in

Bots for Bookkeeping | Enrico Palmerino from Botkeeper

Accountants work analyzing financial reports on a laptop.

New technologies, such as cloud computing, have made entrepreneurship more possible than ever. But as more and more people are becoming CEOs and developers, fewer and fewer are becoming accountants—services that are sorely needed to make operations function.

Enrico Palmerino encountered this problem with his own startup. “Our accounting department just couldn’t keep up with the growth or the complexities of the business,” Enrico says.

He eventually discovered a large part of the problem was the lack of technology solutions in this field. This led him to found Botkeeper, an automated bookkeeping service for accounting firms powered by AI and machine learning.

On this edition of UpTech Report, Enrico discusses how he’s using this advanced technology to give the accountants more time to focus on their clients.

More information: https://www.botkeeper.com/


Enrico Palmerino is the founder and CEO of Botkeeper, which provides accountants with an automated bookkeeping platform using human-assisted machine learning and AI. Botkeeper’s platform delivers the fastest, most accurate, and low-cost bookkeeping available to accounting firms seeking a scalable CAS solution for their clients.

Botkeeper has raised $50M from marque investors and serves several thousand companies and 80+ accounting firms, processing 30M+ transactions per quarter.

Enrico also serves on the Board of OnProcess Technology and Fidelity Bank and is an investor and advisor in several prominent tech companies. He has been recognized as one of the Top 100 Most Influential People in Accounting, Top 30 Most Inspiring Business Leaders, and 2020 Golden Bridge Entrepreneur of the Year.

Prior to Botkeeper, Enrico was co-owner and Managing Director of SmartBooks, a cloud accounting firm, which grew from 6-40 employees in 3 years before he was bought out in 2015 so he could launch Botkeeper.

Early in Enrico’s career he was ranked 2nd among Bloomberg’s Top 25 Entrepreneurs Under 25 in recognition of his co-founding ThinkLite. Enrico grew ThinkLite from dorm room start-up to $8.5M ARR before graduating college and successfully exiting the venture.

DISCLAIMER: Below is an AI generated transcript. There could be a few typos but it should be at least 90% accurate. Watch video or listen to the podcast for the full experience!

Enrico Palmerino 0:00
The way that we approached it was, well, if I had hard evidence and prove that it works, and if it doesn’t mess up the client and you get you get crazy results, then accounting firms will adopt it like they, they don’t argue with numbers.

Alexander Ferguson 0:18
Welcome to UpTech Report. This is our apply tech series UpTech Report is sponsored by TeraLeap. Learn how to leverage the power of video at Teraleap.io. I’m excited to be joined by my guest today Enrico Palmerino who’s based in Boston. He’s the CEO and co founder at Botkeeper. Welcome Enrico. Good to have you on. Great to be here, Alexander, thanks for having me. No botkeeper is focused on providing an automated bookkeeping support to accounting firms, helping their businesses, the ones that they serve, using a powerful combination of machine learning AI and skilled accountants themselves. How we understand Enrico, when you guys started out, what was the problem that you initially saw, you’re like, Alright, let’s let’s solve this, let’s find a way we can solve this.

Enrico Palmerino 1:01
So I experienced an accounting problem, I think, from both sides of the table, I founded a company in college that grew pretty quickly. And our accounting department just couldn’t keep up with the growth or the complexities, the evolving complexities of the business. So then my next venture was a cloud accounting company to solve that problem for SMBs. And tech companies like like my prior one. And while there, I realized that a couple things were happening, one, there was a supply shortage of accountants, like fewer accounts going into accountancy, the there was a lot of retirement happening in the academy, the average age of an account is 55. Just kind of just crazy if tells you anything about the population that’s doing it. And then but small business formation was growing exponentially. So there’s this supply demand gap between demand for Financial Services and the ability to do it. And then you take that and you couple it with basically an abstract problem. So accounting firms, and like the cloud accounting firm, we had software companies like Wolters Kluwer and Thomson Reuters built platforms to assist and automate and manage tax practices and audit practices. But everyone’s left the cast practice, the client accounting services practice or bookkeeping practice, just off to the side and ignored. And they’ve been forced to fend for themselves with a bunch of disparate siloed apps that they have to integrate together and stack on top of each other. So I want to build a unified platform that the functionality of all those apps in it, and visa be those app that those feature set, extract and automate the the data extraction from the client, the SMB, bringing that into the platform, and then use machine learning AI to do a bunch of the compliance base like basic bookkeeping processing, freeing up the limited supply of accountants to do the complex advisory consulting, etc, for the client.

