In this episode of the Accounting Technology Lab, Randy Johnston and Brian Tankersley interview Jeff Siebert, Founder and CEO of Digits, about the evolution of modern accounting, real-time financial intelligence, and the future of advisory services. The Accounting Tech Lab is an ongoing series that explores the intersection of public accounting and technology.
View the video below:
Or listen to the podcast using the below player:
Transcript (Note: There may be typos due to automated transcription errors.)
SPEAKERS: Brian F. Tankersley, CPA.CITP, CGMA, Jeff Siebert, Randy Johnston
Randy Johnston 00:00
Two, one. Welcome to the accounting Technology Lab. I’m Randy Johnston with co host Brian Tankersley, and we are so excited to have a special guest here today, Jeff Siebert of digits. We’ve known Jeff for years. His reporting tool that he’d created was super and earlier this year, he released a full blown AI General Ledger in the last couple of weeks, new auto reconciliation. I mean, this guy’s on fire. And I thought, are we lucky enough to get him as a guest? And it turns out, we are. So Jeff, good day. Welcome to the accounting Technology Lab.
Jeff Siebert 00:35
Randy Brian, thanks so much for having me. It’s great to be here.
Randy Johnston 00:38
Yeah, no, I think a lot of our guests and listeners may not know background on you, so it’s probably appropriate, if you don’t mind doing a little bit of an intro.
Jeff Siebert 00:49
Yeah, quick background. So I personally had no background in finance or accounting. My whole life has been tech. I taught myself C when I was 12. Grew up in Baltimore, Maryland, and just started building software. And so long story short digits is actually my third tech company. I built a business in document collaboration way back in 2000 708, we got acquired by box. I built a mobile developer tool platform. In 2011 we got acquired by Twitter that now runs on 6 billion smartphones around the world, basically every active phone on Earth. And then at Twitter, I became head of product in 2015 and so left Twitter started digits in 2018 and it really came actually from that journey of doing building businesses and then seeing how Twitter ran internally. And what struck me was the difference in data quality between what we had on the product engineering side and what we had on the finance side. And in product engineering, you have Google Analytics, you have AB testing tools, you have Grafana. You can see what all your servers are doing live in the moment. You can see what your users are doing. It’s just incredible real time insight. And on the finance side, as the business owner, I was waiting two to three weeks to get a black and white PNL from our accountant at the time, and I was like, something’s wrong here, right? Like, why don’t I have a real time finance dashboard? And so that was the premise for digits when we started the business, of course, then I fell deep down the accounting rabbit hole. I read two textbooks, cover to cover. We hired UCLA, professor of accounting to give our team a private class. Talked with hundreds of accountants, and what we really learned is it wasn’t the people. The people are, of course, hard working, disciplined, diligent, really want to help these business owners, and they were being held back by software. And so that became the premise for digits, is, can we build sort of the next generation of accounting software?
Randy Johnston 02:33
A beautiful bit of background. Now, our listeners know that I wrote AI on lisp 50 years ago this year,
Jeff Siebert 02:41
I love that
Brian F. Tankersley, CPA.CITP, CGMA 02:43
I’ve been around Randy. Randy actually had dinosaurs roaming the earth. He had to worry about getting getting caught by Raptors when he exited his office every day. So it was an interesting time.
Randy Johnston 02:55
But it turns out, I’ve been doing AI for 50 years, waiting for this evolution to occur. And, you know, Brian has a phrase that he’s used is he wants accountants to get real time data with very little real time work. Yes. And you know, we knew that was coming. In fact, when you announced the digits GL and showed it at different events, scaling new heights, being an example, it was clear what, what was going to happen here. And you know, so from that vantage point, I know you’ve got some tenets that are very near and dear to your heart at digits. And I think it’s good for our listeners to hear some of those things in terms of your fundamental core values and direction,
Jeff Siebert 03:41
yeah, so I can, I can go through a couple of them, and then we can dive in deeper. Of course, the first is that AI is a tool. It’s not a solution. It’s a technology, just like a database is, and you need to define and design the entire system to actually make use of it. And so the big difference between digits and sort of any ledger that’s come before, it is actually the data architecture. And so if you look at QuickBooks, zero, NetSuite, SAP, etc, etc, etc, they are all relational databases architected 20 plus years ago, where each transaction is a row in the database. And so when you spend money on Uber, for example, they just see u, B, E, R, they don’t know what Uber is. And so this was actually our core realization back in 2018 when we started digits, is to design it for an age of machine learning. Now what everyone calls AI, you had to change the data model. And so digits is a vector, graph, data architecture. Everything’s an object. And so Uber is an object, your bills, your invoices, your customers, your vendors, your transactions, they’re all objects in our system that allows our models to basically build a semantic understanding of how the money flows in each business. So that’s the first one is like how we understand data. The next I’d call out is basically extreme data privacy. So I’ve built my whole career around data privacy. Helped create the Netflix documentary The social dilemma. We do not sell the data. We use it only to train the internal models. We actually are able to train firm specific models. For accounting firms, they train only on your client’s data, and that really helps encode your firm’s best practices. So we’ve gone to extreme lengths in our data architecture.
