November 13, 2024 • 33 mins
In this episode of Canopy’s Practice Success Podcast, Davis Bell, CEO of Canopy, interviews Botkeeper CEO Enrico Palmerino. They discuss Enrico’s journey from tackling accounting challenges as an entrepreneur to founding Botkeeper, a platform automating bookkeeping through AI.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Davis Bell (00:06):
Hello and welcome to Canopy's Practice Success
Podcast.
I'm Davis Bell, the CEO ofCanopy, and I will be sitting in
as host for today's podcast.
I, I'm excited.
They don't usually let me dothese things.
They don't let me out in publicoften, so.
Um, I don't know what happened,but, uh, glad, glad to be here.
I'm super excited to sit downwith Enrico Palmarino, who's the
(00:28):
CEO of Botkeeper.
And today we're going to chatabout Botkeeper and what the
accounting profession can learnfrom the world of tech and
venture capital, as well as wesee AI taking the profession.
So welcome, Enrico.
Great to be here with you.
Enrico Palmerino (00:40):
Thank you, Davis.
Appreciate it, sir.
Yeah, super
Davis Bell (00:43):
excited.
So, um, I, I thought maybe, um,And I've got two questions.
I think they're probably superrelated.
One, I just want to know aboutyour background prior to Bot
Keeper.
And my next question is whatinspired you to create Bot
Keeper?
And I think those are actuallyvery related.
So maybe, you know, you justtake them in whatever order you
want.
Enrico Palmerino (01:04):
So, uh, background.
Um, I was a quant major, uh, incollege and while at Babson, uh,
which is an entrepreneurialschool.
So, uh, everyone's alwaysstarting businesses there, which
is a lot of fun.
And I started a company thatautomated how you, uh, Analyze
and design lighting systems.
And eventually we became a bigdata aggregator in doing that
(01:28):
and saw gaps in the marketplacefor certain lighting products
and then patented those and gotinto manufacturing.
So now we were a softwarecompany and a manufacturer and
we were doing that across theglobe, which meant.
For all of you accountants thatare listening, we had the
world's perfect storm ofaccounting challenges, deferred
rev rack, complicated assetdepreciation schedules, foreign
(01:50):
currency conversions, multientity, and of course, whip and
bomb and, you know, inventory,you name it.
So the business grew fast.
Um, Our accounting couldn't keepup.
It was a constant struggle.
And so when that company gotacquired, I was like, I'm going
to solve, uh, I'm going to makea virtual finance department for
(02:11):
businesses like me.
did that with a husband andwife, uh, um, who had started
smart books.
So I teamed up with them andinvested in, uh, their practice.
It was like four or five, six,seven people, something like
that at the time.
Then a few years later we were,uh, you know, five plus million
dollar CAS practice.
(02:32):
Um, and that was just CAS.
So that was like big at thetime.
We were fast growing and then wewere struggling with all the
same issues like Multiple appsthat don't talk to each other, a
G sheet to kind of keep track ofwho's working on what accounts
and what apps they're using.
You lose someone and became wayharder to find and hire people
like in a very short period oftime, there's like the
(02:52):
accountants, um, missing.
And, uh, the idea of like, itwas just like, God, it would be
so amazing.
Like there, every other industryhas one platform with all the
feature and functionality youneed to do the job, but CAS.
And so that would be great ifyou could have.
Those features in a singlesystem.
(03:13):
And if it wasn't a singlesystem, all the data would,
you'd have all the data and allthe data would tie out.
And with all that data, um, youcould build really powerful
machine learning and AI.
And if you could do that, youcould automate away a lot of the
processing, which would alsohelp solve the, uh, you know,
human shortage.
And that was the, the impetusfor, for Bachiever.
Davis Bell (03:36):
So to kind of summarize, you start this
business in college thatintroduces you to the pain, um,
of sort of dealing withaccounting firms from the, from
the entrepreneurial.
sort of perspective.
Then you went to go try to solvethat by, by becoming a partner
(03:59):
in a firm and then youdiscovered all the pains
associated with, with running afirm.
And so that, then you went andstarted, so it's your next
business about, you know, thepains or experiences as a tech
company or you just, you'regoing to stay put here.
Enrico Palmerino (04:12):
I have a couple ideas for solving pains
with hiring, recruiting and allthat cause of having like a high
growth company.
You know, a company like oursthat wasn't.
A huge challenge.
Um, but there's plenty to do yetto come with bought deeper.
So, uh, you know, maybe, maybein a decade from now.
Davis Bell (04:32):
Well, I think, um, you know, I, I think my
perspective is the productsbecome great when they're being
solved by people who've reallyunderstood the pain, uh,
associated with the users ofthose products.
And sometimes you can just goand figure that out.
And sometimes you experiencethem, um, you know, in, in your
own life, which Something that,that you did.
(04:54):
So kind of honing in on botkeeper a little bit.
You, you mentioned some of thoseproblems.
It sounded like, you know,disparate systems data that's
not, you know, talking to, toeach other.
Um, the, the challenge of, uh,bringing new hires up to speed.
(05:15):
Would you, would you considerthose to be the key challenges
or would you, are there othersthat, that you would mention
that drove to kind of the, thefounding.
Enrico Palmerino (05:23):
I, I think it, it, it all like in a weird way
is entangled, right?
So disparate systems and havinga system switch between a half a
dozen or a dozen apps to dobookkeeping for a single client
times 20 clients.
That's a whole lot of systemswitches on a given day or on a
monthly basis.
It's also hold a lot of securityvulnerabilities.
(05:44):
Um, that you're opening yourselfup to like that much more
potential for margin erosion asevery one of those apps, you
know, increases cost.
Um, and then the, the data tieout and the siloed datasets
becomes more and more of a majorproblem when you want to start
leveraging AI.
So.
If you end up having a bunch ofniche AI applications that don't
(06:08):
talk to each other and requiresome sort of syncing of data,
then you're wondering like, isthe AI making good decisions
based on bad data or is itmaking decisions on yesterday's
data instead of today's data?
And does the data actually tieout and therefore is the
decision the right one to bemade?
We all encounter this on a dailybasis when we go to like have a
quarterly board meeting orexecutive meeting where like two
(06:31):
departments are presentingdifferent data numbers.
But imagine like the complexityand the exacerbation when the
speed of reaction is automatedor AI is doing it.
And then do you have visibilityinto what data set was used to
make the conclusion?
And you could just kind of seelike this, this compounds.
And then the other part of itis, um, stress.
(06:53):
Cause like in a world where likeyou can't find and hire people,
then the people you do hire, youwant to make sure that you take
care of them and that theyreally like what they're doing
and they're not being likeunnecessarily like burnt out or
stressed out.
And when you have a shortage ofpeople, that means more work for
the people you do have.
And.
I feel like accounting is justone of those things where like
(07:15):
the second you come up for air,like to breathe at the end of
the month, like you've got tostart over in the next month and
you're doing the next cycle.
And if you lose a person or anemployee, um, or someone takes
vacation or sick time, it's likethe whole world can come
crashing down.
And so the other aspect to likeone of the big things we want to
do with Blockkeeper was can weautomate away as much of this
(07:37):
like processing?
That needs to be done or thatcould be done on a daily basis
so that you could lose anemployee and someone could cover
the gap pretty easily becausethey have to only be trained or
brought up to speed on like 20percent of the transactions or
the reconciliation happensautomatically in the statements
fetched and or the month endreview.
(07:59):
Like the AI is showing you whatyou need to touch or treat and
address and instead of saving itall for the month end and
Compressing the time frame to dothe work.