Alexander Ferguson 2:53
Okay, so if I can understand correctly, you starting in the accounting field, you saw then the issue there was not enough new accounts coming into the field, but the need was there. So like are how do we solve this, building a technology that I guess empowers or enables to take away a lot of the the basic work that person doesn’t have to do, allowing a smaller team be able to do much more by using technology.

Enrico Palmerino 3:21
And accountants are very skilled by very smart individuals. And unfortunately, you can’t apply the skill of the critical thinking until after the books are in order. So you really, it’s like a scramble to mimic a machine and process high volumes of data as fast as you can to only then spend very little amount of time like analyzing it providing advice during the complicated accounting side of the equation only to start right over again it cramming data in, you know, because it’s just a cycle every single month, this happened. So I just wanted to free up the skilled people to be more skilled in critical thinking,

Alexander Ferguson 3:55
CPA firms, do they tend to enjoy technology and say, yes, let’s bring on that tech.

Enrico Palmerino 4:02
No, they have historically been laggard tech adopters, who are as a very like pessimistic or cautious when it comes to technology adoption, because ultimately, they they’re trying to help and support their clients, and they don’t want to do anything that could negatively impact them. So they’re very, very careful when it comes to choosing technology.

Alexander Ferguson 4:23
But my own actually CPA in his firm, he said he’s had trouble trying to bring around his other partners on the fact of we should really adopt these new technologies like no, why would we change? So I’m what I’m just kind of curious that the pushback that you get of Wow, we could actually have technology manage all these menial tasks. We don’t need to has it been an uphill battle to bring this tech?

Enrico Palmerino 4:45
So the way that we started barkeeper was actually like this problem you described. I haven’t been in the accounting space knew I was going to be facing it head on with building this automation platform for accounting firms. So the way that we approached was, well, if I had hard, like evidence and prove that it works, and if it doesn’t mess up a client and you get you get crazy results, then accounting firms have adopted like they, they don’t argue with numbers and data. That’s what they do. That’s Yeah. So what we decided to do is Alright, first, let’s build the platform for ourselves. So let’s, let’s hire accountants, on our end, let’s build a platform for ourselves. Let’s support the SMB directly. And we’ll learn and we’ll be able to like understand and iterate and acquire a lot of data visa V, acquiring all these SMB clients onto the platform. And then we’ll be able to start testing the platform will be able to show enough evidence that this, this is going to work to accounting firms to have them test it and coach us on what it needs and, and what we need to evolve it to what feature set and then once we get to that point, then we can actually start selling the platform to accounting firms, and ultimately achieve the vision that we have for botkeeper,

Alexander Ferguson 5:56
Infosys is starting as a service base. Basically, you have to use the technology for yourselves. But you’re just hire accountants and doing this to understand the product building it and now you can launch it to two CPA firms.

Enrico Palmerino 6:08
Correct? Because I figured it would be a lot easier if I could say, look, we have hundreds of clients on the platform, we have each one of our callings can support this many more clients than you’re supporting right now, if you had the same platform, you’d have the same advantage and be able to do what we’re doing.

Alexander Ferguson 6:26
Any stats you can share with kind of where you guys are today.

Enrico Palmerino 6:29
Yeah, so over the over that course of call it four years, we reduced the human labor component and basically processing a client’s books by about 63%. And that’s continuing to grow. So like is it like year over year, we tend to lower the number of hours that a human needs to call it supervisor review the machine are assessed by about 30 to 35% year over year. And it’s you know, 35% of a smaller number each time it’s the you know, going from in school going from a 95 to 97 is incrementally more difficult than, you know, going from a 90 to 95. But yeah, we’ve we’ve been able to achieve, I think a very high degree of automation. And one of the cool things you see it just in our pricing. So when we entered the market, we we were selling a this bookkeeping package SMBs it’s a 399 a month, $399 a month, which was at the time, awesome, it was like a third or fourth of the cost of market rate. And yet we still had market margins. In today, we sell that same package to accounting firms for $39 a month. And we have twice market margin. So it’s a it’s this, you couldn’t explain it any other way, then, you know, tech is doing what it’s dreamed to do.

Alexander Ferguson 7:49
Let’s just dive real quick, into the tech stack itself help me understand more about how it works for those that are more curious about the tech itself. Sure.