Randy Johnston 05:21
Well, it’s super three points, and I was actually zoned off there for a minute on old database architectures, because our listeners have never heard us say this in our 250 plus episodes that we’ve recorded, because I had the good fortune of being around the original designs of databases like IMS, like DB two, like amazing SQL, like Sybase, and one of the original database designers in the market and I co authored a book together, and his way of describing this rows of data that you’re talking about, it’d Be Like going up to a parking garage and taking your car apart and putting the fenders up on the fifth floor, the engine on the third floor, the body on the first floor, and then you had to reassemble the data. Now, this vector database that you’re talking about, there’s some attributes. And again, our listeners are pretty technically astute, but describe some of the real benefits for the listeners today on this vector database model, and why this ability to identify an Uber invoice or a utility invoice accurately is such a big deal or the
Brian F. Tankersley, CPA.CITP, CGMA 06:36
or the similarity between things,
Jeff Siebert 06:38
so that Brian is exactly the point with with a vector graph data model, you can build a similarity understanding. And so let’s say you have a business. They have an Uber transaction that’s booked to ground transportation, right when they spend money on Lyft, what happens? Right? A traditional ledger would have no idea. But because of the object graph we can understand across our global data set that Lyft is semantically similar to Uber. And so if you have booked Uber, you know how to book Lyft, right? It’s also ground transportation. And so that’s probably the, the most concrete example I can give, but it allows you it’s it’s hard to picture, because this isn’t just like a simple, two dimensional graph, this is an n dimensional graph. This is what the deep learning models work on, and so they’re able to detect really fine grain patterns across multiple dimensions that allow them to really understand how to book the transactions.
Randy Johnston 07:34
So with that explanation, then, Jeff, I think I would extend this to say you would understand curb or Ola or grab or Didi just as easily.
Jeff Siebert 07:46
Perhaps I am not as personally familiar with those, but I think conceptually, yes,
Randy Johnston 07:52
I thought that would be the case. Now, the other beautiful thing that you said earlier, which is, in my mind, super important, you can, in effect, I won’t use the word customize, but you can recognize firms patterns, and that tells me that you’ve got a multi layer data model where we can have this global view of all data, all firms, but we don’t share it or cross pollinated in any way, shape or fashion, but then We have this isolated area that’s about you the firm. And I suspect you might even have the ability to be you the user. But that,
Brian F. Tankersley, CPA.CITP, CGMA 08:30
well, well, if you, well, if you think about this, I mean, realistically, you know, if you’ve got a telemarketing operation, telephone expense is a different thing than it is if it’s an insurance company where you’re just using it, it’s an ancillary thing to Office. And so I appreciate that you’re that this gives you a level of industry potential, industry knowledge that you know to in context, say that’s here, just like, just like, you know if I’m, if I’m doing, if I’m doing home, if I’m doing repairs on an office and it’s just my office, then that’s repaired maintenance expense. Okay, on the other hand, if I’m if I’m a construction contractor, that’s cost of sales and job expenses. And so it’s a even though it’s the same vendor, Home Depot, or Lowe’s, or 84 lumber, or whoever it is, it’s a whole different place it goes for the reason. And so this context engine you’re talking about with the vector database, this ability to figure out context is a very, very big deal.
Jeff Siebert 09:33
So this is exactly right. The way we describe it is, you can think of digits as a layer cake. This is how we get these different levels of context. And so our top tier models train individually on each business or each client for a firm. And so if that client has seen the transaction before, digits is effectively perfect at booking it, because it knows exactly what was done for that business, and it will mimic that work. Then, then the next tier of models, it falls back to the firm model that. Because a lot of firms do specialize in practice areas or different geographies, industries, etc, right? And it’ll look at how they have treated that transaction across their other clients, and then mimic that work. This is a superpower for firms onboarding new clients, because you can start a new client fresh on digits, and it’ll mimic the work you’ve done on your other clients, which is really so Jeff.
Brian F. Tankersley, CPA.CITP, CGMA 10:20
So Jeff, is it contextually aware enough that it can extract, let’s say, if you’ve got 20 power bills for 20 different profit centers where it can extract that, that account number and then assign it to the proper to the proper expense Expense Center.
Jeff Siebert 10:36
So this is, this is the holy grail we’re building towards. As you know, it’s all about context. If the model can get access to the context, it’ll detect the patterns. And so we are basically beginning to pull in more and more detail into the model. And you mentioned our automated bank reconciliation launch a few weeks ago. So that’s one of those steps we now ingest the bank statements. And one of the big challenges for years has been checks, because right in the bank feed, you just see check, 123, it’s useless, right? A lot of banks now in line the images, the check images, into the statement. We can now extract the image, look at what the check was for, and then book the check. And so that’s exactly the sort of journey we’re on. Is pulling in those cost center data, account numbers, whatever we can get from bills, invoices, etc, to keep making the model smarter. And then as it falls through the layer cake, it basically uses that context at each level to make the best guess it can. Yeah.
Randy Johnston 11:31
So I think a reasonable statement there might be not quite yet, but we have that on our radar
Jeff Siebert 11:37
right exactly so, so publicly, we’ve announced we’re at 97% accuracy, and that remaining 3% right, it gets, like, asymptotically more difficult to get perfect. But that’s what we’re racing towards, is just trying to pull in all the context we can to figure out that remaining 3%
Brian F. Tankersley, CPA.CITP, CGMA 11:54
it’s like, it’s like, in physics, you’re trying to approach the speed of light, and it gets harder and harder
Jeff Siebert 11:58
every time. Yeah, exactly. Interesting.
