In this special episode, we feature two interviews recorded live during Newcomer’s Breaking the Bank Summit, a financial technology summit held this week in San Francisco.
We’re including two of the most dynamic discussions here, beginning with Gabriel Stengel of Rogo and Jeff Seibert of Digits, followed by an interview with Josh Reeves, CEO of Gusto.
The episode kicks off with a breakdown of the event, highlighting the key debates that emerged between Rogo and Digits around the trustworthiness of LLMs in fintech, as well as Reeves’ perspective on the service intensive business and going shoeless in the office.
After getting our hosts’ reactions, we dive in to live-recorded audio from the event.
For a full selection of discussions from the summit, including video of each talk, visit the Newcomer Youtube at youtube.com/@newcomerpod
Timecodes
00:00 - Intro
09:43 - Rogo + Digits Discussion
28:13 - Josh Reeves, Gusto Interview
00:00:00
Everybody, welcome to your episode of Newcomer.
00:00:03
Tom Doton here, joined by Madeline Renbarger and Eric
00:00:06
Newcomer of Newcomer. We're all in San Francisco right
00:00:09
now, but we are not all in the same place, actually.
00:00:12
Madeline and I are holed up at our friend Volley's offices,
00:00:16
which is as close to newcomer HQ in San Francisco as you can get.
00:00:20
I would say this is the Newcomer Satellite office.
00:00:23
Absolutely. You gotta come in.
00:00:24
We can't all be in the same place for security reasons.
00:00:27
We don't want all newcomer employees together.
00:00:29
It was dangerous enough this week.
00:00:31
Yeah, well, right. Good transition to the reason
00:00:34
we're here, which is that we are fresh off the Breaking the Bank
00:00:37
Summit, the Fintech Summit annual thing.
00:00:40
Although what is it the second year you've done it?
00:00:42
Well, the first year we called it the Newcomer Banking Summit,
00:00:45
and we realized fintech was much more exciting than banking.
00:00:48
So in some ways this was the first year and in some ways it
00:00:51
was the second. You know, that was a little
00:00:53
awkward on stage. It's like people are coming
00:00:55
back, but this is the first time.
00:00:57
But yeah, this was, yeah, our first pure Fintech summit with
00:01:00
some returning faces. Right, there were some returning
00:01:04
returning faces, like Jackie Reesus.
00:01:05
Yeah, Matt Harris from Bain Capital Ventures came back for
00:01:08
presentation round two. Big hit of both events I would
00:01:12
say. All in all, it was a lot of
00:01:13
people across the fintech world. It was fun kind of getting to
00:01:18
talk to them in between sessions at the after party as well,
00:01:21
which was exclusive. We try not to talk about the
00:01:25
after party, but. Don't tell them I.
00:01:27
You know I ran into like. The guy who'd gotten viral
00:01:29
because he was trying to have people take out debt to make
00:01:33
investment accounts, I don't know if you saw that the other
00:01:35
day. I think it was like Basic
00:01:36
Capital, who I had no idea was even going to be there there.
00:01:38
So that's what I love about these events.
00:01:40
You know, we try to get great founders there.
00:01:42
And then I'm like, Oh yeah, I've I've heard of your company.
00:01:45
So, yeah, it was a good crowd. And then excited about what we
00:01:47
have to share about what happened on stage.
00:01:50
Conversations ranged from a ton of different topics around
00:01:53
fintech. I mean, the premise was, you
00:01:54
know, fintech is back, baby. And then, you know, we had to go
00:01:57
on stage and really ask, is it back?
00:01:59
But stable coins certainly are back.
00:02:02
The big resonating theme of the day was everyone is super pumped
00:02:06
about stable coins. Probably.
00:02:07
Talking about stable coins is back.
00:02:09
Tripe a done, you know a billion dollar acquisition, billion plus
00:02:12
dollar acquisition of Bridge. We had the Bridge CEO as the
00:02:15
second person on stage. Jackie Reesus at Lead Bank is a
00:02:20
bank that's really backed stable coin fintech companies.
00:02:23
So we definitely open the day strong with people who are
00:02:26
excited about. It, I mean, later in the day,
00:02:27
too, we had, you know, Eric Gleiman from Ramp who has
00:02:30
launched with Stripe and Bridge for a staple coin card.
00:02:34
Yeah. Yeah.
00:02:35
I mean, in the background of all this, you know, Tether is making
00:02:37
more money than per employee that almost like any company in
00:02:40
history off of stable coins. So everybody sees that there's
00:02:45
like a ton of money to be made. And of course, Congress was
00:02:48
passing legislation right as we were holding the event to make
00:02:52
stable coins much more legal to do or clearly legal.
00:02:55
So yeah, that I think that was theme A.
00:02:58
We've decided the selects for this episode will be from Theme
00:03:02
B, which I thought was equally interesting.
00:03:05
Yeah, theme BI mean there was a big debate that I feel like is
00:03:08
still kind of unsettled, which I find interesting.
00:03:10
So we'll hear different perspectives on it later.
00:03:12
But around, you know, the idea of AI enabling startup founders
00:03:17
to go after service businesses, you know, like accounting, legal
00:03:20
tech services, but you know, accounting, especially since
00:03:23
this was a fintech summit and where the merits are around, you
00:03:26
know, fully pitching your startup as a service business
00:03:30
that can be fully enabled with AI, or rather, you know,
00:03:32
sticking to software and you know, being the software
00:03:34
provider for these businesses. Matt Harris at Bain Capital
00:03:37
Ventures gave a presentation which I think captured some of
00:03:40
the venture capitalists sentiment well, which is
00:03:43
financial services businesses are enormous.
00:03:46
You know, you think about accounting or broader services
00:03:50
businesses like law. There are many, many, many
00:03:53
billions of dollars of money to be made there.
00:03:56
But of course, they're human intensive businesses, the kind
00:03:59
that tech companies traditionally shy away from.
00:04:02
But the argument is that now, thanks to large language models,
00:04:05
start up founders should go after those categories.
00:04:08
And Harris was arguing basically, definitely good for
00:04:11
start up founders, maybe not as good for venture capitalists
00:04:15
because they're still gonna be, you know, you might out operate
00:04:18
a law firm, but you're not necessarily going to build, you
00:04:20
know, a Facebook. So that was Harris's argument.
00:04:24
And then as you're going to listen in these conversations, I
00:04:27
put that question to Digit CEO Jeff Siebert and Rogo CEO
00:04:31
Gabriel Stengel in our first conversation.
00:04:34
And then I talked with Josh Reeves, the CEO of Gusso,
00:04:38
probably one of the most experienced founders we had on
00:04:40
stage and asked them, you know, do you really want to run a
00:04:44
services business? Jeff at Digits had really stood
00:04:48
up an accounting practice to build their QuickBooks
00:04:51
competitor. And then Josh Augusto certainly
00:04:55
has built some of these services businesses.
00:04:58
But again, I think to sort of see if they can build software.
00:05:01
So where did you guys net out in terms of the argument?
00:05:05
Matt Harris, back to his presentation.
00:05:06
He ended very strong in an interesting way by basically
00:05:09
saying I'm tired of nibbling around the edges of fintech.
00:05:12
You know, this is a $33 trillion opportunity, which I guess means
00:05:16
all banking around the world. Well, it's also including
00:05:18
accounting, it's including accounts receivable, it's
00:05:21
including, you know, health insurance brokerages, as much
00:05:24
money as possible that you know, any money industry I.