What if like you do that on adaily basis?
There's no reason you shouldn'tbe able to sure and like all of
that was just like more and morereason like Someone's got to
figure this out
Davis Bell (08:17):
Yeah
Enrico Palmerino (08:18):
And the daunting part was you know as I
started like look at I'm like Ibasic gotta build We have built
one company that builds 10companies worth of technology
and features and functions.
And now we're going to pull thatoff.
But um, as we got into it, Ithink we just looked at it like,
(08:39):
I always use Excel as like theexample because 99 percent of
Excel users use 1 percent ofExcel.
So I just need to figure outlike of these functionalities,
like what were the most criticaland important ones.
And then get rid of a lot of thenoise.
Um, and if I did that, you know,there is a chance that we could
actually build 10 companiesworth of tech, um, under one
company umbrella.
Davis Bell (09:00):
Yeah.
I mean, I, um, I think, youknow, there's, there's sort of,
you know, in software, there's,there's sort of two approaches,
right?
The first is you take this verynarrow point solution approach,
you know, you're going to dothis very narrow thing.
And that's really appealing.
Um, initially because it's justmuch easier to get off the
ground.
It's easier to be billed.
(09:20):
It's to build.
It's easier to be good at onething, but over time you then
run in, you've created thisproblem for your customers,
which is then they need to gobuy a bunch of other things.
And I think, you know, howyou're, the way you're talking
is very similar to the kind ofapproach we've taken here, which
is.
We're not just going to be thatone thing.
We're going to try to be kind ofeverything.
And, you know, there's a trendyname for this in Silicon Valley
(09:44):
currently, which is the compoundstartup, which whenever someone
puts a new name on it, you know,they've invented it.
Obviously you've been at thisfor a while.
We have, um, but it's harder,right?
It's see, I just call it thesuite,
Enrico Palmerino (09:56):
which suite's been around for a long time.
Exactly.
Yeah.
Meta suite has been around andMicrosoft suite,
Davis Bell (10:03):
like, right.
Enrico Palmerino (10:04):
But suites stick, so, you know, you can
round out the suite, has alongevity and lasting power.
Whereas I think like the nichecompany, your ability to expand
is difficult, and that's whymost niches get acquired and put
into a suite.
Davis Bell (10:20):
Yeah, I, I, um, you know, the, the, it's sort of
like, you know, um, you, you,you, you live, I'm trying to
remember the exact phrase, butit's sort of like you, You know,
the phrase you either, um, die ahero or live long enough to
become a villain, it's like youeither die at point solution or
live long enough to becomeplatform.
(10:40):
Right.
So, so tell me, tell me justfrom like the, you know,
entrepreneurial heroes journey,because then I want to, I want
to talk a lot about AI, which,which you've already sort of
mentioned.
Um, what, what was the hardestthing you, you know, you set
out, you, you bit off a lot,right?
You're not just going to take apoint solution approach.
You're going to try to buildthis.
(11:01):
this platform.
Um, what, what was the, what wasthe hardest thing about doing
it?
Enrico Palmerino (11:08):
I think the most difficult thing about the
Bok, about Bokyem as a conceptis you You had to build a plane
while you were flying it.
There was no, there was no suchthing because AI requires data
and the AI and the data we needto do automated accounting is
private financial data.
(11:30):
And it's exactly that it'sprivate.
There's no like mass data setsomewhere of private company
data.
Um, and you can get like somepublic company data out there,
um, but that doesn't helpbecause like small businesses
are structured very different.
Their charts of accounts,everything's totally different
than the, those on wall street.
So we needed to acquire clientsin order to acquire data sets,
(11:54):
in order to build.
And then we knew that like, wedidn't want to be in the direct
SMB market because like theeconomics and AI world, the
winner is the, who has the mostdata and like, you can't get the
most data if you're doing one toone sales.
So we needed to find a way toget one to many, um, in the
(12:14):
accounting channel.
So this, like the difficultaspect was we had to go.
Create enough value to have abusiness to get people to sign
up and then using that to getthe data to build dream product
that we wanted to and then tolike Go from building the dream
product that we wanted to andhaving to like do it direct us
(12:35):
and be So you get the proofpoints so that the you know
world's laggard pessimistic techadopters known as accountants
would eventually be like trustthe product enough to be able to
buy it and And use it in theirpractices on their clients,
which is a risky thing thatProcess and that complex
(12:55):
complexity of you know, thistale of two cities and two
worlds Took eight years like ittook a lot of money.
It took eight years of time But,you know, we finally got there
and we launched infinite and nowit's like, you know, I talked to
our investors and I'm like,you've been at botkeeper almost
arguably it's almost 10 yearsnow, like next year we'll cross
(13:18):
the 10 year mark.
And they're like, do you stillgot it to keep going?
And I'm like, yeah, you guysdon't understand the last eight
years or nine years where theprelude, like this is like
infinite launch.
This is chapter one, you know?
We're, we're amped up and readyto go.
And, and we see it too.
And, you know, it took us eightyears to, I think, acquire a
(13:39):
couple hundred accounting firms.
It took us, uh, three quartersto acquire more than that.
Um, since, since then, so it'sexciting.
Davis Bell (13:47):
I read this quote that was talking about, um,
Ultimately, your only moat,competitive moat is obsession,
um, because people who areobsessed with a problem are
willing to go spend eight years,right?
Like if it's someone who's justtrying to build a little app and
flip it to some big company and,you know, make some money.
You know, they're just not goingto do that, right?
(14:08):
They're not going to spend eightyears, um, suffer all, all the
pain and, and build this slowbuild and you know, that you,
you can build a generationalcompany that way, but, um, you
know, it's, it's a slow burn,right?
And the, the platform approachesa slow burn too.
Like it's just the nature of it.
Um, but to your point, like.
(14:29):
Once you've done it, you know,you, you can really camp out
there for decades cause it'sjust super hard for anyone else
to come along.
And
Enrico Palmerino (14:36):
I mean, that's the moat, like, you know, people
are like, well, what's stopping?
Like, I'll tell you what'sstopping.
Eight years of development, um,a hundred and something million
dollars and billions oftransactions that we needed to,
billions of transactions, tensof thousands of companies and
hundreds of accounting firms ofdata and transactions to build
(14:57):
what we built.
And the unique thing about thisis like when we first started,
The only thing we were competingagainst in terms of like client
and data acquisition was likeold school human labor
accounting, which meant we couldcharge three 99 a month per
(15:21):
client to do this.
And like, that was a third or aquarter of the cost of like the
labor.
Today, BotKeeper sells that samething for 69 bucks a month.
So, so anyone today that's like,Hey, I'm going to go do, I'm
just going to follow BotKeeper'splaybook, do the exact same
thing.
You'd hemorrhage so many moredollars than we did.
(15:44):
Because you'd have to be, like,you'd have to start with human
labor to process and get thereand train the models.
And you'd have to do it at sucha loss.
Like, you know, you know, youknow, like we know we were
barely breaking even or had likeminimal gross margins and we
were selling it at three 99until the tech caught up, like
(16:05):
if you had to come in and startselling it at 69 today and you
were doing it by human labor.
Like you just, you'd have to,you know, raise two or three
times the amount of money thatwe did in order to pull it off
and get to the same point.
So it's a good place to be.
It's just, it's a hard place tobe.
But like you said, I'm, I'mobsessed with it.
I like, we, we love what we'redoing.
(16:26):
Uh, you know, we've got a greatteam, uh, that makes it a lot of
fun to work with.