Randy Johnston 12:01
You into physics. I was going to mathematics for infinity. It’s the same concept. So now, Jeff, another moving part is it’s clear that, you know, a very popular platform into it has both QuickBooks Desktop and QuickBooks Online. And one of the tricks is the desktop product has more inventory and costing capabilities, right? Yeah. And so the most common question I was asked in the fourth quarter, and again, I’ve had the fortune of speaking to 1000s of attendees face to face in the fourth quarter, was going to have to have to get off QuickBooks, desktop, what platforms can I go to? And I need inventory costing, and I’d like integrated payroll? Yeah, probably heard that more times than you know you’re gonna, you’re gonna become an old guy with gray hair or no hair, like I’m turning into. But what’s, what’s your, uh, response to that particular need?
Jeff Siebert 13:01
Yeah, it’s a really great question. It’s obviously a super important feature for a lot of folks today. We don’t have inventory built in. We do focus primarily on local professional services, online businesses, digital businesses, basically non inventory tracking businesses. This has been a huge feature request, and something strategically we aim to partner on so there’s some great companies in the space nowadays. For example, Katana, I would highly recommend folks check them out. And it’s similar to payroll. I know Intuit has sort of gone through all of the adjacencies and tried to build them all. We would much rather partner with the strongest players in the space. So for example, we have a great partnership with gusto on the payroll side, and so we will likely partner with the top inventory products and do tight integrations, versus trying to build that and try to be best in class at multiple things understood.
Randy Johnston 13:52
And I wanted to make sure that our listeners knew that demarcation, because that is one of the things that I’ve said about your platform, really, since inception, and and I hope that you are so successful that you can acquire inventory costing and acquire payroll, and have that same development attitude around it. So now we’ll talk more about product a law, I think, a little later in our time together. But I’d like to turn attention to your second major tenant, which was about privacy, and by the way, that was the setup with this inventory costing payroll partnering thing. Because it’s one thing to be on your own platform and control the data. It’s another thing all again, to use outside AI engines or outside sub processors, which, of course, with the Intuit announcements of the 100 million dollar partnership with open AI and data flowing back and forth, you know, there’s questions about, where does the client data go? Now, I’m going to overstate this slightly, Jeff, but one of the tenants of a CPA license. Is the fiduciary protection of client data. And there seems to be a lot of CPA professionals that have forgotten that.
Brian F. Tankersley, CPA.CITP, CGMA 15:09
And I will tell you somebody that’s taught CPA review, I will say that this is, this is the holy grail of CPA. Okay, it’s that you will not, without authorization, disclose this information to anybody for any reason. Okay, this categorical thing here, okay, and this is the thing that freaks us all out, because, you know, it’s, it’s one of those things that it’s what the privacy is, one of those things where, if we don’t if we don’t watch it very carefully. You know, the problem is, it’s all embedded in 50,000 80,000 words of privacy policies and EULAs and everything else. And only one document has to say we do this, and they can bury it wherever they want. And so I just want to ask you categorically, are you selling information to anybody for any reason whatsoever.
Jeff Siebert 16:04
No, we do not. We do not sell the data. Okay, so
Brian F. Tankersley, CPA.CITP, CGMA 16:07
it’s not going to be used to, it’s not going to be used to market loans to anybody. It’s not going to be used to sell insurance to anybody. It’s just used to, it’s just used to process it internally. And you have separate models available for firms. So if they don’t want to use somebody else’s data training, they can’t, they can, right?
Jeff Siebert 16:28
That’s exactly right. And so yeah, we in our privacy policy in terms we of course, have the right to use the data internally to improve the product. That is, by definition, what we need to build the models and make the product better. But yeah, we don’t sell the data to anyone where we do use sub processors. Obviously, we host on Google Cloud, as does most of the internet these days. All the data is encrypted at rest. Google does not have access to it. We do have some features that require foundation model API’s. Our contracts with those vendors prevent them from using the data, training on the data, et cetera. So we’ve taken this very, very seriously. And this dates back to my earlier career in mobile development. I mean, my prior company, we had access to live DAU of all the top apps in the App Store, like the crown jewels of these multi billion dollar companies. And so we were very serious on this from day one,
Randy Johnston 17:19
yeah, and that makes great sense, because topic for another day, but I bet you’re excited about the new Google TPU sevens. That for lots of reasons based on that comment, but you know, I would just say that over the past 18 months, but more so of the past seven months, I can make the vein stick out in the neck of a CPA professional when I suggest that the Intuit QBO not opt out problem, and Intuit Payroll not opt out problem, when people know about it, they’re pretty aggravated, And they’re usually looking for another path, and because of that, I’ve suggested that your digits platform is another landing path. So this strong piece of privacy is a big deal. Now this is a little bit of a right turn, but today, do you have a relatively good way to migrate alternate vendors data, or is it better just to, you know, hoover up the data from existing bank and credit card feeds.