Enrico Palmerino 7:58
So if you looked at accounting, if you were cloud accounting firm, you would go and you license a bunch of different applications and app to say do dashboards an app for receipt capture an app for bank statement fetching an app for dog stores an app for task management, an app for so on, and so communication, so on and so forth. We’ve taken an each of those apps have a license fee, and those fees stack up and they get pretty expensive. And they’re incremental, every incremental client you put that fee is not going away. We basically built all that feature and functionality functionality set into one unified platform where those pieces are integrated. So a task that you let’s just say attach a document to that document is the exact document that sits in the document management component. And it’s not an old or outdated version of it. Like you can see the versions have been changed over time. So you don’t run into like those issues. But better yet, the task is triggered automatically when something happens. So you don’t have to go in and create a new task every time or set up a template of tasks. And oh, by the way, when you complete the task, in some of the cases, the task auto zeroes itself out. And you can complete the task sometimes right from within the task. So it’s not a go to a task. Here, go switch platforms complete, come back and check it off. But that’s just one piece. other pieces are we’re extracting data off of documents using OCR in an automated fashion. We’re pulling in transactional data straight from banks and using algorithms to pre calculate information like here’s how much cash you have on hand across all your bank accounts. Here’s how much credit you have out there. Here’s what your net caches and things that to do it manually every single day would just be very time consuming. And then another aspect of it is the machine learning and AI so as you can hear where we’re while we’re AI company, automation comes from a variety of software developed, integrations RPA, workflow etc. But on The machine learning side we can take all the transactions from the banks and the credit cards, etc, into our system. And, you know, apply the appropriate logic to automatically categorize and classify the majority of the transactions without a human’s involvement.

Alexander Ferguson 10:13
It sounds like the majority of what you’ve done is taken technology that exists and actually, in some places is just very common. But you’ve applied it in a unified way. So one, it just all these different things, and they’d have to go to different solutions for bringing in one solution. But you’ve also applied it to one industry, right? Just Just accounting, because then then it serves that specific purpose. But then on the machine learning side that that last piece is you also knowing the industry? Well, you you’re able to categorize better and better and better, obviously, as you have more and more transactions come through, Correct,

Enrico Palmerino 10:45
yeah, so our data set keeps growing. And one of the nice things about supporting accounting firms is they have many clients each. So as they onboard all their clients, we get high volumes of clients. And each of those clients comes with all of their historic data. So it’s not just the data from here for that we’re learning off of, you plug in, if you’ve been in the business for five years, we get five years of history pulled into the system. So each incremental client we’re learning from every month that goes by, we’re learning from that incremental set of data. And the one of the beauties of our model is we have this AI coaching network. So while the machines are doing a great job at processing a majority of the transactions, there’s going to be exceptions and questions and things that it doesn’t have a high enough confidence rate to do, we kick that out to the expert accountant at the accounting firm. So the the accounting firms that use us our training, their accountants are training our machine and making it smarter over time. And if you could relate it to like Google’s CAPTCHA, which you know, to do the past, we’re all training their algorithm to do the photo, you know, photo based AI, for accountants that use our platform or training our system to be better at automating the accounting for them, which makes them happier, because they do less. So it’s this beautiful, you know, color

Alexander Ferguson 12:03
approach. Yeah, win win win. Can you hear anything about your current percentage and, and even pontificate on the future of getting that percentage down to zero? where people don’t even need to review them anymore? Will we be able to get to that point.

Enrico Palmerino 12:19
So I don’t, I don’t think you can ever get down to a point of zero, just because the way that basically the way that machine learning works is we are many, many model approach. So we have models that look at the historic transactions of a client, if there’s a history and a pattern there. We know what to do with it. If there isn’t a history and a pattern. So it’s a new transaction, we look across all of our other clients to say, Can we draw a correlation at a high enough confidence to say we know what to do with us, but they’ll always be new vendors entering the market, new items being sold new transactions, so you’ll inevitably always have transactions that come into our system that have a low confidence. So always need to be some level of human review. But hopefully, over time, it is, uh, continues to trend down and I just say, you know, this year, we expect another, you know, 30 plus percent reduction in human time. I don’t I see that trend happening again next year. It’s just gonna always be like, maybe it’s 20 to 30% reduction of a smaller and smaller number.