And all of us, like, you know,we've kind of been working to
the, for this thing for like solong and like see it come out is
just like, yeah, it's fun.
All right, now let's go.
Davis Bell (16:38):
Well, yeah, I mean it's it's a nice thing to be on
the other side of, right?
I mean eight years of kind ofslugging it away and and you
know slowly acquiring firms andthen finally you get to this
inflection point like it justthere's there's there's nothing
better.
That's why they say it's a
Enrico Palmerino (16:55):
ten year overnight success.
Davis Bell (16:57):
Yeah Exactly.
And yeah, I mean it really istrue, right?
I mean, it's it's you get theseflash in the pan You You know,
companies, but they're justexceedingly rare.
And the minute you dig into it,it's just, you know, there was,
there was that like decade ofjust sort of building and
grinding and failure anditeration.
(17:18):
Just kind of how it goes.
Um, yeah, so I think they saylike, it takes like
Enrico Palmerino (17:22):
on average 15 years for a company to get to,
if you can get to a hundredmillion dollars, those who did
it, it takes them like 15 years.
And there's been like a handfulin like the history of time, a
handful of exceptions.
Who have done in like eight,nine or 10.
Davis Bell (17:38):
Right.
Enrico Palmerino (17:40):
So
Davis Bell (17:40):
yeah, open, open AI is probably, you know, a more
recent, uh, exception.
And even Slack,
Enrico Palmerino (17:45):
Slack was one of them.
Davis Bell (17:47):
Right.
Um, yeah, but there's not,there's just not that many of
them, um, to your point.
So you, you mentioned, you know,AI it's, it's, you know,
impossible to have aconversation about either
accounting or software andcertainly accounting software
without talking about AI.
Um, So tell me, tell me first,you, you, um, you made a point
(18:12):
with which I violently agreearound AI and it's the ability
to leverage AI.
Relative to being a platform,um, versus a point solution.
So talk, talk, talk to us alittle bit about that.
Enrico Palmerino (18:29):
So, um, to me, I look at like the future of AI
is where it's going to be verydifferent than the AI we use
today.
Like the future of AI, a singleprompt delivers task completion.
Um, and, and that's not rightnow, like right now you're a
prompt is completing like anaspect of a task and like an
(18:53):
arguably like a tiny component,like if you consider bookkeeping
as the task, like your prompt iscompleting one sub segment of
like is actually completing a,an aspect of a sub segment of
bookkeeping.
So, and that, that works ish,because previously, like with an
(19:14):
app stack, you have peopleplaying traffic cops in terms of
figuring out organization andtiming and sequence, like you go
into the app, you do the thing,you take the completed aspect of
it, you move it into the nextapp, you do the next part of the
equation, take it out and againand again.
But in an AI world, you have allof these things that are going
to be ideally idealisticallyhappening automatically.
(19:36):
But if they can't talk to eachother and you see this all the
time and like marketingagencies, and that's all they do
marketing, they have a hard timegetting the data actually tie
out.
If you've got all thesedisparate apps that aren't
integrated with each other andthere's no, I can't see that
ever being fixed.
There's 1600 apps in the zeroand QBO marketplaces.
The cost for one company tointegrate with all 1600, not
(19:58):
going to be the case.
So you're going to have allthese like siloed data sets,
siloed apps, performing siloedtasks and functions that have no
purview or context into anythingelse that's going on.
And that just creates a world ofhurt.
Like, because you're not likethe, basically the, the, the
data traffic collides.
(20:20):
Did the thing get done on timeor did it get completed before
like the other component wascompleted and I had the
remaining data that it needed tocomplete its own task and what
was used or was the context thatgoes into the completion of a
task.
And that's where I think likethis, there's a reason why the
biggest companies in the worldthat are the winning, the AI
race are suites.
(20:40):
Like Meta is acquiring companiesthat will round out its suite to
give it the most context for itsAI, whether it's Instagram or
WhatsApp.
Um, or, uh, uh, Oculus, likethey're, they're trying to take
all the components that youwould need to create a very
powerful AI platform.
(21:03):
Same thing with Microsoft.
Even like zoom is trying toexpand its offering to get into
it.
But Microsoft was like, Hey, wealready have most of the
business suite offering and thenvideo and, you know, met
conferencing, all that, like isrounding it out.
Google is another one.
Open, look at what open is doingand how they're expanding.
But you know, they're at the,the, the.
(21:25):
For us, like, I think that'sgoing to be one of the
advantages we have down the roadis having everything in a
singular suite allows the AI tohave context, which allows it to
do like one of the new featureswe, um, have in beta right now
that's coming out at the end ofthe year is insights.
Where you can ask our AI,anything you want to know about
your firm, how to be moreefficient and it can look at all
(21:48):
the data that's there and comeup with that.
Who's the top performer whocould take on more clients?
Like where are bottlenecks inyour process?
And it will.
be able to see and deliver theanswer to those, and in an ideal
world, you could tell it tojust, like, fix that.
Right?
And it would just go and fix, ordo, like, go find me another six
(22:09):
hours of time savings and makeit happen.
Um, and it can do it, and youcan't do that without context.
Um, and I think, you know,ultimately context is rounded
out data sets, and Uh, andintegrated data sets where the
data actually ties out perfectlybecause if it doesn't and you
(22:30):
have these AI tools that areproducing an immense amount of
data and ingesting an immenseamount of data, you just start
to have, like, we've all seenthis where your database gets
riddled with bad data and youend up having to, like,
basically throw it out, like, orhit pause entirely and do a
massive cleanup project and thatcleanup project becomes near
(22:51):
impossible when You know, genesequencers and like in a totally
unrelated space when they go offthe wall, like someone creates a
prompt, um, that has themsequencing, like out of order or
putting it in the improperfolder, they potentially create
petabytes worth of data thatends up needing to be scrapped
(23:13):
because it's and redone becauseit just can't be fit into the
database in the right way.
And there's no way to clean itup.
Cause it's just too much dataand it's easier to start over.
So I think, uh, I think theworld's really exciting for AI,
but I think you're going to seethis trend of those players who
are niche apps trying to roundout, um, and be more of a suite
(23:36):
or more comprehensive orwholesome in their area.
Um, and you're going to seeprobably more suites emerge in
general in the, the accountingsector.
And you'll see acquisitions.
I think you'll, now that themarkets are starting to open up
again, you're going to see theplayers that are in there, like
merging together to, you know,Create the suite or to round out
(23:57):
the missing components.
Davis Bell (23:59):
Yeah, no, I, I mean, I think you very articulately
explained how we think about itand how I think about it.
I just, I, you know, typically anew innovation favors, um, The
scrappy, you know, new, newentrant, right?
Because the, you know, the, the,the, the person who's already
out there shipping product justhas a hard time innovating,
(24:20):
taking advantage of it.
But I think the nature of AI issuch that its value comes from
data at the end of the day.
I mean, that, that is the fuelfor AI, right?
Like if you just plug AI intothe software and there's just
nothing there.
It, you know, it like I sort ofthink about this framework is
(24:42):
sort of like, you know, uh, dataand action.
So like, but the action has tobe preceded by data, right?
Like it, it doesn't know whataction to take.
The, the, the action will bebased upon, you know, in our
context, it's like, all right,well, what, what documents did
(25:02):
they submit last tax season?
Right.
Well, therefore these are thedocuments they're going to need
next season.
And if it was blank, the AI isnot going to know what, like,
right.
There's no pattern
Enrico Palmerino (25:15):
to recognize.
Right.
Yeah.
Davis Bell (25:17):
Yeah.
