be a set of kind of companies that actually using our infrastructure to do things at scale with customers. And then we have APIs and things like that to kind of help them with that. But again, like I don't know like I don't know if the AI is like non-deterministically creating a bunch of invoices. It's probably actually writing some code that's creating the invoices from their database. And obviously it's like up to them to kind of go check that code.
26:07 >> Are you seeing more fraud? Does the invoice thing specifically, I feel like it's much easier for, you know, fake companies to send you an invoice and try and get you just to pay them >> Yeah. >> money on accident? >> I think the biggest and like probably worrying thing about AI is like it's made fishing at scale like much more possible, right? Like you can be this you familiar with social engineering? It's like someone can message you and pretend to be someone, but now they can do it like at massive scale, right? They can call you and it can be an AI bot and you don't realize it's AI.
26:39 We definitely seeing that. Like I think there's a ton more fishing ton more malware happening especially this year. >> human sounding voices call you or have you you actually personally >> pick up my call. I I continuous texts from people and I do get continuous calls, but yeah, at this point my phone is like pretty unusable. I get like 10 plus spam calls a day. >> In terms of, you know, running your own business, what I don't know what what is your AI stack, or where do you get the most value out of AI? Is it sort of like the strategic thinking? Is it more as an assistant? You know, where where do you see value, and where are you like, "I'm super relying on humans for that."
27:14 >> you know, the funny thing about being a CEO of a scaled-up companies I have a lot of humans that do things already, right? Like I already have an assistant. I already have a chief of staff and things like that. you know, we are using a ton of AI internally for like a bunch of things, right? Like obviously engineering has a ton of it. I think the interesting thing is there's obviously like the agents that coding and things like that, but we've actually also invested a lot in agents like helping with the quality of things. So, like we have this one agent that like every time someone does a new database query or makes a change to database query, they'll like test it in production to see like, you know, is this a non-performing database query?
27:52 And they'll like flag that before it gets merged. So, there's a there's a lot of these things that you can do that, you know, create quality. Cuz I think again, a year ago everyone was like, "Oh, agents are going to create a bunch of sloppy code." Which I think is definitely like one way to look at it, but I think there's also agents can do a lot more to create >> code agent. >> I mean, it really is like that. You're like, "Hey, check the database queries.
28:16 Like make sure that, you know, these patterns of like sloppy code, whatever it is, right?" Like you can actually you know, people say sloppy code, but you can kind of understand like what makes code sloppy, right? Like there's this kind of idea of like interconnectivity in code. I'm an engineer, so this might get boring for people, but like you know, a piece of code that calls this way and then calls back that way is like generally worse than like something that's like kind of like directionally separate. But you can test for all of these things when you do new code, and you can say like, "Hey, this is like, you know, badly modularized code. You least you think about changing it like this. Actually, an agent can go even further and go fix it for you, right? So, so, lots of stuff in engineering. and then we definitely like yeah, every team is using it slightly differently. one of the things that's kind of interesting, at least I've noticed within Mercury, is there's a yeah, there's a lot of these kind of custom tools for each it AI tools for each department. We've ended up building a lot of it ourselves. so, actually we moved off Fin for our customer support and we built our own kind of internal customer support. And it's all kind of built into our command agent. So, you can be like, "Hey, send this payment."
29:23 Or if you're like, "Why is my payment not sent yet?" Or like, "What Yeah, why is this taking so many days?" Like the AI will automatically respond. And that's like a layer one of the >> Was it a cost decision? You guys got acquired by Salesforce, right? Yeah. >> It wasn't a cost decision. We realized like to make our agent really good, it needs to do things, right? Like it was like for a while we were using Fin and it couldn't do anything. It would just be like it would look at our support center and it would like try to automate like answering questions. But the next level of automation was like, "Let's go do things." Like, "Let's go, you know, either make the payment or check like go check in the database and like go look deeper." and all of these things that can theory you could use some third party to do it, but in practice like so much of the work like the intelligence layer is on the front on the front end models. And if the integration layer is like built being I ask like why have a middle person? Like then you have to kind of communicate with the middle person to say like, "Oh, this is how you like go do X or Y." We really wanted the smooth experience between like so the command agent well, yeah, you can make a payment, you can ask questions that are kind of read only. If you get stuck on something, actually elevates to a human.