00:05:27
Don't know if you guys heard this, but it elicited a woo.
00:05:29
Yeah, from Jackie Reese's. I at least heard it.
00:05:32
She was right in front of me while he was she also.
00:05:33
Sent me an e-mail saying she loved that quote too.
00:05:35
So yeah, she clearly agreed with that idea.
00:05:38
Well, who wouldn't want to be part of a $33 trillion
00:05:40
opportunity? But but I think that kind of
00:05:42
speaks to the challenge that fintech has had, which is like,
00:05:45
is this just some sort of, you know, skimming of, of businesses
00:05:48
that the big banks and other financial institutions aren't
00:05:51
already taking? Or is it like truly disruptive,
00:05:53
right? Is fintech, Venmo transfers,
00:05:57
weird loans, helping people, you know, finance their burritos?
00:06:01
Or is it, you know, the core of the American economy that big
00:06:05
services companies have been able to deliver?
00:06:07
Or a bunch of small ones, Honestly, a lot of small
00:06:10
businesses doing accounting and law and all these things that
00:06:13
they want to tackle. Right.
00:06:15
And like bringing it back to like large language models, like
00:06:17
is this the entry point to this larger business, to this larger
00:06:21
opportunity or you know, are we still going to be about these
00:06:24
kind of marginal disruptive plays that can build real
00:06:28
businesses like Klarna or a firm, but aren't, you know, I
00:06:32
don't think are really necessarily taking 33 trillion
00:06:34
as the opportunity there. I will say, I mean the success
00:06:38
of Rogo so far, you know, they're basically building an AI
00:06:42
agent that can automate the work of an investment making analyst,
00:06:45
which is very specific, but has taken off, especially in an
00:06:49
industry where, you know, unlike accounting, you can get most of
00:06:52
the way there and the people will say, you know, this is good
00:06:55
enough, we got this done. And so that is one place where I
00:06:59
can see the service automation be very promising.
00:07:01
But that being said, it is a pretty specific service that
00:07:04
they're automating there. So we'll see how that expands
00:07:06
accounting. Of course you know it's tricky
00:07:08
'cause you don't want to mess that up.
00:07:10
I, I think that's a key point, right, Gabe with Rogo is, you
00:07:15
know, they're trying to replace like the Goldman Sachs analyst
00:07:17
or even just, you know, give you a sort of cheap analyst that
00:07:20
that has a rundown on companies you care about.
00:07:23
You know, you imagine the bankers already like going on
00:07:25
ChatGPT and be like quick before this client meeting.
00:07:27
Tell me the download on this random like oil and gas company
00:07:30
I've never heard of. And so there, it's like Rogo.
00:07:33
Doesn't need to be 100% crack. They just need to be better than
00:07:36
sort of the lazy version of this.
00:07:39
And obviously they have, you know, aspirations to be much
00:07:42
higher quality than that, but they don't need to be 100%
00:07:45
correct. They need to help you get pretty
00:07:47
broad coverage and be able to move quicker than you can right
00:07:50
now. Whereas accounting, as Jeff was
00:07:52
saying, you know, you need to get it right and you know, a lot
00:07:55
of what they would need language models for is sort of the boring
00:07:59
business of tagging all your expenses, which is somehow is
00:08:03
still a task that's like cannot be fully solved by technology
00:08:08
right now. And so he talks about how they
00:08:10
use machine learning for a lot of the tagging because he finds
00:08:13
it much more accurate. And then when they really can't
00:08:17
figure it out, they sort of see if a language model can sort of
00:08:20
crack it, but they're not really relying on it.
00:08:22
So Madeline, I think exactly like you're saying, it really
00:08:26
depends what business you're in, how much you trust language
00:08:29
models that aren't totally consistent.
00:08:31
You know, Jeff made the point like, well, ultimately you can't
00:08:35
sue the AI. So if you mess up my accounting
00:08:37
books, like, you'll sue the CPA. So there's a different stakes.
00:08:42
You know, the last thing I'll point to before we throw it to
00:08:45
you can listen to them. I did all that reporting on
00:08:48
bench, the accounting firm in Canada that blew up where the
00:08:52
founder basically said, oh, if you hadn't pushed me out, this
00:08:55
would might have turned out differently.
00:08:57
And you know, you can read those stories and it had a lot to do
00:09:00
with the particular investors and executives.
00:09:04
There was this real lesson that building a startup that hires a
00:09:07
bunch of accountants is is a big headache.
00:09:09
And so building a tech company where you're going to try and
00:09:13
hire all the people and then slowly figure out how to do it
00:09:17
better with software, it's hard to guarantee that you can bridge
00:09:20
that gap. And then one day you may find
00:09:22
yourself just operating a regular accounting firm and your
00:09:26
investors want you to have, you know, the margin profile of a
00:09:30
hot start up. And then you're in big trouble.
00:09:32
Good recap of the event, guys, and I think now we should just
00:09:35
kick it on over to two of the highlights of the event.
00:09:43
Very excited about this panel. We're going to have a great
00:09:46
discussion about what is possible with artificial
00:09:49
intelligence in fintech. We have two of the very well
00:09:54
financed cutting edge startups here.
00:09:56
Both of you can certainly claim that.
00:09:58
Gabe, you want to start with Rogo and just give the quick
00:10:01
like what you guys are doing? Happy to.
00:10:03
Thanks for having me. I'm Gabe.
00:10:05
I'm the CEO and founder of Rogo. We're training an AI analyst for
00:10:08
investment banks, private equity firms, and hedge funds.
00:10:11
We're a Series B company backed by Khosla, Thrive, Eric Schmidt.
00:10:15
A number of other folks deployed a, you know, a number of public
00:10:19
investment banks, large alternative asset managers,
00:10:21
large public equity investors. And you're, you're replacing the
00:10:24
banking analyst, right? Not yet, not yet.
00:10:27
Banking analyst right now, I mean, everything can be broken
00:10:31
up into augmentation and automation and we're much more
00:10:34
on the augmentation side still, even though, you know, I think
00:10:36
we and a lot of folks see that changing quickly.
00:10:39
And Jeff, the Digit story. Yeah.
00:10:42
So two months ago, we launched the first feature complete
00:10:44
replacement to QuickBooks in 20 years since 0 came out.
00:10:47
It was a long build. We started the company in 2018
00:10:50
and we're basically in R&D stealth mode ever since.
00:10:53
And so our goal has been to basically automate and make
00:10:56
small business finance real time intuitive, actually helpful to
00:10:58
the business owner, not put them in a position where they're
00:11:01
waiting two to three weeks to get a black and white PNL, which
00:11:03
is sort of. You're coming for a QuickBooks.
00:11:05
Space 100%, yeah. And there there are lots of
00:11:09
problems with QuickBooks. Some of them have nothing to do
00:11:12
with what a foundation model could produce, and some.
00:11:15
So dude, how do you break down Yeah, how much of what's making
00:11:18
digits different has to do with things foundation models can
00:11:21
produce? So.
00:11:22
It is pretty fundamental. I would broaden it to machine
00:11:25
learning. So when I started the company in
00:11:26
2018, our premise was can we build the first GLA general
00:11:30
Ledger for accounting that's ML native.
00:11:33
We honestly got very lucky with this whole AI wave.
00:11:35
Now it's AI native. Great.