Jeff Siebert 18:29
No so people get surprised when I say this, we will help you move off your existing Ledger in about five minutes and digits will get your entire historical data, every transaction, every line item, every piece of metadata, etc, and you can pick up exactly where you left off. So there’s no, like, multi week implementation process or anything like that. You can sign up and get started within minutes.
Randy Johnston 18:54
Yeah, and Jeff, I’d heard you say that publicly before, but I thought it was good for our listeners to know that in the in the podcast here. So you know what we’re hearing here, friends is kind of a one two punch of some, you know, very strong AI capabilities in this AI powered ledger paired with the right attitude around data privacy. So you know, the only thing I might have as a technical worry is scalability. How big do you think you can grow?
Jeff Siebert 19:29
This is a great question, and again, draws back to my prior company experience. So with Crashlytics. Crashlytics today now runs on 6 billion monthly active devices. Does two to 3 billion events a day, literally trillions of events in the data system. And so the scale of that is just unfathomable. In comparison to finance. We have businesses on digits with millions of transactions. And the core engineering team at digits was the core engineering. Team that built Crashlytics, and so we designed this from Day Zero to be a high scale ledger as a concrete example for that stripe. As you know, when you when you have a business running on stripe, you get all these purchases coming in, all this line item detail, no one today is really booking that into Xero or QuickBooks you’re doing roll up journal entries in order to keep the data volume low, we have a native stripe integration. It pulls in every single piece of data, every purchase, every customer, detail, etc. Because if you want to be an ERP, you need all your customers and purchases in the ERP. You can’t hide them behind a journal entry that sort of destroys the whole value.
Randy Johnston 20:37
Yeah, that makes great sense in terms of an illustration, because it’s well, not like Shopify, who had an unfortunate outage with the Black Friday 25
Jeff Siebert 20:46
weekend? Oh, no, I actually didn’t see that.
Randy Johnston 20:49
That’s crazy. Yeah, they were down. No Shopify transactions. It’s like, Oops. We took a
Brian F. Tankersley, CPA.CITP, CGMA 20:55
little too that’s a bad day.
Randy Johnston 20:57
Yeah, it’s a real bad day. So, so I think on this point I want to transition out of technical into product here in just a minute. But are there any other technical questions we should have asked you, other things that you get asked frequently, or other things that you think are core to know, because one of the disadvantages of our podcast is, I tend to be the hardcore technician and Brian tends to be the hardcore accountant. And if neither one of us were in the room, Brian would sound like a hardcore technician, and I’d sound like a hardcore
Jeff Siebert 21:33
No, I love this, and I know I love your depth of knowledge in the space I should have you on our more tech focused podcast.
Randy Johnston 21:41
It’s kind of a funny deal. You know, as I sometimes say, I slept in a Holiday Inn last night, so I think I can play an accountant.
Jeff Siebert 21:49
Well, no, this has been very thorough. We touched on scalability, touched on data privacy, touched on data architecture. I’d say maybe the only other thing is the big like differentiation with digits, and I believe, versus everyone else out there in the space, is we build and train our own models. And so yes, there are aspects where we’re using the foundation llms that everyone is using, but a lot of the core value at digits comes from our own model development, our engineering team, our ML team, quite literally wrote the O’Reilly books on machine learning. Over the past five years, we’ve spent over five years training these models. And so it’s not just AI bookkeeping, like how to classify the transactions. We also have our own document extraction models. Our bill pay models are incredibly accurate at pulling out all the information from your bills. And so this has been a
Brian F. Tankersley, CPA.CITP, CGMA 22:35
major I can testify to that, by the way, I can testify to that I put my VoIP I was telling you earlier. I put my VoIP telephone bill into it, and it pulled out the information on the vendor. Pulled out their tax ID number. It pulled out their their bank account information. It set up and said, Do you want to pay this via ACH, which is 50 cents? It went through and did all of that instantly, and it was, it was impressive
Randy Johnston 23:05
that point, you know, building and training your own models is, you know, impressive to start with, because Brian and I both have the privilege of teaching a lot of artificial intelligence, and we talked about the frontier models, and they the ability to protect the data and what you have to do and so forth. And you know, by my count, there’s 1000 plus models out there. But for example, in open router, today, I’m running 579 models, so I can run results side by side. And when I look at the new Gemini version, versus the Gemini pro three versus chat, GPT, five, one, it’s clear that Gemini has forced open AI into code red, because they may have gone around them in this release. But then you look at what’s happened with Mistral, who are clearly running on edge devices, a small enough model to run on a smartphone. You know, this whole world of AI is something you’ve lived in for a while, and obviously we’ve lived in it a while watching it, trying to help you understand where you can use AI effectively with your clients and in your firms, as we see it, yep.
Jeff Siebert 24:14
And what’s really interesting is the open source Chinese models aren’t that far behind, yeah. And so you can imagine a world in a couple years where it’s effectively commoditized, where you have extremely smart models that are open source and comparatively cheap to run. That’ll change a lot of things.
Randy Johnston 24:31
Yeah, it will. In fact, when this was first breaking we were recommending people run a lot of open source the llama models, when they were broadly available, if you wanted to remain private, because you could install it and run it privately and so forth. Well, Jeff, so far, great. Let’s turn attention, if we could, to the bay, the digits, product itself. And again, this is not trying to be promotional friends. What we’re trying
Brian F. Tankersley, CPA.CITP, CGMA 24:56
to do, by the way, let me say this, there is no financial relationship. Relationship between Randy and me and digits, okay, zero. Okay. We are doing this because we are interested in the technology. So let me just say that categorically at today, there is zero relationship.