So it, it, it, and I agree withyou too, that, um, where we're
at today, you know, it, it, tome, like one, one way to think
about it, that, that resonateswith me, I think about AI today
as like a not super smartintern.
So it's like, you know, ifyou've ever worked with an
intern, you're like, Hey, canyou go like, Google a list and
(25:40):
like come back and they'll comeback and you're like, well,
that's not what I wanted.
And you send them back and maybethe third time you're like,
okay, finally, and you're, andyou're sort of like, did I
actually really save any time?
I mean, I, I'm trying to helpthis intern out, you know, get
some experience, but maybe Ididn't save tons of time and
maybe the first time you did.
And, you know, so it's sort ofnets out, but there's a big
difference between that andjust.
(26:01):
You know, working with aseasoned career professional who
you can just pick up, you know,slack them or text them or call
them and be like, I need you todo this complex thing.
And they just come back and it'sdone.
And I think that we're in thedumb intern phase, but the rate
of progression is such that Ithink, you know, and that's,
(26:22):
this is what they call agenticAI, where it's like just doing
it, you know, maybe at a promptand maybe later, just not, it's
just recognizing it needs to bedone.
It's hard to know exactly whenthat's going to come to
fruition, but I think, yeah, Ithink we're, I think we're
seeing the, the opportunity andthe near term future very
similarly.
Enrico Palmerino (26:40):
Yeah.
And it's, um, to your point,the, the AI revolution, I think
most of the benefit or valuethat we're going to see out of
this is going to be theincumbents will become stronger,
um, because they're the ones whohave the vast data sets, the new
players, to your point.
One, haven't built their own AI.
Like this is another thing too.
I always point out most of thesenew AI companies, they're just
(27:03):
leveraging open AI or anotherlike existing AI model, which in
general, all of those are alsoLLMs or generative AI, which
doesn't bode well foraccounting.
Like you can do some narrativestuff, but you want, if you want
task specific AI, like that justtakes a whole lot of coding and
you have to do that yourself.
(27:24):
And if you want high accuracyaccounting, you That's the only
way to do it.
Um, at least like that's whateveryone always, that's what
anyone says is the only way todo it today.
Does that change?
Maybe generative AI over time,um, can get better at it, but
like how it's architected, itdoesn't look like it's ever
going to head in that direction.
(27:44):
Cause it is general, um, andgeneral doesn't bode well for
specific, uh, accuracy that'sbased on like machine learning
versus based on, um, inventive,uh, Synopsis of, of, of
datasets.
So yeah, I think the big playersare definitely gonna have the
(28:05):
advantage.
The small players, it's a riskytime to use like a brand new AI
company because open AI made achange earlier this year that
killed 1200 AI companies justbecause they got rid of one of
their APIs.
Um, Historically, you never hadto worry about that.
If you're using a brand newstartup on the block, because
(28:26):
They built their own softwareand you just had to worry how
much runway they have.
Now, if it's an AI brand new,they call them, uh, on the
Silicon Valley, they call themrapper companies like you, you
just, you're, you're wrappingsomeone else's AI with like a
concept.
And that has like, there's likeno moat to that.
I don't think, um,
Davis Bell (28:45):
yeah,
Enrico Palmerino (28:45):
but, uh, but it'll be interesting.
I, you know, I think a lot'sgoing to get done and there's so
much unlocked potential that wejust, we weren't leveraging and
that companies were sitting onin their data.
Um, and it's just going to takethose incumbents to be willing
to invest and explore andchallenge the status quo.
Davis Bell (29:04):
Yeah.
Well, I agree with you andmaybe, maybe one last topic for
conversation.
We've been geeking out a littlebit on the more, the software
side of things.
Bringing it back to accounting,I think, you know, my, um, my
mental model for AI inaccounting has been, um, the
exosuit, and I first called thisa cyborg, but we've got some
(29:26):
nerds here on staff whocorrected me, and, you know, if
you think about the movieAvatar, the humans walk around
in those, they're inside thisrobot, and it can pick up a 8,
000 pound boulder or whatever,move really fast, and, you know,
You know, that's, that's sort ofhow I conceptualize AI.
Um, because I mean, so much, Idon't want to venture a guess,
(29:47):
but an enormous amount of the,the actual human labor in a
firm.
Is menial, low value add,manual, just sort of, you know,
pushing paper from one place toanother.
Parses.
Yeah, and I think people don'tlike to do that, and it's a bad
use of their time.
(30:07):
Or they like it
Enrico Palmerino (30:08):
for all the reasons we wish they didn't,
which is, it's easy, I don'thave to think.
Yeah, sure.
Monday through Friday and justkind of like wake up and get
paid.
Davis Bell (30:19):
Yeah.
So I think that like most peopleget into the profession cause
they're passionate aboutactually accounting and they
want to think strategically andthey want to help people
strategically and then they justget mired down in the weeds of
this stuff.
So what, what would you agree?
Disagree?
Would you modify that?
Like how, what's, what's yoursort of model for it?
Enrico Palmerino (30:41):
No, I agree.
I mean, for the longest time we,we used to, um, Uh, compare Bot
Keeper to, uh, Iron Man.
Tony Stark is nothing, really.
I mean, he's a brilliant person.
But like, he becomes a superherowhen he has his suit.
And we would see the A.
I.
s being a suit.
And on its own, it can't domuch.
(31:03):
And like, even like the Iron Mansuit, it requires Tony to
command it, or direct it, or dostuff.
And that is the differencebetween, arguably, like, A.
I.
actually being an intern, Anintern, conceptually, like,
while it might, uh, he or shemight take, like, you know,
directions from you to do stuff.
The reality is, that person canalso Go do things for you
(31:25):
without you even asking becausethey recognize they're like self
starting Take initiative.
They recognize a problem.
They attempt to solve and youwe've all had those interns who
are like Hey, by the way, I didthis because I thought you might
like it like, oh, that's coolLike we don't need it But like
the self starting nature of andthe fact that did is like really
cool Ai can't do that right now.
Ai is only going to do thethings you specifically tell it
(31:47):
to Every time you tell it to dosomething so until it can start
to like problem solve for youwithout It is going to be an
exoskeleton or, you know, theIron Man suit that's just
amplifying, uh, the intelligenceand the use of our people.
And arguably, it's like, kind oflike the Iron Man suit.
If you took You took, uh, your,your weakest, like, you know,
(32:13):
engineer, like, you know, yourweakest Tony Stark from a
intelligence standpoint, you putthem in the suit, he'll be able
to do good stuff, like pickthings up and fly.
You take a brilliant person, youput them in the suit and they
can accomplish like insanethings.
And I think we're going to seethe same thing happen with AI.
The smartest people out there,the best accountants are going
to take accounting up notchesand levels that like we've never
(32:36):
thought possible.
And those who are not.
The, you know, strongestaccountants, they're going to
now be able to like be like thestrongest accountants or be a
really great senior accountant,thanks to AI, but they're not
going to, they're still notgoing to play.
It's not like, it's not like theall, uh, intensive, like plain
(32:58):
feel leveler, uh, that somepeople think it is, there is
varying degrees of, of skillsetand use that, that amplify its
outputs.
Davis Bell (33:07):
Yeah, no, I totally agree.
Hey Enrico, this has been superfun.
We'd love to do it again.
Um, and just encourage everyoneif they want to know more about
Botkeeper, botkeeper.
com super cool company, greatproducts, great leadership, and
I'm excited to see what, whatyou guys get up to next.
Enrico Palmerino (33:26):
Thank you so much, sir.