30:29 so, you can say either you ask a question that the agent can't respond or you say like, "Hey, can can I get like a human support?" Or you know, so, it actually goes all the way from like automate an action, have an AI try to do things for you, go all the way to a human, all in like one place. So, I think that's like a very smooth experience. I don't know if many other people have done it, but that's kind of how I think about it. Agents at Mercury, it's not just like, you know, it it should fall back all the way to a human.
30:55 So, that you can like pretty much get any any questions answered. It shouldn't be like a separate kind of task to go talk to a human. >> The 1 billion annualized revenue milestone. >> Yeah. >> I mean, you're you're being very precise that this is not like annual recurring revenue, obviously. You're sort of a banking business, but we're in the the era of, you know, sort of sloppy financials. How sticky is like the revenue? Or like what just what can you tell us? You you talked a little bit about sort of the product lines, but yeah, what I don't know what >> >> And also, you invest in a lot of startups, so I'm after that curious about your observations about startup financials generally.
31:31 >> I mean, our revenue keeps going up every year and like, you know, it's I guess it's sticky in that sense. It's growing in that they stay, you know. Yeah, so, our revenue is kind of based on like deposits and payment volume and, you know, people using Mercury. So, at the end of the day, it's kind of like secondary it's like a it's a second order effect of people using Mercury more. So, we also grew our customers to 450,000, which is I think it's up more than 50% like 60% or something like that. So, we're seeing this kind of acceleration of usage.
32:03 and yeah, again, we're seeing more and more entrepreneurs, but actually in the last kind of couple years, we've gone beyond tech. So, yeah, we obviously started as like tech startups and like we were the only mainly for them. And now about one in three kind of startups use us. we've grown to kind of e-commerce, professional services, and really like a broader set of kind of SMBs. and that's where we're seeing a lot a lot of growth.
32:26 I think yeah, one thing before I started Mercury and I was like very curious about it is like, you know, obviously, because we serve small businesses, many of them fail. and that's actually our main churn. Like our main churn is people start a business and then they don't >> They don't see there's really no agents for this. Like, "Oh man, you're destroying your business. It's going to die if you keep doing this. Like, we'll save you." I I guess nobody wants a bank that's so prescriptive.
32:52 >> we really do want to help people succeed, but also like that's part of being a business, right? You You take a risk and you want to start something and maybe you it doesn't work out, right? Like I don't I think like zero 100% success would actually mean that we're not doing and we're not experimenting as like humanity. but so I wanted to say like, yeah, every cohort, like every month we get a new set of businesses and we kind of look at like how do they grow over time, right?
33:16 Which kind of ties to your question of like, you know, is this recurring or And we've generally found like every the set of businesses fail, right? Whatever, 20% or so fail within a within a year, but the ones that succeed kind of tend to scale both spend and deposits with us over time. so yeah, every month actually, yeah, every month becomes like an additive thing over time, which which is I guess goes back to kind of is this recurring or not. I do find this kind of pretty funny that everyone calls all their revenue ARR now and like, you know, I grew up in the day where like it's no ARR unless it's like contracted revenue for like 3 years and blah blah blah. So, anyway, I think that's like all all thrown away now. So, I always say annualized revenue, it's nice and precise, but yeah, tends to, yeah, we we have a positive NRR and it tends to grow over time.
34:04 >> The credit card business, like how key is that? I mean, obviously, I don't know, Ramp is a juggernaut, has a huge valuation. I think was it today valued at like $60 billion? Are you You see Ramp as a direct competitor? How how core is that to your business? >> yeah, we launched corporate credit cards kind of end of 2022. It's been going really nicely, and yeah, most of our customers use the Mercury corporate card. yeah, most of the time like we're not competing against ramp, we have really Yeah, when we first started, we were really competing against SVB. That was our main main competitor. Nowadays, it's mostly JPMC and a few other kind of incumbent banks. So, yeah, because we're we're really competing for like in people's heads, like I need a bank account, who should I sign up to?
34:52 At that point, like it's then a natural thing to say, "Okay, you know, I also need to issue corporate credit cards." And like they're all kind of part of this platform and, you know, we'll attach >> more products over time. >> to acquire the customer at the banking stage and then >> Yeah, actually that's the only way to use any of our other products. So, you have to use bank to get the books or payroll. >> Oh, interesting.