Randy Johnston 25:13
And I would go a little bit further. The best I can tell Brian and I may be some of the last surviving independent consultants to the profession, because we are quite aware of back fees being taken by many that are not disclosed. And again, Brian’s hair is standing up on the back of his neck, because if his Becker, you know, CPA training, the independence rules live inside his veins.
Brian F. Tankersley, CPA.CITP, CGMA 25:45
I’m not going to say I’m not going to say categorically, that I’m independent from everybody, but I will tell you that my opinion and Randy’s opinion are not for sale to anybody at any price. And I say this because we’re saying a lot of good things about Jeff and digits. Okay, we do try to accentuate the positive wherever we can with everybody, but I will. But because we have been, we have been so effusive in our praise thus far, I want to make sure that everybody gets this okay. And you know, you can, you know, if you don’t believe me, come look at my bank records. There’s not a nickel that’s come from digits to me.
Randy Johnston 26:20
Okay? And so, you know, I have another rule that I’ve used for some time. If you don’t have something nice to say, don’t say anything at all, the grandma rule. And, you know, and I’ve put you a little bit on the spot, Jeff, about the, you know, lack of costing, slash inventory, slash payroll. So it’s not a criticism, because having designed and helped build inventory and costing systems at all scales of legacy GL products, I know how hard it is to design those elements. Well, I’ll just you probably already know this, but I’m just going to say doing a GL, A, R, A, P, is pretty trivial compared to doing a costing inventory, Bill of Materials roll up, just saying, so appreciate it. Yeah, so you got a way bigger hill if you go that’s that’s why the third party doesn’t bother me, because people that understand, that know how to conceptualize it. So that’s why I want to turn attention to because I don’t want to understate or overstate the capability of digits today. And I’m just looking at what I would call the value base. Because you know where you’re selling the starter at $35 a month right now, core at $100 a month, and then you have the Professional Edition, which is a negotiated price. Just help us understand the three levels of product, what you’re trying to accomplish and what we should expect that’s in the product, and stuff that is specifically outside the product. And that’s pretty clearly enumerated, in my mind, on your website at digits.com/pricing,
Jeff Siebert 28:07
that’s exactly right. And so yeah, what we’re aiming for with the three tiers is our starter plan, $35 it’s really geared towards simple businesses, solopreneurs, anyone who’s sort of getting up and running and likely, honestly has more of a tax focus like That’s why they’re sort of doing their books, because their books, because their businesses are pretty straightforward and so super easy to get started with. You literally sign up. You link your bank accounts, cards, etc. We pull in all the data in real time. We book it in real time. Anything that the models are low confidence on. This is another advantage of predictive models, is we flag for you in the inbox, because the worst thing we could do is make a mistake booking it to the ledger. So we just throw out anything that’s low confidence, it goes to your inbox, and then you can correct it. And digits learns instantly. And so when you go and decide, hey, yes, I want to book that to ground transportation to follow the example we’ve been using, then boom, it’s done. And if a new transaction comes in a minute later. That’s similar to that digital mimic what you just did. So you don’t need to worry about waiting a week or a monthly close or whatever for it to learn. It learns instantly. The core plan $100 a month now cheaper than QuickBooks, plus, which is amusing to us, is really designed for the sort of core SMB market folks focused on monthly accounting, working with their external accounting firms. This has been really, really well received by the market. Full power of the ledger. You have all the integrations we support 12,000 banks, cards, payroll, Stripe, I mentioned already, ramp, etc, etc. And so you connect everything in and you get custom reports to send out to investors. Yes, go ahead.
Randy Johnston 29:43
I’m sorry. Hate it to interrupt you, but before you go on, it’s $100 month, and we have covered that also in our podcast accounting Technology Lab on GL accounting. But one thing that’s not listed on your website. Jeff. Is number of users. Because obviously, if it’s the Qbo pricing at $115 that’s five users. How many users? Is $100 for you?
Jeff Siebert 30:11
Sorry, unlimited users. I philosophically do not believe in charging per user, and the reason is, I want everyone at the business to experience digits and get to know digits, and that’s how we spread by word of mouth. We’ve also designed digits to be very collaborative, and so part of the other benefit of this object graph data model we talked about is we have object level permissioning, and so you can share your marketing spend with your head of marketing, and they can see that live in real time, and they won’t see payroll or cash balances or whatever else. You can share your T and E with your office manager, and that’s all they’ll see. And so when you when you bring that mindset to it, I don’t want to charge per user. I want everyone at the company to have access to at least a part of digits so they understand the finances of the business. And to give a really concrete example, so we use it internally for digits. We have shared our software application spend, like, all the SAS stuff we use with the entire company. Everyone at the company can see real time everything that’s booked to SaaS. And that helps us stay honest on, like, what we’re paying for. What do we need? How many seats of those products do we need? Etc, etc.