00:11:37
OK. But we primarily do traditional
00:11:40
machine learning. If you look at accounting as a
00:11:42
field, it is predictive, It is not generative.
00:11:45
You do not want an LLM hallucinating your books.
00:11:48
And so we custom train and run our own models in production.
00:11:51
We do fall back to LLM Foundation models as sort of a
00:11:54
worst case scenario. But our story to the industry is
00:11:57
we are automating the bookkeeping.
00:11:59
We automate 93% of it today. And what's unique about the
00:12:02
accounting industry is they want that they are as a profession
00:12:06
trying to up level the profession away from bookkeeping
00:12:08
and into advisory and sort of client communication work.
00:12:12
And so that's what they're going for, I mean.
00:12:14
Basically, you use machine learning tag expenses, get it
00:12:17
100% right, and then if you're tagging system doesn't work,
00:12:20
you're like, all right, we'll try an LLM and see if it's sort
00:12:22
of more free thinking style can tag it itself.
00:12:26
That's the right way to think about it.
00:12:27
So if a business has seen the transaction before, our
00:12:30
predictive models are effectively 100% perfect.
00:12:32
If the transaction is novel to the business, we then fall back
00:12:36
to different tiers of models that ultimately result in an
00:12:39
agent. Like what does your bookkeeper
00:12:41
do? If you have something totally
00:12:42
new? They Google it.
00:12:43
What does the agent do? It Googles it.
00:12:47
Gabe, that I don't know. I don't take that as the most
00:12:49
optimistic view of what LLMS can do.
00:12:52
It's sort of like, oh, we have our accurate ones and then we
00:12:54
have our like guy who Googles things and like sometimes he's
00:12:57
helpful. Like you're much more leaned
00:13:00
into, like LLMS can deliver real value to bankers.
00:13:04
What's possible right now? How accurate is your software?
00:13:07
It's. Funny, Jeff and I were were just
00:13:09
discussing it backstage because we have two very different
00:13:12
products and very different users.
00:13:13
If you get financials that are not 100% accurate and you are a
00:13:17
small business and you don't know how to go in and audit
00:13:19
those books yourself, that's a pain.
00:13:21
If I give something that's, you know, good enough for an
00:13:23
associate to check, you know, maybe as accurate as their first
00:13:26
year analyst who's not perfectly, perfectly accurate,
00:13:28
that's actually very value additive.
00:13:31
Part of the, the benefit we had is that when we went in, folks
00:13:35
were used to using products like Chacha, BT and a lot of our
00:13:38
early users realized these tools can be very valuable without
00:13:41
being 100% accurate. And so I mean, the last thing I
00:13:44
would ever say when selling is that we are in 100% accurate
00:13:46
tool. What I would say is you can get
00:13:48
a lot more value than you think before they're 100% accurate.
00:13:50
You need to be better than someone who's right out of
00:13:52
college, and you need to be better than a lazy banker using
00:13:56
chachi. Exactly.
00:13:57
And as we learn, right, the way your intern learns to be an
00:14:00
analyst, learns to be an associate, VP, director, when
00:14:02
we're at director level, we're not going to get wooed by a
00:14:04
better paycheck somewhere else and all the enterprise value
00:14:07
isn't going to walk out the door and go to a competitive bank.
00:14:10
And what I mean, what's the coolest thing you can do today?
00:14:12
Like where do you think Rogo is really excelling?
00:14:15
Like what is the sort of query I?
00:14:17
Mean it does things why you should use.
00:14:19
I was an investment banking analyst at at Lazard.
00:14:22
I did buy Side M and A coverage for healthcare companies and I
00:14:25
didn't know anything about healthcare companies and we
00:14:28
would help large biopharma companies spend $5 billion to
00:14:32
buy biotechs and it can do almost all the work I did in the
00:14:35
analysis of those companies, right?
00:14:37
Like looking at their R&D pipeline, thinking about the
00:14:39
different you sort of oncology areas that might be additive for
00:14:42
Johnson and Johnson or or someone similar and then
00:14:45
preparing a presentation on why it would be additive to J and JS
00:14:48
overall M and a strategy. I mean we can put together the
00:14:51
materials diligence, the company, look through their
00:14:53
financials, do a do a lot of what these folks are doing.
00:14:58
Jeff, I mean, we saw sort of the presentation from Matt Harris
00:15:01
earlier, which is like get into the services business.
00:15:04
And I, I wrote last year a fair bit about like bench accounting,
00:15:08
which tried to sort of do accounting plus the actual
00:15:11
accountants. I mean, you're going after
00:15:14
QuickBooks, right? You are not building an
00:15:15
accountant team. You've, I think you've used some
00:15:17
accounting firms to sort of understand your product, but why
00:15:21
not go I guess whole hog and say we will be your accountant.
00:15:24
We know how to use our software better than anybody.
00:15:27
We'll take advantage of it and do the whole thing.
00:15:29
This is such a good question. So and just to clarify, we
00:15:31
actually do have an accounting team.
00:15:33
We do have an accounting firm. Some of you may have seen our
00:15:35
billboards. We do full service accounting
00:15:37
for hundreds of top startups now, OK.
00:15:39
That is not the business that we are not scaling that OK.
00:15:42
Yeah, because. We are.
00:15:43
I didn't think I was that wrong. Yeah, yeah, yeah, we do it, but
00:15:45
we're not going to do it at scale.
00:15:46
So we have. Capped the client, got it, got
00:15:48
it. The reason we did that is we
00:15:49
need to present a model firm to the industry, right?
00:15:52
And so if you look at the accounting industry, they're of
00:15:54
course traditionally relatively risk averse, relatively slow to
00:15:57
adopt new tools. The cloud transition took them
00:15:59
15 years. I'm not joking.
00:16:01
And so we want to accelerate that a bit.
00:16:03
And so we did build out a model firm to show you how you can run
00:16:07
digits as a practice that is not the core business.
00:16:09
And so I do think there's been a trap in the space.
00:16:12
If you look at some of the well known sort of previous offerings
00:16:15
that have tried to build a services business with some
00:16:18
internal tech. And I mean you even see Atrium
00:16:20
that failed at that in the legal space.
00:16:22
I think it's incredibly hard to bring together a software
00:16:25
business and a services business and actually scale it to high
00:16:28
margin. And so we've been really
00:16:29
disciplined on, we hire software engineers.
00:16:32
It is a software product like there is no slippery slope of
00:16:35
humans doing any of the work. Why?
00:16:37
Why has QuickBooks been able to hold on so well?
00:16:40
Yes, it's a marketplace product, believe it or not.
00:16:44
And when QuickBooks came onto the scene in the early 90s, it
00:16:47
also had taken them five years to build.
00:16:49
They were sort of building on the backs of the personal
00:16:52
computer revolution. For the first time, you could
00:16:53
really do accounting for your own business, like in your own
00:16:56
office. And the problem is you need both
00:16:58
the business owner and the accountant to really buy in on.
00:17:02
I'm going to use this software. And now 30 years later,
00:17:05
accountants view it as a career choice.
00:17:06
It's a religious preference. Like I'm a QuickBooks
00:17:08
accountant, I'm a zero accountant, I'm a NetSuite
00:17:10
accountant. And so that's the barrier to
00:17:12
break down. Fortunately, there is a macro
00:17:15
trend in our favour. So 33% of fewer people are
00:17:19
pursuing Cpas. Gen.