35:09 >> Or corporate credit card, etc. >> Well, doesn't that limit the sort of ambition of the corporate credit card? >> not saying we'll have that forever, but it also simplifies things in some ways, right? Like we already got you on board, the money is there, it's not Yeah, like our underwriting for the corporate credit card is very simple. Like it's like you have money in your bank account, like it's tied in, like it's, you know, we we don't take many losses there. So, it's a simplification, it's a focus thing, like we really make that like a very smooth experience for our customers.
35:38 Yeah, eventually many of these will be kind of stand-alone products, but right now it's very much tied to like everything comes in a bundle, basically, with banking being being the first kind of starting point. >> You invest in a lot of startups, you're an investor in Etch, which is doing very well, and Decagon, and sure, many others. Are you doing any fintech investments right now? Or like, what you know, how do you feel building this environment where it feels like it's been a fintech winter?
36:05 >> Oh, is there a fintech winter? >> know, is there? You know better than me. >> Every year we change our mind about these things. I find it I find the whole thing kind of funny. You know, I kind of think about my investing and like, you know, Mercury is like let's let's go after big markets where like things are clearly broken. And if you Yeah, think about like big markets, right? Like fintech is a huge market, defense, energy, obviously intelligence, which is like AI goes after. So, I think there's always opportunities in big markets. what Yeah, what did happen is 2021 had this like insane wave of fintech investing. I think 20 to 25% of all investing was in fintech.
36:44 And then, you know, from that have emerged obviously companies like Mercury that have reached reached a certain scale and like become breakout. Obviously, a lot of companies didn't work out and then kind of I I'd say like it kind of like explored a lot of the idea space. And now, you know, now like these obvious kind of fintech ideas were kind of some extent burnt out. This new wave is like AI obviously. And I would say like even a year ago it was, you know, I guess we had like some stablecoins stuff, but AI wasn't a big enabler to fintech.
37:17 Actually, like fintech is one of these like relatively few things that's not like massively disrupted by AI. you know, I think it's not it's not helped that much, but it's not hurt that much by AI in the way that like obviously like B2B SaaS has been kind of >> so, you all can rewrite your application you know, all right. >> so, and yeah, we were just having this conversation literally last week where like someone was like, "Hey, I'd rather just write this whole B2B app than go through Mercury procurement and get like, you know, negotiate a contract and get everything signed." Like it's just become much easier to write those apps, but yeah, it's it's still pretty hard to build a fintech company.
37:55 >> They're like, "Our procurement is so brutal." They're like they'd rather just build it themselves. >> like you can build really complicated things and like especially if it's internal use, like you don't need it you know, you don't need every feature in the world. You just need the features that you really need, right? So, yeah, so I think that's like been the case with fintech. I think it's going to change to some extent, but yeah, I I feel like the incumbent fintech players like Mercury are kind of yeah, very much because we're I guess new enough and we're still kind of innovating. We're jumping on this kind of building AI features.
38:27 But there'll be some cool ideas. I mean, I'm invested in this company called Natural that's building a kind of payment rails for for agent-to-agent kind of fintech. Yeah, there's there'll definitely be innovation. you know, I've always invested in a pretty broad way. Like I I mostly invest because it's kind of fun to learn new things and help entrepreneurs. So, you know, I don't want to invest in just things I know I'm good at. You know, I have like I'm invested in Stoke Space and they do like rockets that go to space. So, I I think like having like a broad set of investments is like what I do. but yeah, fintech is definitely a part of it.
39:02 >> You know, there there's a strong set of investors who are investing betting on AI native services. You know, it's like think about a Crosby or it's like an AI native law firm. There are there are obviously then sort of the roll-up efforts where it's we're going to buy a bunch of you know, businesses doing X, Y, and Z and turn them into AI businesses. are you seeing that in your world? Like how much I guess do you face competition from roll-ups of regional banks that are going to become AI native >> that's actually part of why we haven't seen as much AI in fintech. That like fintech is not labor services. Hm. Look, so actually one place where there has been a lot a lot of AI is accounting, right? Because there is a lot of accounting and bookkeeping that is manual.