Randy Johnston 31:19
Yeah, that’s a that’s a beautiful thing. So, you know, that’s one thing that if users went to your page, they would not see unlimited users, which, in my mind, is another core benefit you have. And I think in a consultation call I did with a firm yesterday, I said, Look, part of the deal on these systems is I need to have everybody using it with security bars, because you just described the security bars, and your model supports that isolation and exposure of the data in a proper way. But you know, of this particular group had about 100 users in their business. And they had 17 users, and they should have had 100 Well, maybe not the gender, okay, 99 users, you know, on the system, but they were going to be charged in one model by user, which I disagree with, and in another model, it was you already own the licensing, and you can use it so with digits friends, remember that it is unlimited users, which is another key deal here. And also you might address Jeff, the accountant access relationship as well. But I’m sorry to derail your explanation, because you were doing just what you should. But it was like, I need to call it the couple of three
Jeff Siebert 32:45
things, no that this is great and this is great feedback. I will maybe clarify our pricing page to highlight the unlimited users. Yeah, digits is unlimited users. It is unlimited connections. And the core plan is up to a million transactions a year. So a massive transaction number that should serve most SMBs, and that’s really the difference from our sort of professional plan. Yes, it’s custom pricing. If you’re doing more than a million transactions a year, we want to understand what your business is and how we can best work together.
Randy Johnston 33:13
Yeah, yeah, that makes good sense, because having dealt with the mid market and big ERP products, so say Gentec and NetSuite and SAP and Oracle Financials and epicord, you know, all those large competitors. You know, it’s a different ball game, but when it’s professional services, to me, the product almost has unlimited scale, with maybe one other little caveat, presence in foreign countries and where your data center is located, and currency and all those type of things. So where are you playing in that?
Jeff Siebert 33:47
Yes, so these are on our roadmap. So today, digits is us only and single entity. We don’t do multi entity consolidation. Today, you of course, can create multiple accounts and do the books. That’s fine. We just don’t consolidate. On our roadmap are both. So we’ve gotten tons of requests, as you can imagine, for foreign currency support and multi entity consolidation. So those will be coming. I can’t announce timelines yet, but today, all of the data is stored in the US, in Google’s us, central one, which is located in Iowa, all encrypted at rest. And so, yeah, that’s sort of where we stand today. Yeah.
Randy Johnston 34:22
And, you know, just a side point that our listeners have never heard from me. During the time that the AI models were being trained in Des Moines, they used 20% of the city’s water supply. So a lot of times people are thinking about power and not water when it comes to these AI data centers. But you know, it hoovered up a bunch of water in that window. But the reason for putting Jeff on the spot, on that friends, is multi entity consolidations and foreign court currency handling, including your EU triangulations on the euro. Those are not. Not trivial bits of code to write, because fortunately or unfortunately, I’ve had experience with all those easier than inventory and costing. There we go harder than general ledger. Okay, so when you’re weighing out your roadmap there, those are good places to go next, because we do have a lot of international entities. But I wanted to be clear, not today, and so my Canadian listeners here, sorry, not today. You know, value added tax and all those type of things. There’s, there’s so many parts on these big systems, but the core system going back to let you get on track again at $100 unlimited users. It’s stunning how much capability is in there. And I actually had zoned off and forgotten your million transaction limitation. But cool,
Brian F. Tankersley, CPA.CITP, CGMA 35:52
you know that’s, that’s a good, that’s a good problem to have because, because, for context, here, you’re talking about a 1200 page bank statement it’s 60, 6060, lines. Okay, so nobody you know, so you’re getting the bank statement, and it’s saying ka chunk when it hits the desk, if it’s in paper. So, so that’s, that’s a big deal. So let’s talk a little bit about payables. I told I spoke with my love for your payables engine thus far. And I want to challenge you and ask a quick question here. One of the things that some of the other tools, like maker sub and others do is they’ll extract the detailed information for each line item on an on a bill. Are you, are you to the point yet where you’re doing that and and you know where, where does that fall in your roadmap? If you’re not?
Jeff Siebert 36:40
Yes, so we already do that. That was actually one of the core reasons we built our own models, because, as we looked across the space, was there a model we could use to do this? It did not exist. That’s why, basically, none of the bill pay products have had line item extraction. So that’s the part we built first, and so it’s already in the product. If you go and you upload a bill, you can click the Split Line Items button, and it’ll instantly show you the line items for you to then book separately however you want.
Brian F. Tankersley, CPA.CITP, CGMA 37:06
Cool, very cool. Okay? And then once you map those to an item in your, in your in your system, then it will always map to that item or expense in the future.
Jeff Siebert 37:16
So we don’t, this is great feedback. We today, we don’t auto split the items, because we’ve seen most people are just booking their bills as single, like bulk entries. But I think that’s a great point. Is we should, if you are splitting the items, we should then default to split the items and then mimic how you booked them. That is great feedback I can bring
Brian F. Tankersley, CPA.CITP, CGMA 37:34
to the team. Well, the reason, reason I think about it, is, you know, you one of your core, one of your core things is franchises. And you know, if you, if you look at that, you know, if I think about your average restaurant, the biggest invoice they have is that bi weekly, bi weekly delivery they get from the jobber that has 500 items on it. And you know, those all go to the same place every single time. But if you can’t break out those 500 you’re not for me. So cool.