00:17:21
Z does not want to be an accountant, which you may or may
00:17:23
not blame them. And 75% of Cpas are at
00:17:26
retirement age. And so they're seeing this
00:17:28
talent crisis as the sort of final motivation of, OK, we need
00:17:32
to actually change software, We need to automate our role, We
00:17:34
need to up level the profession. So I think that trend is very
00:17:37
favorable. Gabe, you know, want to sort of
00:17:40
give a guide book for other companies, you know, you're,
00:17:43
you're, you're sort of at the cutting edge of using language
00:17:46
models in your business. I'm curious how much you think
00:17:50
what we've seen with ChatGPT is an endorsement of text based
00:17:54
exchange interfaces. Like do you think the regular
00:17:58
person or your banker customer wants to type inquiries and get
00:18:02
great responses or do you think that's sort of an entry point
00:18:05
over time? It's more and more like
00:18:06
software. Look, I think we're lucky in
00:18:09
that we, we have a lot of users who use the product we have
00:18:12
today. I don't know if that means we
00:18:14
have, you know, any predictive abilities about what that UX
00:18:17
should be 5 years from now. For me, the abstraction is what
00:18:21
is the easiest way for me to communicate with someone on my
00:18:23
team. It's probably Slack, e-mail,
00:18:25
text message, phone call. You know, if I ask for a
00:18:28
deliverable, I get sent to PDFI review it.
00:18:30
Occasionally I'll jump into the PowerPoint, the Excel backup
00:18:33
myself. I think that's a very human way
00:18:35
to work and there's a lot of throughput in that, you know,
00:18:37
you can communicate, indicate a ton.
00:18:39
I think at the limit, that's what you want these interfaces
00:18:41
to look like. If you're building an assistant,
00:18:44
if you're building, you know, an analyst replacement, if you're
00:18:47
building Jeff's business, you know, I, I'd rather just have my
00:18:50
books integrate directly into, you know, whatever systems those
00:18:52
need to go into. I don't know anything about.
00:18:54
No, this is actually, this is great because this is such a a
00:18:58
product specific difference, right?
00:18:59
Like as a startup founder, business owner, I don't want
00:19:02
another person to manage. I don't want another thing to
00:19:04
talk to, I just want the accounting done right and so we
00:19:07
very explicitly have no chat UI in the product the data comes
00:19:11
in. It is booked where?
00:19:12
The output of Rogue, oh, I mean analysts would produce a report,
00:19:15
so it's not that. Yeah, I got you, but.
00:19:17
The way you would iterate might be, Hey, why'd you do this?
00:19:19
Hey, add a page on this. Hey, Are you sure?
00:19:21
You know, that was the post money valuation for the company
00:19:23
and in the comp set. But I mean, it's, it's, you
00:19:25
know, for, for Jeff, it's like, I don't want to hire an FPNA
00:19:28
person, right? We're getting to the scale where
00:19:30
it's like, you know, we need someone and I really don't want
00:19:32
to do it. I'd rather just have those
00:19:34
systems work and I don't want to talked about it with someone
00:19:36
either. Right.
00:19:37
Are you an evangelist when you talk to other fintech founders
00:19:41
of like they should be applying foundation models to their
00:19:44
business? Or are you more on the side of
00:19:46
like, we picked the right area, it makes sense where we're doing
00:19:48
it. And like, I don't know if it
00:19:50
makes sense for your business. Yeah.
00:19:52
I was I yesterday at Khosla Ventures is one of our big
00:19:55
backers and, and they had a big summit yesterday and, and
00:19:58
something that one of the speakers kept hammering was it's
00:20:00
very hard for incumbents in a space to innovate, to brainstorm
00:20:04
the sort of, you know, not just the incremental way to apply
00:20:06
LLMS to your business, but the whole new business model that
00:20:09
emerges. And part partly why digits is so
00:20:11
fascinating is because it's innovating on the business model
00:20:14
in addition to, you know, the actual technology.
00:20:16
But I mean, I was, I was pushing Jeff in our we had lunch
00:20:19
together to catch up before this.
00:20:20
And I'm saying it sounds like you're not using.
00:20:22
LLMS enough and he had a great retort and he obviously knows
00:20:26
his business better than I do. But I mean, I think anyone who's
00:20:28
not trying to use these tools in in kind of a step change way as
00:20:32
opposed to an incrementalist way is is waiting for someone else
00:20:35
to Jeff. You're an incrementalist.
00:20:39
This is great. No, I mean to be clear, we do
00:20:41
use LLMS, we do use agents. We've been running in in
00:20:43
production for 18 months now. I'd say we are extremely
00:20:46
disciplined on making sure we can guarantee the accuracy of
00:20:50
the output. And so for an example like where
00:20:52
we do use LLMS, we actively prevent them from doing math
00:20:56
because yes, they've gotten better at math, but they can
00:20:58
still slip up. Our team literally wrote an RPN
00:21:01
calculator tool that we give to the LLMS and we tell them if you
00:21:04
need to do math, do not try use this tool.
00:21:06
And so there are a bunch of things you can do to put
00:21:08
safeguards around it. What?
00:21:10
Do you think about that? Well, you know, I, I think it's
00:21:13
kind of the analogy I would use and it's going to be a really
00:21:16
strained analogy, so bear with me.
00:21:18
But the sort of Jeff Bezos thing of, you know, your margin is my
00:21:21
opportunity, your accuracy threshold is someone else's
00:21:24
opportunity because they're going to build something that's
00:21:25
not going to work today or tomorrow.
00:21:27
But all of a sudden the base models will be good enough where
00:21:29
you could skip all these layers of complexity and structure to,
00:21:33
you know, make up for that accuracy.
00:21:35
And suddenly, you know, they will have some of that more
00:21:37
human reasoning and be capable of so much more.
00:21:39
I don't know if that's right in. Jeff on like the continued
00:21:43
improvement of the models and if you spend too much time fixing
00:21:47
the models today, you're missing out on the, I mean saving the
00:21:49
problem is. How do you thread the needle?
00:21:51
How do you get enough adoption and traction today to, you know,
00:21:54
validate putting more in R&D for the future?
00:21:56
And like, depending on how accurate you need to be, yeah,
00:21:59
you might need to do some anachronistic things or like,
00:22:02
you know, not use it, just throw LLMS at every problem, right?
00:22:04
It's not, you know, you don't just want to use LLMS for
00:22:07
everything. But I do think that like, you
00:22:10
know, if if you aren't, if you aren't willing to, to accept
00:22:14
some faultiness, some bugginess, some hallucination in parts of a
00:22:17
product like someone else will. And then someone else will
00:22:21
potentially figure out the new product paradigm or UX that's
00:22:23
going to work when the models just get, you know, 10% more
00:22:26
reliable. Jeff, we're in this moment
00:22:30
where, you know, people are like, oh, we don't need software
00:22:32
anymore. You're going to use like a
00:22:33
coding agent to build it. Like I'm going to build my own
00:22:36
custom model. Like why isn't the next digits
00:22:38
just like a home brewed version of 0 instead of like your your
00:22:43
company? Right now this, this is a really
00:22:46
good question. And Can you imagine the models
00:22:47
get so good that you don't need the accounting software at all,
00:22:50
right? And like, could the model just
00:22:52
you, you give it your data and it does your books?
00:22:54
I, I don't think that is the future you want because
00:22:57
ultimately this is still a workflow.