39:47 I mean, I'm invested in several and obviously we have Mercury Books which has like a big kind of AI component as well. but a lot of fintech is like payments and banking and and and even lending which isn't like really a labor-intensive thing. There are pieces of it like I don't know, mortgage brokerage where I've seen like a few people, wealth advisors another one where there's been like a bunch of AI stuff. but yeah, the core core like banking payment lending is not really like human services.
40:15 Yeah, wherever you think about human services are like I don't know, like 50% of the cost or something like that. That's where like you know, yeah, with lawyers like human services are 90 95%. >> thinking about getting into categories where you wouldn't have because of the human element like wealth advising or something like that? >> we haven't yet, but yeah, we are definitely like exploring and I would say we will probably inch our way there, right? Like if you keep making command better and better, yeah.
40:44 They can already give you like some some help with like your finances, but yeah, maybe we'll give you advice on like certain things. Yeah, wealth advisory being completely AI is like still kind of a question mark on like whether you can you can actually like do an AI fueled trade for a customer. I don't think the the laws are kind of worked out with that. So, you probably still need a human in the loop on the advisory step even.
41:08 But yeah, but there's definitely like things we'll kind of step our way into, but it's not I wouldn't say like we've launched anything or like a cool thing we're doing right now. >> Right. What's your bubble read at the moment? Or we mentioned you know, the fintech run-up in 2020 2021. There's a lot of anxiety like this week about you know, Anthropic's financials. I don't know how much how much as a startup founder do you worry about that and like I don't know, you're just a plugged-in guy. Like what's your read?
41:35 >> tricky because like it really does feel like 2021 nowadays, you know? I don't know. >> If that is a bubble yes, then yeah. >> But like you know, are we in like December 2020 or are we in December 2021? That kind of like I think both things can be true that like AI is transformational, it's going to be huge and like you know, going to fuel like a massive GDP growth etc., but we could also be in a bubble, right? Like I think the I see a lot of things where I'm like, oh, like I'm not sure about that, right?
42:04 Like there's I don't know. I'm going to pick on Y Combinator, but yeah, there's like if you you a pulse in Y Combinator, you're worth like 50 million now. And like if you have any traction at all, you're worth 80 to 100 million. and that's, you know, that's like super early stage companies that are just like and some of them will be successful, but the problem is like all of them can't be successful. Like you know, they've got like we've got such a built-in like so many of them have to be successful to justify like 100 companies with 50 million plus valuations every 3 months, right?
42:35 and I'm not picking on Y Combinator. This is happening like everywhere in the industry where it's like, you know, you had 5 million revenue, that's like a billion-dollar valuation or something like that. So I think there's obviously something real and I think we'll see it, but I also think like we're over-invested and like we do in every cycle. I think on the macro big markets level, it's mostly about like have we overbuilt data center capacity? Like I think that's what matters to Nvidia and yeah, Amazon, like all of these like big companies. Actually, I'm not sold that we've overbuilt data center capacity yet. Like I just feel like I mean, even if you look at Mercury, like we're demanding more and more tokens. I think every company is demanding more and more tokens. I still think we're like relatively early in how we use AI to build things and like what level of penetration we have in like big companies and enterprises and like across across the kind of economy.
43:28 So I'm personally skeptical that we like we have overbuilt data centers yet. I mean, I'm not saying it's not going to happen, but yeah, this year is definitely not happening. I feel like next year we're probably still going to use most of the capacity we built. >> I mean, a lot hinges on OpenAI and Anthropic can continue to consume a lot of the data center capacity, right? They're big users. >> I don't I mean, it's not it's I mean, at least when it comes to data center capacity, it's actually like kind of fungible, right? Like we had this moment where SpaceX AI, I don't know what the all the the >> changing the The main company is still SpaceX.
44:02 >> Yeah. Anyway, they overbuilt data center capacity, right? But then they just sold it to Anthropic, right? So it just actually the main thing that matters is is there enough demand out there and like it doesn't matter that much if it's going to Anthropic or OpenAI or Google or like some open source model, right? Tokens are tokens, data centers are data centers, and like we'll as long as there's enough demand out there it will be sent to the right people. Like individual companies might be hurt there, right? Like maybe Anthropic doesn't have as good a IPO or something, but I think the thing that matters the most to the economy is whether the data centers are used or not because there's so much data and so many companies involved in that.