Jeff Siebert 38:04
Now this, that’s great feedback. That would be easy to do, and we would just default to splitting them and then booking them correctly.
Randy Johnston 38:09
Yeah. And so Jeff, as we think about there are specialty things, and I’m going to go to our first episode with you on this where Brian specifically called out the knowledge of vertical industries, and I think he illustrated with dental as an example, or this one probably fits more in construction or food franchises and so forth. This type of Accounting has been the domain of mid market products because of many things, but it feels like to me, not today, but future roadmap, your ability to do the analysis of items is it should be stunning again. Our listeners have never heard me say this before, but I actually wrote the original Hardy’s point of sale personally, and I was surprised. And I also did the McDonald’s point of sale so and Taco John’s and, you know, so I’ve been around point of sale in restaurants a long time, but I was surprised how, when we were trying to do the customization of a fast food order and to analyze the calorie account and to let the person know the amount of change that they had to count out. So if it was, you know, 87 cents, we would say three quarters a dime and two pennies, because the operators needed that. I guess on days we don’t need the penny thing, but you got it. There’s just these specialty things like rolling up the Z tape from the point of sales. And you know where your attitude around square was, I want all the transactions I can see a future vision where point of sale could also pull up all that analysis, whether it’s in restaurant or something else.
Brian F. Tankersley, CPA.CITP, CGMA 39:57
And I think about Yeah, and I think about that morphing into. A back office where you do the inventory accounting and the margin calculations and restaurants and I mean, there’s a lot of there’s a lot of interesting stuff here that that this contextual awareness makes possible,
Jeff Siebert 40:14
that that is exactly right. And so you’re correct. It’s not a today thing, but that is definitely the vision. And our model architecture has the ability to specialize by industry type or business type, and so that would allow us to build in that direction. Yeah.
Randy Johnston 40:27
So I know we could go on for another hour without much hesitation, but I want to be cognizant that we may only have five or 10 minutes left together. What other key features do you want us to know about across the digits platform? Because, again, up at the custom level, you know you’re reporting doing consolidation of financials, and you’ve already laid out you could run multiple instance of digits to do some of that. But what are other key things that users would find surprising or unexpected in digits today, not necessarily roadmap items.
Jeff Siebert 41:07
So So Today, our focus is on automating as much of the month end closed tedium as possible. And so we talked about the real time finances. You plug in your accounts, all the data comes in and is classified to the ledger in real time. There is no bank feed and digits. There are no rules. You don’t have to go and create and manually maintain rules. We automatically surface anything that the models didn’t weren’t confident on in your inbox. And then, as you just go through your inbox and book the items, the models are learning instantly. So that’s sort of the bookkeeping flow. Then there’s the reconciliation flow, we have also fully automated this. So for supported banks, there’s about a dozen banks who make their statements, their PDF statements, available via API, we automatically pull those statements and we’ll reconcile your accounts for other banks. Of course, we can’t access them by ourselves, but you can drag and drop the PDF onto digits, and we take it from there. And so the document extraction models pull out every transaction in the statement, map it to your ledger and literally reconcile to a pixel bounding box on the PDF. This is very, very cool, because you can then mouse over the transaction in digits and see it highlighted right on the PDF. So it makes the audit process just extremely fast as you go through, so that’s reconciliations, then we have a whole bunch of review tools automatically built in. So we call out all new vendors in a period. So let’s say you’re closing November, boom. With one click, you see all new vendors in November, and you can verify how they were booked and if there was a mistake or whatever, or something unexpected for that business. We show you all vendors and customers that have been booked across different places in the chart of accounts. Sometimes that’s intentional, of course, often it isn’t. And so that makes you give you one click spot to clean that up. And then we do all the variance analysis, flex analysis, etc, for you. And we attribute everything to the actual object level details. So the vendor, customer, department, exactly, et cetera, whatever it is. And so you can mouse over marketing and be like, Oh, marketing spend jumped 20% why? Because, well, this transaction with Google AdWords, whatever it might be. And so you don’t have to run detail reports, you don’t have to export the data, you don’t need to pivot it. We do all those pivots in the product. So you can just mouse over stuff and see exactly what’s going on.
Randy Johnston 43:24
All right, so as you’re describing these productivity tools, I’m reflecting on the need for transactional babysitting. Should be declining, so a misstatement that we’ve heard over the last three years since chat GPT was released, is AI is going to replace the accountant. And so the question here is, does digits replace the bookkeeper or accountant? Do you have an advisory path talk to me about that kind of thinking, because I could hear people thinking through the wires on this Oh, does that mean I no longer have a job, right?
Jeff Siebert 44:08
Yeah, that is not the case. I am convinced that it is not possible for AI to replace accountants. Because, as you know well, as you well know, the profession is a profession of edge cases and judgment calls and complexities for each individual business. These models like they can train, they can get as good as they can, but they’re never going to be able to reach 100% that’s why I described it as an asymptote. What our mission is is to eliminate the tedium, because 90 to 95% of that work is just pure tedium that you’re repeating every month and that we can reliably eliminate. And so the way we talk about digits is, unlike QuickBooks, zero, et cetera, you really have to use those products. You’re spending hours sort of putting the data in, pulling the data out, interpreting the data, et cetera. Our goal is for you to supervise digits, and so it allows you to reshape your role as. Reviewer and obviously trainer, by doing work in the product, you’re training the underlying models, and then allow you to spend way more time working with your clients on the advisory side, which, as we talk with folks, is where they all want to go. So I actually see this as hopefully a huge win for the profession, because even the folks on the junior side entering, instead of looking at a career path where they’re going to be spent doing data entry for potentially years, I think this can really shorten that and get allow them to really sort of accelerate their move on the advisory side and learning how to work with business owners.