Alexander Ferguson 13:18
That makes sense. Yeah. What I’m fascinated with here is like, there’s solutions out there already for SMBs. And consumers that do this, your bank transactions come in, and it can even help start to categorize for you, for the end user. But it’s still up the CPAs. Like, Oh, wait, hold on a sec. How am I helping my customers if they have solutions, but we don’t. And that’s what it sounds like, that’s what you’re doing is you’re empowering the CPAs, that you have the same solutions that are out there. But now you could actually providing a great service with less time for you. And my getting the scenario, that environment quick,

Enrico Palmerino 13:51
kind of what I would say is, the analogy that I think everyone can relate to is let’s just say like you have an American Express card. And we’ve all seen this, like, it’s great that it has some sort of like automated grouping of like, these are tech expenses, these are meals and entertainment. But how often do you go and then you’re like, I bought skis. And it put it as a tech company. Like that’s totally wrong. It Yeah. And it’s all it’s trying to do in your scenario. And like the the individual scenario is determined whether it’s one of five things, but the average chart of accounts for a business could be 100 variations, and every business has its own unique Chart of Accounts. So you’re, you’re dealing with a level of complexity that the consumer just doesn’t have to deal with, which is why you know, we’ve had to raise a lot of money and hire a lot of engineers to build to build it to this point.

Alexander Ferguson 14:42
When when you had decided to build this was this pretty straightforward. Can you share anything about like the beginning pieces and where you guys are today?

Enrico Palmerino 14:53
So definitely not straightforward and it’s a bit I would say like yo I thought we would be where we are I thought we would have achieved where we got to today, like two years ago, because you just never really are building like, this is a while ago, I didn’t realize this, or the data sets not clean, or the integrations or API’s break sometimes and having to, you know, figure out how the proper ETL that, you know, structures, the data accordingly. So, I think the approach, I think most entrepreneurs, if you’re willing to iterate constantly, and evolve and tweak things, I think that’s how you succeed, I find very rarely does someone set out and whatever they thought in the beginning was exactly what it is, you know, kind of going forward. So we’ve been highly iterative, constantly evolving it. And I think that the beauty of our businesses, I don’t pretend to be the smartest person in the room. But I’ve really good at finding the smartest people out there and bringing them on our team and have them you know, kind of take it to the next level.

Alexander Ferguson 15:53
How big is the team today? So I think we’re about 300 employees. So pretty good sized team. Pretty, pretty good size there. But those who want to hear more about the journey, stick around for part two, where we’ll be hearing America’s founders journey. But coming back to this kind of concept of lard at large here of CPAs, trusting in technology. We all want to rely on technology and say, Oh, well, I’ve nothing to worry about now. I mean, is that the case? How are you? How are you approaching that when when CPAs are starting to adopt technology and use it?

Enrico Palmerino 16:25
So our approach, kind of our mantra has always been that humans and AI are better together. And I don’t believe that AI as as it’s defined as artificial intelligence really only exists and universities and in gaming, where the consequence of an error is low, like chess, like what’s the, what’s the worst that could happen if the chess AI loses. But when you’re dealing with a business’s financials, and like financial data, I mean, they’re making decisions off of that data, like paying a bill too early and messing up cash flow would be, you know, highly consequential. So that the idea that the machine is doing things that it is highly competent on by itself, that’s great. So it can be autonomous. But more often than not, the machines doing aspects of the workflow are making recommendations to a skilled human, and basically augmenting the individual and so for, for CPAs. It’s great, because we these are smart individuals and accountants, you don’t have to be a CPA to be an accountant. These are smart individuals. And if we can just assist them to leverage their you know, skill set, or strikes better and more efficiently. That’s the goal.

Alexander Ferguson 17:36
It’s it’s a powerful understanding here of in non concerning world, personal world play game, AI is fine. But if you had to make decisions, you’re not trusting AI completely do to make decisions for you. But it’s helping you get to the decision faster, potentially, it’s making decisions faster.

Enrico Palmerino 17:55
Yeah, prime examples. One of the other things that we automate is at the end of the month, like doing a review of all the book, so ironically, machine does a chunk of it, then the machine recommends to human, something, the human says Yay, or nay, or kicks out for one of the accounts that the accounting firm to do. And then at the end, another machine basically audits, everything that was done, because there’s human error introduced there, at some point in time to pull out like, say you had 1000 transactions, effectively, it will pull out like 3% of them that thinks there’s a high probability of error on now for a human tab to have like, looked at every single transaction, they’re not going to do it. First off, and they’re going to spot check, maybe, but if instead of having to do that, you were handed, here’s 10, transactions, like review, these take time to review them. And there might not be anything wrong with these time. But these 10 have the highest probability of having a mistake.

Alexander Ferguson 18:51
Wow. So it’s actually multi layered integration of checks and balances, both human and AI in multiple places.