44:38 >> If there was a big overbuild out, I assume it would be might be bad for you on a multiple basis, but it'd be good for you on a now all these tokens are super cheap because >> Like Mercury? >> I guess you don't want your businesses to do poorly. You're I guess you care about your business. >> data center buildout, like then tokens will get cheaper, we'll get more applications. I think it's probably good for Mercury. I mean, we're not tied to the like we're not tied to like what's happening in the macro market. I mean, maybe I guess multiples will fall and I was going to finish my point. I do think that in the private sector there's places where we have overfunded companies and you know, if you it's very hard to like have any company that doesn't get like 10 competitors in it, right? Like it used to be the case that like if one VC backed a company in a space, maybe a second VC would back a competitor and that and then everyone else would say, "Oh, we funded this space, right?" Like now it's like we'll get 10 companies and or maybe 20 cuz like especially in these big spaces. So there'll still be winners, but like, you know, that has to play out, right? Like probably all 10 of them aren't going to win. Two of them will win. So yeah, it's going to be a messy time if there's a downturn in the private kind of market. Similar to 2021.
45:46 >> The the counter to that is just like some of these acquisitions have been so unexpected. It's like, "Okay, if the fifth or the 10th place, you know, model company can get acquired, you know, it's like Poolside can have this insane deal with Nvidia, Hugging face. You know, then investor, you know, in in cursor, you know, obviously they were doing really well. They were strong in coding, but it was a very competitive space and there was some weakness and then a sort of historic M&A. You know, I I agree with what you're saying, but >> like even these acquisitions are like, you know, companies that started 3 years ago in an environment that was 3 years ago. Like companies starting now are like, you know, we have to see what the environment is like 3 years from now or 10 years from now.
46:28 and it just feels like it's overdone. I might be wrong. and they'll as I said, there'll be tons of winners, but there'll also be there has to be a ton of losers. Like I just don't see how we can't have that and at some point that like comes to roost and in this market like when everyone's kind of so frothy, that's not coming to roost kind of thing. Eventually it has to. >> So you pausing investments right now or >> No, I'm just I I think it's good to like, you know, the funny thing about I do find investors kind of funny because like, you know, 2022, 2023, 2024, that like a lot of investors were paused or doing so much less. And I was like that was the best time to invest, right? And then and then at the peak markets like people invest like crazy, you know. I've always tried to be like fairly consistent. I'll do something like 20 investments a year and whatever the market is, I'll do 20-ish investments a year.
47:18 >> Did Mercury experience a token maxing phase? Did And did you have a pullback on how much you let sort of every rank and file employee use Yeah, we coding or the these agents and sort of LLMs? >> I mean, maybe we just have like more sensible people, but I don't know. We've never had like a I guess even now we're like pretty free with like people should use it and you know, I'm actually acting CTO right now as well. So every now and then like, you know, we have a month and I'm like, oh, you know, who were the big token users? And I go talk to them and say like, hey, what did you do? and some of them are like using us for like, you know, they did they built something big or like they were doing a bunch of things and I'm like, oh, that's kind of cool that you did this." And some of them are like, "Oh, I was trying out Fable in fast mode and like it's just like you know, it's 3x more expensive than I thought and I'm not going to do that this month." So, there's definitely like >> Does that mean for the company like are these big bills?
48:11 >> No, it's not too crazy. So, yeah, we're very profitable. We've been profitable for 4 years. So, it gives us like I guess some luxury to explore. We're not trying to like kind of and I I would say we're really still in exploration phase. Like if we everything's just moving so fast. So, if you normal but I would say like that you know, the models weren't that good until like December last year. So, like we're less than a year of like especially in engineering at least like we're less than a year of like really really applying the this AI to I guess to people token maxing actually as well but I'd say it got really good this year. So, and then now we have agents and like you know, we we have our own kind of internal agent that like does a lot of coding in Mercury.
48:51 I just feel like every month we're like learning new things. So, I would say right now we're much more still like I wouldn't call it token maxing but like very much like yeah, go use it. Go try to break it. Let's see where the limits are and like what's possible and every month that changes. So, I think it's like a fun phase. Like I feel like either direction like token maxing or token meaning or like it's like not the right time right now. Like I think you got to go like case by case. Like does this make sense? Did you you know, did we drive more productivity or something like that?