I appreciate it.
Davis Bell (33:28):
Thanks Enrico.
Hello and welcome to Canopy's Practice Success
Podcast.
I'm Davis Bell, the CEO ofCanopy, and I will be sitting in
as host for today's podcast.
I, I'm excited.
They don't usually let me dothese things.
They don't let me out in publicoften, so.
Um, I don't know what happened,but, uh, glad, glad to be here.
I'm super excited to sit downwith Enrico Palmarino, who's the
(00:28):
CEO of Botkeeper.
And today we're going to chatabout Botkeeper and what the
accounting profession can learnfrom the world of tech and
venture capital, as well as wesee AI taking the profession.
So welcome, Enrico.
Great to be here with you.
Enrico Palmerino (00:40):
Thank you, Davis.
Appreciate it, sir.
Yeah, super
Davis Bell (00:43):
excited.
So, um, I, I thought maybe, um,And I've got two questions.
I think they're probably superrelated.
One, I just want to know aboutyour background prior to Bot
Keeper.
And my next question is whatinspired you to create Bot
Keeper?
And I think those are actuallyvery related.
So maybe, you know, you justtake them in whatever order you
want.
Enrico Palmerino (01:04):
So, uh, background.
Um, I was a quant major, uh, incollege and while at Babson, uh,
which is an entrepreneurialschool.
So, uh, everyone's alwaysstarting businesses there, which
is a lot of fun.
And I started a company thatautomated how you, uh, Analyze
and design lighting systems.
And eventually we became a bigdata aggregator in doing that
(01:28):
and saw gaps in the marketplacefor certain lighting products
and then patented those and gotinto manufacturing.
So now we were a softwarecompany and a manufacturer and
we were doing that across theglobe, which meant.
For all of you accountants thatare listening, we had the
world's perfect storm ofaccounting challenges, deferred
rev rack, complicated assetdepreciation schedules, foreign
(01:50):
currency conversions, multientity, and of course, whip and
bomb and, you know, inventory,you name it.
So the business grew fast.
Um, Our accounting couldn't keepup.
It was a constant struggle.
And so when that company gotacquired, I was like, I'm going
to solve, uh, I'm going to makea virtual finance department for
(02:11):
businesses like me.
did that with a husband andwife, uh, um, who had started
smart books.
So I teamed up with them andinvested in, uh, their practice.
It was like four or five, six,seven people, something like
that at the time.
Then a few years later we were,uh, you know, five plus million
dollar CAS practice.
(02:32):
Um, and that was just CAS.
So that was like big at thetime.
We were fast growing and then wewere struggling with all the
same issues like Multiple appsthat don't talk to each other, a
G sheet to kind of keep track ofwho's working on what accounts
and what apps they're using.
You lose someone and became wayharder to find and hire people
like in a very short period oftime, there's like the
(02:52):
accountants, um, missing.
And, uh, the idea of like, itwas just like, God, it would be
so amazing.
Like there, every other industryhas one platform with all the
feature and functionality youneed to do the job, but CAS.
And so that would be great ifyou could have.
Those features in a singlesystem.
(03:13):
And if it wasn't a singlesystem, all the data would,
you'd have all the data and allthe data would tie out.
And with all that data, um, youcould build really powerful
machine learning and AI.
And if you could do that, youcould automate away a lot of the
processing, which would alsohelp solve the, uh, you know,
human shortage.
And that was the, the impetusfor, for Bachiever.
Davis Bell (03:36):
So to kind of summarize, you start this
business in college thatintroduces you to the pain, um,
of sort of dealing withaccounting firms from the, from
the entrepreneurial.
sort of perspective.
Then you went to go try to solvethat by, by becoming a partner
(03:59):
in a firm and then youdiscovered all the pains
associated with, with running afirm.
And so that, then you went andstarted, so it's your next
business about, you know, thepains or experiences as a tech
company or you just, you'regoing to stay put here.
Enrico Palmerino (04:12):
I have a couple ideas for solving pains
with hiring, recruiting and allthat cause of having like a high
growth company.
You know, a company like oursthat wasn't.
A huge challenge.
Um, but there's plenty to do yetto come with bought deeper.
So, uh, you know, maybe, maybein a decade from now.
Davis Bell (04:32):
Well, I think, um, you know, I, I think my
perspective is the productsbecome great when they're being
solved by people who've reallyunderstood the pain, uh,
associated with the users ofthose products.
And sometimes you can just goand figure that out.
And sometimes you experiencethem, um, you know, in, in your
own life, which Something that,that you did.
(04:54):
So kind of honing in on botkeeper a little bit.
You, you mentioned some of thoseproblems.
It sounded like, you know,disparate systems data that's
not, you know, talking to, toeach other.
Um, the, the challenge of, uh,bringing new hires up to speed.
(05:15):
Would you, would you considerthose to be the key challenges
or would you, are there othersthat, that you would mention
that drove to kind of the, thefounding.
Enrico Palmerino (05:23):
I, I think it, it, it all like in a weird way
is entangled, right?
So disparate systems and havinga system switch between a half a
dozen or a dozen apps to dobookkeeping for a single client
times 20 clients.
That's a whole lot of systemswitches on a given day or on a
monthly basis.
It's also hold a lot of securityvulnerabilities.
(05:44):
Um, that you're opening yourselfup to like that much more
potential for margin erosion asevery one of those apps, you
know, increases cost.
Um, and then the, the data tieout and the siloed datasets
becomes more and more of a majorproblem when you want to start
leveraging AI.
So.
If you end up having a bunch ofniche AI applications that don't
(06:08):
talk to each other and requiresome sort of syncing of data,
then you're wondering like, isthe AI making good decisions
based on bad data or is itmaking decisions on yesterday's
data instead of today's data?
And does the data actually tieout and therefore is the
decision the right one to bemade?
We all encounter this on a dailybasis when we go to like have a
quarterly board meeting orexecutive meeting where like two
(06:31):
departments are presentingdifferent data numbers.
But imagine like the complexityand the exacerbation when the
speed of reaction is automatedor AI is doing it.
And then do you have visibilityinto what data set was used to
make the conclusion?
And you could just kind of seelike this, this compounds.
And then the other part of itis, um, stress.
(06:53):
Cause like in a world where likeyou can't find and hire people,
then the people you do hire, youwant to make sure that you take
care of them and that theyreally like what they're doing
and they're not being likeunnecessarily like burnt out or
stressed out.
And when you have a shortage ofpeople, that means more work for
the people you do have.
And.
I feel like accounting is justone of those things where like
(07:15):
the second you come up for air,like to breathe at the end of
the month, like you've got tostart over in the next month and
you're doing the next cycle.
And if you lose a person or anemployee, um, or someone takes
vacation or sick time, it's likethe whole world can come
crashing down.
And so the other aspect to likeone of the big things we want to
do with Blockkeeper was can weautomate away as much of this
(07:37):
like processing?
That needs to be done or thatcould be done on a daily basis
so that you could lose anemployee and someone could cover
the gap pretty easily becausethey have to only be trained or
brought up to speed on like 20percent of the transactions or
the reconciliation happensautomatically in the statements
fetched and or the month endreview.
(07:59):
Like the AI is showing you whatyou need to touch or treat and
address and instead of saving itall for the month end and
Compressing the time frame to dothe work.
What if like you do that on adaily basis?
There's no reason you shouldn'tbe able to sure and like all of
that was just like more and morereason like Someone's got to
figure this out
Davis Bell (08:17):
Yeah
Enrico Palmerino (08:18):
And the daunting part was you know as I
started like look at I'm like Ibasic gotta build We have built
one company that builds 10companies worth of technology
and features and functions.