00:22:59
It's a real business process that you need everyone involved
00:23:01
with. You need user accounts,
00:23:03
permissioning, notification sharing, like you need an actual
00:23:06
experience. And so even to the extent we
00:23:09
invest in the AII, do think it comes down to you also need to
00:23:12
build better accounting software, right?
00:23:14
And it's the same thing I would say with Figma, it's like, OK, I
00:23:17
don't think Figma is disappearing.
00:23:19
It can help you. It can automate a lot of the
00:23:20
stuff, but you still ultimately want to canvas your team can
00:23:22
collaborate around. And that's how we you finance.
00:23:25
Like if you could picture a Figma for finance, what would it
00:23:28
be? So we try to look at it as a
00:23:30
broader product perspective where the AI is a core tool, but
00:23:34
it's one piece of the entire vision.
00:23:36
Just like if you were to build like, oh, Redis came out now, do
00:23:39
you not need a whole other like class of products?
00:23:42
You sort of still do like Redis is great technology, but you
00:23:44
still need the actual experience around it.
00:23:46
Do. You think we're close to you
00:23:48
having the problem of someone gets an accountant who that
00:23:52
accountant is using language models that are trying to
00:23:54
interact with your software and then you're trying to figure out
00:23:57
how much to serve them or have you started to see that or are
00:24:00
we too early? We're we're a little too early.
00:24:02
The accountants aren't that quite They're not only.
00:24:04
Adopters. But we we think deeply about the
00:24:08
collaborative flows like the accountant is in the picture.
00:24:10
And this is actually an interesting point for folks as
00:24:13
we talk with business owners. They do not want AI accounting,
00:24:18
IE the the AI does 100%. They actually want the AI to do
00:24:21
most of it. And they want their books
00:24:23
blessed by a real CPA. Because you can't sue the AI,
00:24:28
right? You need someone to I.
00:24:29
Need a human I can blame? And so that is actually really,
00:24:33
really important and that's why the industry is not going
00:24:35
anywhere, right? They will up level and be able
00:24:37
to serve more clients. But you still need that human to
00:24:40
have blessed the books. Gabe, the rapper question, I
00:24:43
mean, we sort of in the presentation earlier today, sort
00:24:45
of, you know, we've, we've moved to a point where it's like, OK
00:24:48
to be a rapper. We don't use that term because
00:24:49
it's pejorative. Like where are I, I, you know, I
00:24:53
love to use pejorative words positively.
00:24:55
Like I've been on the case that it's like, no, you differentiate
00:24:58
on the product and software, not necessarily like on the model.
00:25:01
How would you calibrate that in terms of how much you need to
00:25:04
have at some point, your own models, your own thinking, or
00:25:08
we're going to be the best at directing chachi tea and
00:25:11
anthropic and everybody towards our use case.
00:25:14
Look, it kind of goes back to how Jeff answered the question,
00:25:17
which is even if you have a profoundly intelligent model,
00:25:20
you still need the tools to, you know, do the accounting workflow
00:25:22
right. It's like even if you have a
00:25:24
robot that can move the way human hand, they still need a
00:25:26
drill to go, you know, install whatever they're installing for
00:25:30
us. We're building the tools that
00:25:31
some eventual agent intelligence either owned by us or someone
00:25:35
else might use, Whether that's an Excel interface for an agent,
00:25:38
whether that's a PowerPoint interface, whether that's
00:25:40
structured financial data querying or unstructured
00:25:43
financial data querying or integrations into the internal
00:25:46
data systems of the banks and private equity firms we work
00:25:48
with. That said, I think it's very
00:25:51
clear that reinforcement learning is working and POST
00:25:54
training is working on top of these Frontier models.
00:25:57
And with very great specific evals you can actually
00:26:01
outperform the Frontier. You with post training alone,
00:26:05
you can find a lot of technological differentiation
00:26:07
that that makes it more valuable.
00:26:09
Yeah, I I mean, I think. I think that's what early
00:26:11
results are showing. I don't know if there's, I mean,
00:26:13
actually cognition had a great paper that they, I mean, not
00:26:16
paper. It was a Twitter thread, but it
00:26:17
but it was a great Twitter thread that.
00:26:19
Sums up something in the current it was a.
00:26:20
Great Twitter thread. Like maybe a week ago on they
00:26:24
had, they had post trained a model with RFT for, you know,
00:26:27
writing like CUDA kernels or something that that's quite
00:26:30
difficult for O3. And they got a lot of
00:26:32
outperformance with the types of technologies and and
00:26:36
reinforcement techniques that are now diffusing through
00:26:37
industry. And so I think we'll start to
00:26:39
see a lot more of that. It reminds me of when folks used
00:26:41
to overclock their CPUs and now CPUs are fast enough where no
00:26:45
one really bothers as much. Yeah, No, no, that's, I mean,
00:26:47
it's totally possible. I mean, I think the way the way
00:26:50
that I think about RFT and post training is not necessarily that
00:26:53
it's going to make the model smarter, but it's going to sort
00:26:56
of prune the decision tree and make it better at using your
00:26:59
tools, right. So like opening eyes never going
00:27:02
to have your tool structure and it's training data unless you
00:27:04
collaborate with that. And so the same way if you get a
00:27:06
super smart human starting at your firm on Monday, they're
00:27:08
going to need to learn how to use QuickBooks and zero and I
00:27:11
hope digits, you know the model you want to teach it how to use
00:27:14
those tools to Yep. Jeff, last question, last word.
00:27:18
Would you tell a founder today to build a services business
00:27:22
powered by you've done the sort of flirtation with having the
00:27:26
accounting firm? Do you buy that investor
00:27:28
narrative or do you think it's oversold?
00:27:30
I think it's oversold and it's so tough because it is such a
00:27:33
tempting narrative. It's build your services
00:27:35
business, look at what's slow, automate that, move on, right?
00:27:37
You can drive a high margin business.
00:27:39
The change management, the actual people management is so
00:27:43
challenging because it depends on which profession you're
00:27:46
trying to automate. But there's so much built in
00:27:49
just like native practices in that profession that in order to
00:27:52
really automate everything and build the trust, it's a long
00:27:55
journey to get to that margin versus.
00:27:57
Just turns out people outside Silicon Valley, they know their
00:27:59
businesses too. They're hard businesses.
00:28:01
Exactly right. And again with accounting,
00:28:03
accounting is complicated like. There is no way to simplify
00:28:06
that. Great.
00:28:07
Thank you very much. Thank you, Schuber say, about
00:28:13
this conversation, two things. One, Gusto's the rare startup
00:28:18
where I sort of have a rooting interest in that we are a small
00:28:21
business, and every time Gusto solves a problem, I am relieved
00:28:25
that you get to take it off my plate.
00:28:27
The second thing is it's funny to be hanging out with you with
00:28:29
shoes on, you know, do you guys still have you, have you gotten
00:28:32
rid of your no shoes policy or are you still I?
00:28:35
Feel like I have to give context to people here.
00:28:37
So excited to be here. This is a fun space for me
00:28:39
because we had a Gusto holiday party here in 2015, but our
00:28:43
three first offices were only a block away and we were raised by
00:28:47
our parents to take our shoes off at home.
00:28:49
So for many years our offices were shoes off, walking around
00:28:52
in slippers, socks. We passed them out to folks and
00:28:56
then when the pandemic started, obviously people can do whatever
00:28:58
they want at their home. Today, it's more of an optional
00:29:00
policy. But I do describe it as a
00:29:02
tradition that that felt right for gusto at the time.