49:23 Yeah, I think we're like doing like It's like some insane number like 4x more lines of code per engineer per like per month or something compared to like a few years ago. So, >> I think my last question, you know, if you think of Mercury on this like big time line, you know, you're hopefully soon becoming a real bank. Is the goal to build like an institution? Like do you think you know, in a decade the goal is like we have like the best big office in New York City and we're are this sort of you know, it was so much of being a bank is like it's like political and influence and you're part of like at a certain scale like how you know, I feel like Jamie Diamond projects this like I'm another pillar of like society and America's institution. Like is that an aspiration for Mercury or do you or is this a new thing that sort of looks a different way and has a different sort of sense about it?
50:15 >> I mean you make it sound bad. institution >> Wait, wait, I mean does it sound good or bad? I don't know. >> I thought you made it sound bad. >> I don't know. I think it's neutral. It's like some of you a lot of you would want to be Jamie Diamond. I don't know. >> Oh, it's depends. >> I I don't I want to help customers. Like I I know that sounds kind of like a bit cliché maybe, but like it's not that interesting to me to be like an institution for the sake of it. I think it's like yeah. How many people can we help with like building a business, being an entrepreneur? Like I think all of that stuff is really hard and if we can like make it a little little easier for people and save them money and time, I would like that.
50:52 and I want to have fun doing it. If that ends up being and I want to build something that lasts like I want you know, I want to die one day and go like, "Wow, like you know, Mercury's still having an impact on the world." So, that's how I think about it and you know, maybe along the way we become an institution to do that, but yeah, that to me is not the end goal of like having power or control or whatever comes with being an institution.
51:16 but I guess like the positive side of being an institution is that you created like the trusted kind of rock that people kind of rely on and I hope we do that, right? Like I think you know, when we first started like you know, as I was saying like it was crazy to me that someone would put like a million dollars on Mercury. Now we have like several customers of 500 million plus on Mercury, so So, you know, I hope like people keep continuing to trust us and like we kind of live up to that to that trust. so yeah, I'll take that part of being an institution, sure.
51:46 >> Immad, thank you so much for joining the podcast. >> Yeah, thanks for having me. This is fun. >> That's our episode. This is the Newcomer podcast. Thanks for listening. Please like, comment, subscribe. Go find us on our Substack at newcomer.co. If you're hungry for more podcasting, I also have a talk show called the Cerebral Valley Show. We love your comments. Thanks for all your support and see you next week.
Summary
- Mercury has grown to over $1 billion in annualized revenue, driven by a significant increase in customer usage and deposits.
- The company primarily serves startups and small businesses, with a focus on simplifying banking processes and enhancing user experience.
- Immad emphasizes the importance of trust in fintech, noting that many customers have transferred large sums to Mercury shortly after signing up.
- Mercury plans to launch new products, including payroll and enhanced banking features, as they transition towards becoming a fully chartered bank.
- The integration of AI is a key focus, with tools being developed to automate tasks and improve efficiency for users.
- Immad acknowledges the competitive landscape of fintech but believes that Mercury's unique offerings and customer-centric approach will help it stand out.
- He expresses a desire for Mercury to be a lasting institution that helps entrepreneurs succeed, rather than simply aiming for power or influence in the banking sector.
Questions Answered
How can fintech companies gain trust from customers?
The speaker reflects on initial skepticism about fintech's ability to store money securely, but notes that trust can be built quickly, as evidenced by a significant deposit made shortly after launching their service.
How does Mercury plan to balance automation and customer interaction?
Mercury aims to automate as many processes as possible, especially for lower loan amounts, while acknowledging that larger transactions may still require human interaction.
How is AI impacting Mercury's growth and operations?
AI has accelerated product development and hiring at Mercury, enabling faster innovation and a competitive edge in the fintech space.
What trends are observed in startup financials and customer retention?
Mercury's revenue is growing as more startups use their services, but the main challenge is customer churn due to business failures, particularly among new startups.
Are we in a bubble regarding startup valuations?
While AI is transformative, there are concerns about overvaluation in the startup ecosystem, with many companies unable to sustain high valuations.