And now we're going to pull thatoff.
But um, as we got into it, Ithink we just looked at it like,
(08:39):
I always use Excel as like theexample because 99 percent of
Excel users use 1 percent ofExcel.
So I just need to figure outlike of these functionalities,
like what were the most criticaland important ones.
And then get rid of a lot of thenoise.
Um, and if I did that, you know,there is a chance that we could
actually build 10 companiesworth of tech, um, under one
company umbrella.
Davis Bell (09:00):
Yeah.
I mean, I, um, I think, youknow, there's, there's sort of,
you know, in software, there's,there's sort of two approaches,
right?
The first is you take this verynarrow point solution approach,
you know, you're going to dothis very narrow thing.
And that's really appealing.
Um, initially because it's justmuch easier to get off the
ground.
It's easier to be billed.
(09:20):
It's to build.
It's easier to be good at onething, but over time you then
run in, you've created thisproblem for your customers,
which is then they need to gobuy a bunch of other things.
And I think, you know, howyou're, the way you're talking
is very similar to the kind ofapproach we've taken here, which
is.
We're not just going to be thatone thing.
We're going to try to be kind ofeverything.
And, you know, there's a trendyname for this in Silicon Valley
(09:44):
currently, which is the compoundstartup, which whenever someone
puts a new name on it, you know,they've invented it.
Obviously you've been at thisfor a while.
We have, um, but it's harder,right?
It's see, I just call it thesuite,
Enrico Palmerino (09:56):
which suite's been around for a long time.
Exactly.
Yeah.
Meta suite has been around andMicrosoft suite,
Davis Bell (10:03):
like, right.
Enrico Palmerino (10:04):
But suites stick, so, you know, you can
round out the suite, has alongevity and lasting power.
Whereas I think like the nichecompany, your ability to expand
is difficult, and that's whymost niches get acquired and put
into a suite.
Davis Bell (10:20):
Yeah, I, I, um, you know, the, the, it's sort of
like, you know, um, you, you,you, you live, I'm trying to
remember the exact phrase, butit's sort of like you, You know,
the phrase you either, um, die ahero or live long enough to
become a villain, it's like youeither die at point solution or
live long enough to becomeplatform.
(10:40):
Right.
So, so tell me, tell me justfrom like the, you know,
entrepreneurial heroes journey,because then I want to, I want
to talk a lot about AI, which,which you've already sort of
mentioned.
Um, what, what was the hardestthing you, you know, you set
out, you, you bit off a lot,right?
You're not just going to take apoint solution approach.
You're going to try to buildthis.
(11:01):
this platform.
Um, what, what was the, what wasthe hardest thing about doing
it?
Enrico Palmerino (11:08):
I think the most difficult thing about the
Bok, about Bokyem as a conceptis you You had to build a plane
while you were flying it.
There was no, there was no suchthing because AI requires data
and the AI and the data we needto do automated accounting is
private financial data.
(11:30):
And it's exactly that it'sprivate.
There's no like mass data setsomewhere of private company
data.
Um, and you can get like somepublic company data out there,
um, but that doesn't helpbecause like small businesses
are structured very different.
Their charts of accounts,everything's totally different
than the, those on wall street.
So we needed to acquire clientsin order to acquire data sets,
(11:54):
in order to build.
And then we knew that like, wedidn't want to be in the direct
SMB market because like theeconomics and AI world, the
winner is the, who has the mostdata and like, you can't get the
most data if you're doing one toone sales.
So we needed to find a way toget one to many, um, in the
(12:14):
accounting channel.
So this, like the difficultaspect was we had to go.
Create enough value to have abusiness to get people to sign
up and then using that to getthe data to build dream product
that we wanted to and then tolike Go from building the dream
product that we wanted to andhaving to like do it direct us
(12:35):
and be So you get the proofpoints so that the you know
world's laggard pessimistic techadopters known as accountants
would eventually be like trustthe product enough to be able to
buy it and And use it in theirpractices on their clients,
which is a risky thing thatProcess and that complex
(12:55):
complexity of you know, thistale of two cities and two
worlds Took eight years like ittook a lot of money.
It took eight years of time But,you know, we finally got there
and we launched infinite and nowit's like, you know, I talked to
our investors and I'm like,you've been at botkeeper almost
arguably it's almost 10 yearsnow, like next year we'll cross
(13:18):
the 10 year mark.
And they're like, do you stillgot it to keep going?
And I'm like, yeah, you guysdon't understand the last eight
years or nine years where theprelude, like this is like
infinite launch.
This is chapter one, you know?
We're, we're amped up and readyto go.
And, and we see it too.
And, you know, it took us eightyears to, I think, acquire a
(13:39):
couple hundred accounting firms.
It took us, uh, three quartersto acquire more than that.
Um, since, since then, so it'sexciting.
Davis Bell (13:47):
I read this quote that was talking about, um,
Ultimately, your only moat,competitive moat is obsession,
um, because people who areobsessed with a problem are
willing to go spend eight years,right?
Like if it's someone who's justtrying to build a little app and
flip it to some big company and,you know, make some money.
You know, they're just not goingto do that, right?
(14:08):
They're not going to spend eightyears, um, suffer all, all the
pain and, and build this slowbuild and you know, that you,
you can build a generationalcompany that way, but, um, you
know, it's, it's a slow burn,right?
And the, the platform approachesa slow burn too.
Like it's just the nature of it.
Um, but to your point, like.
(14:29):
Once you've done it, you know,you, you can really camp out
there for decades cause it'sjust super hard for anyone else
to come along.
And
Enrico Palmerino (14:36):
I mean, that's the moat, like, you know, people
are like, well, what's stopping?
Like, I'll tell you what'sstopping.
Eight years of development, um,a hundred and something million
dollars and billions oftransactions that we needed to,
billions of transactions, tensof thousands of companies and
hundreds of accounting firms ofdata and transactions to build
(14:57):
what we built.
And the unique thing about thisis like when we first started,
The only thing we were competingagainst in terms of like client
and data acquisition was likeold school human labor
accounting, which meant we couldcharge three 99 a month per
(15:21):
client to do this.
And like, that was a third or aquarter of the cost of like the
labor.
Today, BotKeeper sells that samething for 69 bucks a month.
So, so anyone today that's like,Hey, I'm going to go do, I'm
just going to follow BotKeeper'splaybook, do the exact same
thing.
You'd hemorrhage so many moredollars than we did.
(15:44):
Because you'd have to be, like,you'd have to start with human
labor to process and get thereand train the models.
And you'd have to do it at sucha loss.
Like, you know, you know, youknow, like we know we were
barely breaking even or had likeminimal gross margins and we
were selling it at three 99until the tech caught up, like
(16:05):
if you had to come in and startselling it at 69 today and you
were doing it by human labor.
Like you just, you'd have to,you know, raise two or three
times the amount of money thatwe did in order to pull it off
and get to the same point.
So it's a good place to be.
It's just, it's a hard place tobe.
But like you said, I'm, I'mobsessed with it.
I like, we, we love what we'redoing.
(16:26):
Uh, you know, we've got a greatteam, uh, that makes it a lot of
fun to work with.
And all of us, like, you know,we've kind of been working to
the, for this thing for like solong and like see it come out is
just like, yeah, it's fun.
All right, now let's go.
Davis Bell (16:38):
Well, yeah, I mean it's it's a nice thing to be on
the other side of, right?