00:29:05
You know, yeah, in some ways it's like you've become this
00:29:08
grown up company. Can you give, you know,
00:29:12
obviously payroll is at the heart of it, but talk through
00:29:15
just briefly what the pieces of Gusto's businesses are today in
00:29:18
terms of the focus and where you're really strong?
00:29:22
So something really important to know about Gusto is we obsess
00:29:25
over a small business. So if we have any customers
00:29:27
here, we're honored to serve you.
00:29:29
And just know that there are more dentist offices in the US
00:29:33
than tech startups. And so we really focus on
00:29:36
mainstream small business of the six million ish employers in
00:29:40
America, 2/3 or less than 5 employees.
00:29:43
And so with that customer focus, payroll is our first product.
00:29:47
I always like to say if you don't pay someone, they quit.
00:29:50
So it's a pretty non optional product.
00:29:54
So we're really happy we started there, but kind of broadening
00:29:57
through the back office would be the way to describe it.
00:30:00
We really obsess over customer pull versus company push, but
00:30:05
benefits is a big investment area for us.
00:30:06
There's many, many types of benefits, something I know we're
00:30:09
excited to get into more. We have a pretty broad product
00:30:12
group inside Gusto called Gusto Money that's focused on things
00:30:16
related to cash flow management. For a lot of our customers,
00:30:18
their biggest expense is payroll, paying their team.
00:30:22
And one of the biggest points of stress is when APAR don't line
00:30:26
up and they don't have the money in their bank account at the
00:30:28
right moment to fund payroll. So we launched a bill payment
00:30:32
not too long ago. Actually next Wednesday, we're
00:30:34
going live with invoicing. So you'll see us do a lot more
00:30:37
product investments, more in that B2B fintech part, but
00:30:40
again, really, really focused on small business.
00:30:43
And overall business sort of somewhere between 500 million
00:30:46
and a billion in revenue. Can you?
00:30:48
Describe it that way. We serve over 400 companies.
00:30:51
I mean, one fun way we always grounded in customer we're you
00:30:56
know, in the range of, you know, 8-9 percent now of all employers
00:30:59
in America. And with the business model that
00:31:02
leads to, you know, good revenue, we've been free cash
00:31:04
flow positive for many years. We reinvest that money back into
00:31:07
building new product to solve more pain for our customer.
00:31:11
And I mean, you're an important sort of partner to other fintech
00:31:15
companies. I know.
00:31:16
I think the guidelines CEO was here at one point.
00:31:19
What what do you see in terms of like the partnership strategy
00:31:22
for Gusso? Yeah.
00:31:24
So a big part of our approach to we're here to build a multi
00:31:27
decade company, we're still early in the journey even though
00:31:30
it's, you know, over 10 years. And that to me is a factual
00:31:33
statement, right? We're only at about 89% of
00:31:35
employers. You still have 30% of companies
00:31:37
in the US doing even things like payroll by hand on pen and
00:31:40
paper. So with the obsession on small
00:31:42
business, a lot of what we're doing is replacing manual
00:31:45
process, but there's just a lot of other pain points we want to
00:31:48
help with. And if we're going to get there,
00:31:50
it's not going to be entirely through first party products.
00:31:52
So we've had a playbook, which we will continue to use of
00:31:55
choosing product by product. Are we going to build?
00:31:58
Are we going to partner? 4O1K is a good example of where
00:32:01
we've partnered historically. And then there's also obviously
00:32:04
acquisition. We've done that a few times.
00:32:06
We'll do that more going forward for where we deem something that
00:32:09
was third party becoming first party.
00:32:11
But you know, we're not going to build it all ourselves, and
00:32:14
we're excited to partner where it makes sense.
00:32:16
Has artificial intelligence changed the pace of product
00:32:20
development? I mean, you're, you're an
00:32:22
interesting case in that on the one end, it's sort of the, you
00:32:26
know, I don't know, Klarna's the world where they're like, well,
00:32:28
we'll build everything internally.
00:32:29
Like, you know, payroll could be imagined that way.
00:32:32
On the other hand, it can accelerate your growth.
00:32:34
Like how do you, how are you seeing sort of artificial
00:32:37
intelligence, the ability to build faster impacting gusto
00:32:41
today, I mean? I'm surprised it took this long
00:32:43
to get to AI as a topic. I mean there's so many threads.
00:32:46
I think a lot of folks here are technologists.
00:32:48
I'll have to just choose like on productivity and how we build.
00:32:52
We have 1000 person plus R&D team.
00:32:54
I don't think we're that different than most software
00:32:56
companies in that it's obviously driving pretty meaningful shift
00:33:00
in how we build. You know, the iteration speed.
00:33:03
I think we're going to talk later about, you know, there's
00:33:05
the service as a software, software as a service, everyone
00:33:09
gets SAS. A lot of what we do is taking
00:33:11
stuff that involves compliance, involves manual filings and
00:33:16
digitizing it into software. In the past I said we use
00:33:20
paperless cloud mobile. Now obviously AI is a key
00:33:22
ingredient to that, but that's on the more internal how we
00:33:25
build side on the. Doesn't mean fewer engineers
00:33:28
just to put a fine line on it. Yeah, doesn't mean fewer
00:33:31
engineers. Given the scope of our ambition.
00:33:34
It means like the amount of work we can do per gusty grows
00:33:38
dramatically. But probably in the grand scheme
00:33:41
of things, it means less hiring than if AI didn't exist.
00:33:45
But we're also pretty ambitious, so we're doing a lot of hiring
00:33:48
anyways. But it's because we want to go
00:33:50
solve more pain point for our customers.
00:33:51
And then on this sort of, do you feel the threat of customers
00:33:56
building in house versions of Gus?
00:33:59
Honestly, as a small business that's sort of incoherent,
00:34:01
right? Because it's like I'm, I'm not
00:34:03
going to build an AI team as a five person startup.
00:34:06
It would matter more to your enterprise level competitors or
00:34:09
how do you think about that challenge I.
00:34:11
Mean, I don't think too many dentists want to go, right,
00:34:14
build software. That said, the reality is most
00:34:17
of what we're doing for a dentist office or other small
00:34:20
businesses they haven't had access to in the past.
00:34:23
They've been on their own. They've kind of had to do it
00:34:25
manually or just not do it. And so for us, AI is enabling
00:34:29
products to exist in a small business category that never
00:34:32
existed. I think in mid market
00:34:34
enterprise, you're navigating more of this complexity of
00:34:36
displacing jobs. But our due N is to like grow
00:34:40
the small business economy, right?
00:34:42
And so we want to bring all of the things big companies have
00:34:45
had historically to a small company.
00:34:48
The only way we're going to do that right with a ratio that
00:34:51
makes sense and with a scalability that makes sense is
00:34:54
through leveraging technology. I.
00:34:55
Mean, if I had to distill the angst of a small business, it's
00:34:58
like I want to know that I'm in good standing and I don't know
00:35:01
everything I'm supposed to do. And you guys are, but in the
00:35:04
products, sometimes we like, you know, in New York, you're
00:35:06
supposed to have sexual harassment training.
00:35:08
OK, I'll pay you some money for that.
00:35:09
Like, do you think AI is going to move us further along of just
00:35:12
being able to give me an assessment?
00:35:14
Like, have you done everything you're supposed to do?
00:35:17
My combat fee, am I Good question.