Enrico Palmerino 18:59
Yeah. And at the end of at the end of that, you know, hey, we kicked out these 10, a human has said, yea, or nay to those time, which trains the machine and makes the machine smarter, and so on, and so forth. So it is this beauty, like this beautiful harmony between machines and humans?

Alexander Ferguson 19:15
What can you share as far as current features, upcoming features that you’re excited about? And you would want to bring forth here.

Enrico Palmerino 19:24
So one of the things I’m really excited about is our new our next generation platform. So we’re in this next platform, we’ve done what what I was saying we do a little bit of it now, but smart tasks like this idea that there’s a piece of the equation of the workflow that human needs to do something on and it’s in a task form to do it. And you can execute the task within the task and that the machine kind of zeros out the task after I think is going to be I haven’t I haven’t seen anything like that before. And the fact that we built it very specifically for accountants and the kind of things that they want to handle, I think is going to be really So

Alexander Ferguson 20:03
how will that like affect the How will that play a role than in like a day to day normal of a CPA firm or accounting firm.

Enrico Palmerino 20:11
So, in, in doing the books for like a typical client, you’re you’re having to do systems switching like a lot you’re, you’ve got, you got to go into like QuickBooks zero to see your clients and those are two different gels, then you got to go into like, say, Bill comm see which of your clients are using bill comm, you got to, you know, go here to upload a report, pull data from this other system, and And all the while, you’re constantly going to a task list of things you need to do and then having to go in and out. And it just increases the probability of human error, because all those jumping around, but if you could just work from one place, and you had all the feature set there that you needed to execute, but you can even go so far as execute the item that you need to do in this one location. That would just make I think life a lot simpler. And I think that’s the goal is just simplify the complex.

Alexander Ferguson 21:03
What a powerful PC technology should do is simplify the complex to make our lives easier. So we get back to enjoying what we really like to do. I appreciate the time you’ve taken us through the value set of what you guys are providing, and where where you’re headed, can you can you speak actually, anything to your roadmap of where you guys are going?

Enrico Palmerino 21:24
Yeah, so we kind of look at accounting as being in like four groups. So there’s pre accounting, and that’s your call it pulling in all the data categorization, classification, reconciliation, you know, financial report generation, then you have operational accounting, which is bill pay, and accounts receivable invoicing. And then you start to get into skilled accounting. So, deferrals accruals, you know, rev rack, and then you get into I’d say, like CFO advisory KPIs, which is your dashboards, forecasts, etc. We’re constantly trying to like work up that stack. But features that I see is basically coming out with is soon we’ll have a more autonomous version of AP. So it would just make it easier for you to run and manage bill pay all once again, within this platform, hopefully, you know, get and gain some, some nice insights from where your expenses are in that system. And then we’d like to also make it easier, like build tools for the skilled accountants, so not full stack automation, but a tool that makes running and doing deferred rack and prepaid schedules and asset depreciation schedules a lot easier and manageable in once again, one place, so you know, jumping around from Excel sheets to manually entering into the GL,

Alexander Ferguson 22:41
I get the sense that you have a clear roadmap of of the bigger picture of where you’re headed, and the steps that that you’re headed towards. Well, I’m excited. Thank you again, for for sharing for sharing this. This is definitely This is your first venture either, right that this you’re familiar with the concept of building a business?

Enrico Palmerino 22:57
Yeah, I’ve, I’ve failed a lot more than I have succeeded. So I think I’ve incorporated maybe it’s like 13 companies and I’ve had successes with about three. So yeah, I’ve made a lot of made plenty of mistakes to learn from and still still making mistakes today.

Alexander Ferguson 23:12
I would trust in someone who’s made all the mistakes and knows what to learn from that. They can go from there. To they help out a lot. Advisors are powerful. So those that want to learn more, though about the journey that Enrico has been on stick around for part two of our discussion on our next episode of UpTech Report. Thanks again for your time, Enrico. And for those who want to learn more also go to botkeeper.com. And it looks like you can just get started now you can get a little demo. Is that a good first step? Yep. jump right in and jump right in. Thanks again, everyone, and we’ll see you on the next episode of UpTech Report. That concludes the audio version of this episode. To see the original and more visit our UpTech Report YouTube channel. If you know a tech company, we should interview you can nominate them at Uptechreport.com. Or if you just prefer to listen, make sure you’re subscribed to this series on Apple podcasts, Spotify or your favorite podcasting app.

Creating the Perfect Product Demo | Yoav Vilner from Walnut.io

You Only Need to Ask | Rohit Vashisht from WhizAI