I mean eight years of kind ofslugging it away and and you
know slowly acquiring firms andthen finally you get to this
inflection point like it justthere's there's there's nothing
better.
That's why they say it's a
Enrico Palmerino (16:55):
ten year overnight success.
Davis Bell (16:57):
Yeah Exactly.
And yeah, I mean it really istrue, right?
I mean, it's it's you get theseflash in the pan You You know,
companies, but they're justexceedingly rare.
And the minute you dig into it,it's just, you know, there was,
there was that like decade ofjust sort of building and
grinding and failure anditeration.
(17:18):
Just kind of how it goes.
Um, yeah, so I think they saylike, it takes like
Enrico Palmerino (17:22):
on average 15 years for a company to get to,
if you can get to a hundredmillion dollars, those who did
it, it takes them like 15 years.
And there's been like a handfulin like the history of time, a
handful of exceptions.
Who have done in like eight,nine or 10.
Davis Bell (17:38):
Right.
Enrico Palmerino (17:40):
So
Davis Bell (17:40):
yeah, open, open AI is probably, you know, a more
recent, uh, exception.
And even Slack,
Enrico Palmerino (17:45):
Slack was one of them.
Davis Bell (17:47):
Right.
Um, yeah, but there's not,there's just not that many of
them, um, to your point.
So you, you mentioned, you know,AI it's, it's, you know,
impossible to have aconversation about either
accounting or software andcertainly accounting software
without talking about AI.
Um, So tell me, tell me first,you, you, um, you made a point
(18:12):
with which I violently agreearound AI and it's the ability
to leverage AI.
Relative to being a platform,um, versus a point solution.
So talk, talk, talk to us alittle bit about that.
Enrico Palmerino (18:29):
So, um, to me, I look at like the future of AI
is where it's going to be verydifferent than the AI we use
today.
Like the future of AI, a singleprompt delivers task completion.
Um, and, and that's not rightnow, like right now you're a
prompt is completing like anaspect of a task and like an
(18:53):
arguably like a tiny component,like if you consider bookkeeping
as the task, like your prompt iscompleting one sub segment of
like is actually completing a,an aspect of a sub segment of
bookkeeping.
So, and that, that works ish,because previously, like with an
(19:14):
app stack, you have peopleplaying traffic cops in terms of
figuring out organization andtiming and sequence, like you go
into the app, you do the thing,you take the completed aspect of
it, you move it into the nextapp, you do the next part of the
equation, take it out and againand again.
But in an AI world, you have allof these things that are going
to be ideally idealisticallyhappening automatically.
(19:36):
But if they can't talk to eachother and you see this all the
time and like marketingagencies, and that's all they do
marketing, they have a hard timegetting the data actually tie
out.
If you've got all thesedisparate apps that aren't
integrated with each other andthere's no, I can't see that
ever being fixed.
There's 1600 apps in the zeroand QBO marketplaces.
The cost for one company tointegrate with all 1600, not
(19:58):
going to be the case.
So you're going to have allthese like siloed data sets,
siloed apps, performing siloedtasks and functions that have no
purview or context into anythingelse that's going on.
And that just creates a world ofhurt.
Like, because you're not likethe, basically the, the, the
data traffic collides.
(20:20):
Did the thing get done on timeor did it get completed before
like the other component wascompleted and I had the
remaining data that it needed tocomplete its own task and what
was used or was the context thatgoes into the completion of a
task.
And that's where I think likethis, there's a reason why the
biggest companies in the worldthat are the winning, the AI
race are suites.
(20:40):
Like Meta is acquiring companiesthat will round out its suite to
give it the most context for itsAI, whether it's Instagram or
WhatsApp.
Um, or, uh, uh, Oculus, likethey're, they're trying to take
all the components that youwould need to create a very
powerful AI platform.
(21:03):
Same thing with Microsoft.
Even like zoom is trying toexpand its offering to get into
it.
But Microsoft was like, Hey, wealready have most of the
business suite offering and thenvideo and, you know, met
conferencing, all that, like isrounding it out.
Google is another one.
Open, look at what open is doingand how they're expanding.
But you know, they're at the,the, the.
(21:25):
For us, like, I think that'sgoing to be one of the
advantages we have down the roadis having everything in a
singular suite allows the AI tohave context, which allows it to
do like one of the new featureswe, um, have in beta right now
that's coming out at the end ofthe year is insights.
Where you can ask our AI,anything you want to know about
your firm, how to be moreefficient and it can look at all
(21:48):
the data that's there and comeup with that.
Who's the top performer whocould take on more clients?
Like where are bottlenecks inyour process?
And it will.
be able to see and deliver theanswer to those, and in an ideal
world, you could tell it tojust, like, fix that.
Right?
And it would just go and fix, ordo, like, go find me another six
(22:09):
hours of time savings and makeit happen.
Um, and it can do it, and youcan't do that without context.
Um, and I think, you know,ultimately context is rounded
out data sets, and Uh, andintegrated data sets where the
data actually ties out perfectlybecause if it doesn't and you
(22:30):
have these AI tools that areproducing an immense amount of
data and ingesting an immenseamount of data, you just start
to have, like, we've all seenthis where your database gets
riddled with bad data and youend up having to, like,
basically throw it out, like, orhit pause entirely and do a
massive cleanup project and thatcleanup project becomes near
(22:51):
impossible when You know, genesequencers and like in a totally
unrelated space when they go offthe wall, like someone creates a
prompt, um, that has themsequencing, like out of order or
putting it in the improperfolder, they potentially create
petabytes worth of data thatends up needing to be scrapped
(23:13):
because it's and redone becauseit just can't be fit into the
database in the right way.
And there's no way to clean itup.
Cause it's just too much dataand it's easier to start over.
So I think, uh, I think theworld's really exciting for AI,
but I think you're going to seethis trend of those players who
are niche apps trying to roundout, um, and be more of a suite
(23:36):
or more comprehensive orwholesome in their area.
Um, and you're going to seeprobably more suites emerge in
general in the, the accountingsector.
And you'll see acquisitions.
I think you'll, now that themarkets are starting to open up
again, you're going to see theplayers that are in there, like
merging together to, you know,Create the suite or to round out
(23:57):
the missing components.
Davis Bell (23:59):
Yeah, no, I, I mean, I think you very articulately
explained how we think about itand how I think about it.
I just, I, you know, typically anew innovation favors, um, The
scrappy, you know, new, newentrant, right?
Because the, you know, the, the,the, the person who's already
out there shipping product justhas a hard time innovating,
(24:20):
taking advantage of it.
But I think the nature of AI issuch that its value comes from
data at the end of the day.
I mean, that, that is the fuelfor AI, right?
Like if you just plug AI intothe software and there's just
nothing there.
It, you know, it like I sort ofthink about this framework is
(24:42):
sort of like, you know, uh, dataand action.
So like, but the action has tobe preceded by data, right?
Like it, it doesn't know whataction to take.
The, the, the action will bebased upon, you know, in our
context, it's like, all right,well, what, what documents did
(25:02):
they submit last tax season?
Right.
Well, therefore these are thedocuments they're going to need
next season.
And if it was blank, the AI isnot going to know what, like,
right.
There's no pattern
Enrico Palmerino (25:15):
to recognize.
Right.
Yeah.
Davis Bell (25:17):
Yeah.
So it, it, it, and I agree withyou too, that, um, where we're
at today, you know, it, it, tome, like one, one way to think
about it, that, that resonateswith me, I think about AI today
as like a not super smartintern.
So it's like, you know, ifyou've ever worked with an
intern, you're like, Hey, canyou go like, Google a list and
(25:40):
like come back and they'll comeback and you're like, well,
that's not what I wanted.