00:35:19
And like our customers, including you, do not want to be
00:35:24
stressed or worrying about that. And it's our job to make sure
00:35:27
you know fully if you're good or if there's something you need to
00:35:30
do what that exact thing is. And if we can do it for you,
00:35:33
even better. But yeah, compliance maps to
00:35:35
most of our products, right? Payroll involves a lot of
00:35:37
compliance time tracking, PTO involves a lot of compliance,
00:35:41
local, state, federal rules. And so AI fits in there in terms
00:35:45
of like ingesting. But at the end of the day, it's
00:35:49
a compliance engine where accuracy needs to be to the
00:35:52
like, you know, 6 Sigma. And that's always been true pre
00:35:56
AI, post AI, that's been true since we started the company.
00:35:59
How much do you think your customers want to talk to the
00:36:02
product? Like, do you see a move to text
00:36:05
interfaces now that they're more capable or you're like, we
00:36:09
really need to flow whatever advances we have into actually
00:36:12
the layout and design of the product?
00:36:15
So we always start with like small businesses are busy and
00:36:17
it's way too hard to run and build a small business.
00:36:20
Still, you can attest to that. And so how can we help them?
00:36:23
We can help them by saving time. We can help them by taking 5 or
00:36:26
20 or 30 of the hats on their head off their head and doing it
00:36:30
for them. Pre AI, I would say we did that
00:36:33
through really clean, elegant, easy to use work flows primarily
00:36:37
in a web app or a mobile app. That's gone us pretty far,
00:36:40
right? Like that's what we're known
00:36:41
for. Anyone can use Gusto with no
00:36:43
training, no background in running a business.
00:36:46
We have high NPS, high customer satisfaction.
00:36:48
Our primary way of growing is word of mouth with AI.
00:36:52
It's not about the technology, the interface.
00:36:54
That's what we get excited about.
00:36:55
A conversational interface for a lot of use cases that Gusto does
00:37:00
is a more intuitive, more accessible way to use Gusto.
00:37:04
You like it? You think it's?
00:37:05
Promising. We think it's very promising.
00:37:07
Our interface for that is called Gus, which people should
00:37:11
hopefully get. I think it's somewhat like this.
00:37:12
And then, yeah, well, Penny the pig is.
00:37:14
Oh, that's a. Different logo OK, but.
00:37:16
Then like you're like cracking Penny in half, but like Penny's
00:37:19
being sacrificed for the sake of your conference.
00:37:21
But yeah, conversational interface we don't think
00:37:24
replaces web app, replaces mobile app.
00:37:26
It's just a different surface area.
00:37:28
But like we are taking 14 years of functionality and giving Gus
00:37:32
those superpowers. So we have customers today,
00:37:35
thousands of customers using Gus to go create shift schedules,
00:37:38
right? You want to go change who's
00:37:40
working with today? You know, take Sally off Friday,
00:37:43
add 20 hours to gym. We go do that all for you.
00:37:46
You can just tell us that in a conversational interface, or you
00:37:49
can go through and navigate, but that saves time for a small
00:37:53
business, which they really appreciate.
00:37:54
I wanted to put one of the big themes of today to you.
00:37:58
I mean, we saw Matt Harris give a presentation where he's making
00:38:02
sort of a case that other investors have also made that
00:38:05
it's like there's now, thanks to language models, this
00:38:08
opportunity to go after services businesses and sort of be the
00:38:12
services business while being a tech company.
00:38:15
Then later I had Jeff, the CEO of Digits, who's building
00:38:19
accounting software. He's, you know, hired
00:38:21
accountants to figure it out and then software.
00:38:23
And I think he was much more skeptical of that sort of, you
00:38:26
can be a services tech business. Where where do you land?
00:38:30
You've built a big business and you touch a lot of services.
00:38:34
Like are you bullish on that argument?
00:38:37
So I think we've had a kind of unique perspective to this
00:38:41
because the first product we launched payroll mostly if you
00:38:46
put aside like ADP paychecks, which are a minority of the
00:38:48
market, the majority of companies in America did payroll
00:38:51
by hand or with a local payroll service Bureau, which was
00:38:55
basically a service manual localized business solution.
00:38:59
So a lot of our journey has been taking and did it with health
00:39:03
benefits, right? We are a broker.
00:39:04
There are thousands of brokerage firms across the country that
00:39:07
mostly do things by hand. So we take these complex
00:39:10
compliance centric spaces and we do what we do best, which is we
00:39:15
digitize the heck out of it because if we're going to go do
00:39:17
it for 400 thousand, a million businesses, we cannot do
00:39:22
it ourselves manually. So I don't think it's a zero a
00:39:25
hundred, a hundred zero. Like we have operations teams at
00:39:29
Gusto, but we have over 400 companies today.
00:39:32
Our Gusty to employee or ratio is over 1 to 1000.
00:39:36
That's only possible with the technology.
00:39:38
Right. So you only want to do it if you
00:39:40
can get to the point where it's fairly automated, not sort of
00:39:44
for a long time, keep the humans sort of trying to figure it out
00:39:47
or yeah, how long do you spend sort of like we'll do the human
00:39:50
business to try and get to the point where we can automate it
00:39:53
before saying, oh, I guess this is sort of, at least with the
00:39:55
technology today, a perpetually human.
00:39:57
Business, well, we were joking before, right?
00:39:59
Like if a company is scaling a service offering to large, large
00:40:03
volumes and it's staying service based, that's to me not a tech
00:40:07
company, that's just a service business.
00:40:08
Different multiple profile, different overhead costs,
00:40:11
different operational complexity.
00:40:13
We work backwards from what's best for the customer.
00:40:16
For us to solve these pain points, it has to be technology,
00:40:18
it has to be digital. It's just more accurate, right?
00:40:21
We process several $100 billion a year of payroll taxes.
00:40:24
If we were doing that manually, it would create human error
00:40:27
potential. It would be too time consuming.
00:40:30
So we digitize because it's a better experience.
00:40:33
But when we're sub scale and trying out new products early,
00:40:35
early on, I think it's totally fine to start more manual.
00:40:39
It's about getting that feedback loop going, getting that
00:40:41
learning going. But that's typically with less
00:40:43
than 100 customers, right? If we're going to scale
00:40:46
something, it better have good unit economics, good CAC, good
00:40:49
gross margin and we should have confidence that it can scale in
00:40:52
a, you know, technology enabled way.
00:40:54
Otherwise, we're just going to build a huge operations team.
00:40:57
Somehow in payroll, you're in like 1 of the wildest categories
00:41:02
in Silicon Valley. You have, you know, like
00:41:04
Rippling and Deal, you know, making huge accusations,
00:41:08
obviously Ripley accusing Deal of spying on it.
00:41:11
Have you done an internal check to see whether anyone's spying
00:41:15
on you? And what do you make of sort of
00:41:17
your high profile, sort of same, same category businesses having
00:41:24
so much drama? We we just focus on the
00:41:28
customer. Have you have you checked?
00:41:30
Have you checked? We have a very, very like
00:41:35
intense, like super, super expensive and worth IT security
00:41:39
team. Yeah, but that's because we have
00:41:43
like hundreds of billions of dollars of our customer money
00:41:46
sitting in our accounts. And we, you know, do tax filings
00:41:49
and tax payments and we hold Social Security numbers and a
00:41:52
whole bunch of very sensitive data that we take that job very
00:41:55
seriously. So I can tell you with
00:41:57
confidence, we do not believe anyone is spying on Gusto, but
00:42:02
that's because that's what our customers should demand of us,
00:42:04
that we're obsessed with protecting them in every
00:42:06
possible way. Do you see much competition or I
00:42:10
mean, you're very focused on small businesses and I like that
00:42:13
focus and it gives you a great mandate and sort of a mission to
00:42:17
motivate the company. On the other hand, most
00:42:19
businesses want to keep their customers.