Randy Johnston 45:33
Understood and as it turns out, when we’ve taught people to do client accounting services CAS and notice I separate advisory because I have a completely different path on advisory, but for Cas, which I guess we’ve taught now, 2728 years, at this point, quite some time, we’ve suggested that the teams are managed in pods five to seven People with, you know, some entry level people and managers above them and so forth, and that a typical pod might be able to handle up to about 100 businesses. So, you know, five to seven people, about 100 businesses. So what would be your expectation putting you on the spot a little bit, if I’m managing clients that are running digits. How many people do I need to, kind of oversee this
Jeff Siebert 46:29
for you know? So what’s the ratio? Yeah, so we have a concrete example of that today. There’s a pod of three, I won’t say the firm, but there’s a pot of three using digits to manage 150 clients. And so sort of a one to 50 ratio there. And that’s that’s happening today. I imagine that’ll keep scaling as we sort of fill out more and more of the different edge cases and steps in the product.
Randy Johnston 46:53
Yeah, I appreciate that you know known facts, because, again, we’d like to deal in the facts. One other twist, and it’s not so much of an issue here in professional sales and services business. But how are you handling tax, sales, tax, use tax, type of issues?
Jeff Siebert 47:13
Yeah, so we haven’t gotten that’s another big feature request. We haven’t built that into the ledger deeply at the moment. So right now that is being done on the more manual side. That’s a perfect example of something I think we will get to next year that will then eliminate a whole nother chunk of work. Yeah.
Randy Johnston 47:29
And again, Brian and I are lucky enough we’ve been around the design of all the bigs, the avalaras and the solos and so forth. And there are some products out there that do it. But here’s another reminder for our US listeners. You know, sales tax in the US is very Byzantine. You know, 17 to 18,000 taxkeeping jurisdictions, as opposed to Value Added Tax, GST and so forth, globally. So it’s, it’s
Brian F. Tankersley, CPA.CITP, CGMA 48:00
truly the, it’s truly the death of 1000 cuts. You know, it’s kind of like, it kind of like Afghanistan. It’s where software empires go to die. Is they take on that, and they take on 50 state payroll, and if they don’t really know what they’re doing and haven’t really done their homework, it’s, it’s, it’s a quagmire.
Jeff Siebert 48:19
Yeah, this is exactly right. And this goes back to what I said earlier. We won’t build 50 state payroll. We don’t intend to build the sales tax engine for the US. We’re going to partner with one of the top players in the space and build it into the ledger that way.
Randy Johnston 48:31
Yeah, makes great sense. Because, again, I wanted to differentiate that Jeff in terms of, you know, what’s in, what’s out. Well, you know, you have been such a wonderful guest. And you know, I again, I’ve known you for a few years, and every time I’ve listened to you, it’s like that guy is such a good guy.
Jeff Siebert 48:48
So I appreciate that, Randy, that means a lot.
Randy Johnston 48:51
But are there parting thoughts that you have for our listeners today?
Jeff Siebert 48:56
This has been a really thorough overview. I deeply appreciate it. And of course, your knowledge of the space is beyond parallel. And so no, I would just encourage folks to check it out. You can start a free trial. There’s zero commitment. Hook up your accounts, either your own personal ones, or for a client. A cool tip I have is literally, if you want to just try it, start a trial, hook up your personal credit card, and it’ll just book everything as if you were a small business, which is pretty funny. And so you can see how the model works and like what it’s doing, and then cancel the trial. So just, just check it out. We would love any feedback,
Randy Johnston 49:26
Jeff, that’s a great recommendation. So Brian, parting thoughts or questions for Jeff.
Brian F. Tankersley, CPA.CITP, CGMA 49:33
You know, I Jeff, I’m very impressed with what you and your team have done, and I think it’s, I think it is a new world that you’re leading us into, and I think it’s very exciting. You know, I very impressed with what you’ve done thus far, and I’m really looking forward to seeing what you and your team come up with in the future. Because, you know, the questions we’ve been asking you about things like multiple lines in an invoice. Sales tax and those things, you know, these are questions that that we ask people that are sometimes very, very hard for people to answer, because many times they haven’t even thought about it. It’s very apparent that you have thought a lot about this, and you get a lot of really smart people thinking about it. And I encourage people to go and check out your work and see if they like it, beautiful.
Randy Johnston 50:24
Well, Jeff, thank you so much for investing your time with us today. I know your time is wildly valuable, and we are pleased to get the amount of time with you we did so again. Do start a free trial of digits, at least give it a good run. That’s our recommendation, and we look forward to speaking with you again soon in another accounting Technology Lab. Good day. Thank you.
= END =
Thanks for reading CPA Practice Advisor!
Subscribe Already registered? Log In
Need more information? Read the FAQs