And you send them back and maybethe third time you're like,
okay, finally, and you're, andyou're sort of like, did I
actually really save any time?
I mean, I, I'm trying to helpthis intern out, you know, get
some experience, but maybe Ididn't save tons of time and
maybe the first time you did.
And, you know, so it's sort ofnets out, but there's a big
difference between that andjust.
(26:01):
You know, working with aseasoned career professional who
you can just pick up, you know,slack them or text them or call
them and be like, I need you todo this complex thing.
And they just come back and it'sdone.
And I think that we're in thedumb intern phase, but the rate
of progression is such that Ithink, you know, and that's,
(26:22):
this is what they call agenticAI, where it's like just doing
it, you know, maybe at a promptand maybe later, just not, it's
just recognizing it needs to bedone.
It's hard to know exactly whenthat's going to come to
fruition, but I think, yeah, Ithink we're, I think we're
seeing the, the opportunity andthe near term future very
similarly.
Enrico Palmerino (26:40):
Yeah.
And it's, um, to your point,the, the AI revolution, I think
most of the benefit or valuethat we're going to see out of
this is going to be theincumbents will become stronger,
um, because they're the ones whohave the vast data sets, the new
players, to your point.
One, haven't built their own AI.
Like this is another thing too.
I always point out most of thesenew AI companies, they're just
(27:03):
leveraging open AI or anotherlike existing AI model, which in
general, all of those are alsoLLMs or generative AI, which
doesn't bode well foraccounting.
Like you can do some narrativestuff, but you want, if you want
task specific AI, like that justtakes a whole lot of coding and
you have to do that yourself.
(27:24):
And if you want high accuracyaccounting, you That's the only
way to do it.
Um, at least like that's whateveryone always, that's what
anyone says is the only way todo it today.
Does that change?
Maybe generative AI over time,um, can get better at it, but
like how it's architected, itdoesn't look like it's ever
going to head in that direction.
(27:44):
Cause it is general, um, andgeneral doesn't bode well for
specific, uh, accuracy that'sbased on like machine learning
versus based on, um, inventive,uh, Synopsis of, of, of
datasets.
So yeah, I think the big playersare definitely gonna have the
(28:05):
advantage.
The small players, it's a riskytime to use like a brand new AI
company because open AI made achange earlier this year that
killed 1200 AI companies justbecause they got rid of one of
their APIs.
Um, Historically, you never hadto worry about that.
If you're using a brand newstartup on the block, because
(28:26):
They built their own softwareand you just had to worry how
much runway they have.
Now, if it's an AI brand new,they call them, uh, on the
Silicon Valley, they call themrapper companies like you, you
just, you're, you're wrappingsomeone else's AI with like a
concept.
And that has like, there's likeno moat to that.
I don't think, um,
Davis Bell (28:45):
yeah,
Enrico Palmerino (28:45):
but, uh, but it'll be interesting.
I, you know, I think a lot'sgoing to get done and there's so
much unlocked potential that wejust, we weren't leveraging and
that companies were sitting onin their data.
Um, and it's just going to takethose incumbents to be willing
to invest and explore andchallenge the status quo.
Davis Bell (29:04):
Yeah.
Well, I agree with you andmaybe, maybe one last topic for
conversation.
We've been geeking out a littlebit on the more, the software
side of things.
Bringing it back to accounting,I think, you know, my, um, my
mental model for AI inaccounting has been, um, the
exosuit, and I first called thisa cyborg, but we've got some
(29:26):
nerds here on staff whocorrected me, and, you know, if
you think about the movieAvatar, the humans walk around
in those, they're inside thisrobot, and it can pick up a 8,
000 pound boulder or whatever,move really fast, and, you know,
You know, that's, that's sort ofhow I conceptualize AI.
Um, because I mean, so much, Idon't want to venture a guess,
(29:47):
but an enormous amount of the,the actual human labor in a
firm.
Is menial, low value add,manual, just sort of, you know,
pushing paper from one place toanother.
Parses.
Yeah, and I think people don'tlike to do that, and it's a bad
use of their time.
(30:07):
Or they like it
Enrico Palmerino (30:08):
for all the reasons we wish they didn't,
which is, it's easy, I don'thave to think.
Yeah, sure.
Monday through Friday and justkind of like wake up and get
paid.
Davis Bell (30:19):
Yeah.
So I think that like most peopleget into the profession cause
they're passionate aboutactually accounting and they
want to think strategically andthey want to help people
strategically and then they justget mired down in the weeds of
this stuff.
So what, what would you agree?
Disagree?
Would you modify that?
Like how, what's, what's yoursort of model for it?
Enrico Palmerino (30:41):
No, I agree.
I mean, for the longest time we,we used to, um, Uh, compare Bot
Keeper to, uh, Iron Man.
Tony Stark is nothing, really.
I mean, he's a brilliant person.
But like, he becomes a superherowhen he has his suit.
And we would see the A.
I.
s being a suit.
And on its own, it can't domuch.
(31:03):
And like, even like the Iron Mansuit, it requires Tony to
command it, or direct it, or dostuff.
And that is the differencebetween, arguably, like, A.
I.
actually being an intern, Anintern, conceptually, like,
while it might, uh, he or shemight take, like, you know,
directions from you to do stuff.
The reality is, that person canalso Go do things for you
(31:25):
without you even asking becausethey recognize they're like self
starting Take initiative.
They recognize a problem.
They attempt to solve and youwe've all had those interns who
are like Hey, by the way, I didthis because I thought you might
like it like, oh, that's coolLike we don't need it But like
the self starting nature of andthe fact that did is like really
cool Ai can't do that right now.
Ai is only going to do thethings you specifically tell it
(31:47):
to Every time you tell it to dosomething so until it can start
to like problem solve for youwithout It is going to be an
exoskeleton or, you know, theIron Man suit that's just
amplifying, uh, the intelligenceand the use of our people.
And arguably, it's like, kind oflike the Iron Man suit.
If you took You took, uh, your,your weakest, like, you know,
(32:13):
engineer, like, you know, yourweakest Tony Stark from a
intelligence standpoint, you putthem in the suit, he'll be able
to do good stuff, like pickthings up and fly.
You take a brilliant person, youput them in the suit and they
can accomplish like insanethings.
And I think we're going to seethe same thing happen with AI.
The smartest people out there,the best accountants are going
to take accounting up notchesand levels that like we've never
(32:36):
thought possible.
And those who are not.
The, you know, strongestaccountants, they're going to
now be able to like be like thestrongest accountants or be a
really great senior accountant,thanks to AI, but they're not
going to, they're still notgoing to play.
It's not like, it's not like theall, uh, intensive, like plain
(32:58):
feel leveler, uh, that somepeople think it is, there is
varying degrees of, of skillsetand use that, that amplify its
outputs.
Davis Bell (33:07):
Yeah, no, I totally agree.
Hey Enrico, this has been superfun.
We'd love to do it again.
Um, and just encourage everyoneif they want to know more about
Botkeeper, botkeeper.
com super cool company, greatproducts, great leadership, and
I'm excited to see what, whatyou guys get up to next.
Enrico Palmerino (33:26):
Thank you so much, sir.
I appreciate it.
Davis Bell (33:28):
Thanks Enrico.
Popular Podcasts
United States of Kennedy
United States of Kennedy is a podcast about our cultural fascination with the Kennedy dynasty. Every week, hosts Lyra Smith and George Civeris go into one aspect of the Kennedy story.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com