00:42:21
It's like it's much easier to keep the customer you have and
00:42:23
grow with them than to find new ones.
00:42:25
How do you think about sizing up with your customers?
00:42:29
So we're both in a like very crowded space and a very
00:42:33
fragmented space. And then yes, I think it's even
00:42:36
very simplistic to say small business, medium sized business,
00:42:39
large business, these are huge cuts.
00:42:42
And even within small business, you have different industry
00:42:44
types, different geographies. So, you know, hopefully everyone
00:42:48
gets by now. We are obsessed with small
00:42:50
business. We tend to see they're less of
00:42:53
the kind of frothy Silicon Valley tech kind of drama,
00:42:58
frankly. But there's still legitimate
00:43:00
companies that have lots of good product in market.
00:43:03
Intuit is a company we highly respect and we compete with
00:43:06
multiple fronts. And you know, there's a bunch of
00:43:10
other companies I used to say was only into IT folks that have
00:43:12
built really incredible franchises and small business.
00:43:15
You know, Shopify has a huge part of their company that's
00:43:18
focused there, Square, Toast, HubSpot.
00:43:21
So we feel like we're in good company and I always encourage
00:43:25
more entrepreneurs to tackle the small business category.
00:43:28
The playbook still is generally for SAS just to move up market
00:43:32
and move to bigger enterprise. And in the past, I just meant
00:43:35
small business was left. Do you?
00:43:36
Want 100% business on Gusto? We have a lot of 100% business
00:43:40
on Gusto. Most of them started on Gusto
00:43:43
when they were one employee. So our focus is on small
00:43:46
business. The vast majority will stay
00:43:48
small if some grow bigger. We're honored and happy to serve
00:43:51
them. There's probably a point at some
00:43:54
point where they graduate. But I like to remind folks, you
00:43:57
know, that's very rare, right? Like the number of companies
00:44:00
that start at one or two and grow to 100 and our entire
00:44:02
history is less than .1%. Is a, you know, we all at the
00:44:07
end of the day, live or die by the economy in some ways, and
00:44:10
you have a great sort of vantage into a part of it.
00:44:14
Is small business in America strong right now?
00:44:16
What's your read of the mood among your customers?
00:44:20
Yeah. So we have economists on team.
00:44:22
We publish a lot of content under Gustanomics and a lot of
00:44:25
folks are interested in the data.
00:44:27
I'd say high level two things we track a lot new employer starts
00:44:31
and then net hiring across our customer base.
00:44:34
I'd say on the second that is quite depressed.
00:44:37
New hiring new. Hiring, net hiring.
00:44:39
So just the propensity of our customers to add more to their
00:44:42
team that's been depressed for a few years now on the new
00:44:46
employer starts that really with the pandemic got dramatically
00:44:49
elevated and it stayed fairly high.
00:44:52
But yeah, it's more interesting to track given our size and
00:44:55
scale. We feel like we're pretty
00:44:57
diversified. It's more just net, net a bet on
00:45:00
small business broadly and it tends to be pretty durable
00:45:02
segment. Can you talk about the embedded
00:45:06
payroll part of your business? And sort of, I mean there is
00:45:09
this other trend of, you know, API based businesses right now
00:45:12
where it's like, OK, you're helping other people interact
00:45:15
with your service, but not necessarily building out the
00:45:17
product yourself. How do you think about that as a
00:45:20
piece of Gusto's business? Yes, we're always going to be
00:45:23
driven by customer pull. So I'm very excited our direct
00:45:27
business will keep growing past, you know half a million, a
00:45:30
million businesses in the coming years.
00:45:33
But a couple years ago we noticed and maybe some of you
00:45:36
are part of this trend, but vertical SaaS in particular we
00:45:39
got really excited about where you have companies tackling a
00:45:42
very specific industry could be very esoteric, but if you obsess
00:45:46
over that one category kind of build business in a box, you can
00:45:49
actually create a viable tech company there.
00:45:52
And every one of these folks kept coming to us and saying,
00:45:55
you know, we don't want to build payroll.
00:45:57
No one really wants to build payroll cuz it tends to be quite
00:46:00
difficult. But we really want to provide
00:46:03
payroll to our customer. And we don't want to just keep
00:46:04
routing customers to you. We want to have it be native, a
00:46:07
native product experience. And so we have some great
00:46:10
reference points in embedded in Fintech, especially like Stripe
00:46:15
as a good example. And so that was where the
00:46:17
genesis of embedded payroll came about.
00:46:19
That's what we call Gusto embedded payroll.
00:46:21
And we're really excited there. We have a number of partners on
00:46:25
the banking side. Chase payroll is powered by
00:46:27
Gusto. We recently announced U.S. bank,
00:46:30
we'll be rolling out payroll powered by Gusto.
00:46:32
You know, 0 is another good example on the contact side.
00:46:36
And so anyone that does want to launch or offer a payroll
00:46:39
natively within your product but doesn't want to build in from
00:46:42
scratch, please let us know. We're eager and excited to
00:46:45
partner with you. That whole business is more of
00:46:47
an infrastructure. Business, my last question, you
00:46:49
know, we have a lot of founders here.
00:46:51
You've been at this a while. Like what what?
00:46:54
What's your main piece of advice to somebody starting a company
00:46:57
today? Or what would you do differently
00:46:59
if you were starting Gusto right now?
00:47:02
So I'll answer both. My main advice I'm pretty
00:47:05
consistent with, but I really believe it.
00:47:08
It's imagine the 10 thousandth time you're describing what
00:47:10
you're doing. Will you be as excited as the
00:47:12
first time? Because at that point you can't
00:47:14
fake it, It will show. And you have to have a deeper
00:47:19
interest, passion, borderline obsession with the thing you're
00:47:23
trying to fix, the problem you're trying to make better
00:47:26
because that's what gets you through all the ups and downs of
00:47:28
company building. And in our case, you know, I was
00:47:31
with a team earlier today that's an incubation team launching a
00:47:34
new product. And, you know, we went around
00:47:36
the room and it was an easy first question.
00:47:39
I didn't even have to ask it. Everyone just shared their
00:47:41
favorite small business. And so in our case, you know,
00:47:44
that obsession with small business comes through hopefully
00:47:46
very clearly. It's a big part of our hiring
00:47:47
filter. And as long as they're in pain,
00:47:49
we feel like we have purpose and we have so much to do that we
00:47:52
feel like we're really still early.
00:47:54
There's a lot, lot more ahead. And if I reflect on learnings
00:47:57
from prior chapters, I had a prior start up where that wasn't
00:48:00
the case. And that's where, you know, you
00:48:03
can, you can start a company or if you join a company, you feel
00:48:05
like there's a miss there and wonder what's off right.
00:48:09
And, and I would argue, you know, go back to like, what is
00:48:12
the purpose? What is the reason why you
00:48:14
exist? And it should not be about you.
00:48:16
It should be about your customer.
00:48:18
Great. Well, thank you very much.
00:48:19
Thank you.