How SMBs Can Fix Bookkeeping to Build Trustworthy Data and Scale Finance with Kevin & Drew
In this episode of Future Finance, hosts Paul Barnhurst and Glenn Hopper sit down with Kevin A.Thomas and Drew Hyatt, the co-founders of Omniga, to discuss the often-overlooked but critical challenges facing small business finance teams. While enterprise companies get most of the attention, Kevin and Drew are focused on the backbone of the economy, businesses under $30 million in revenue—and the unique struggles they face when it comes to closing the books, managing data, and generating strategic insight.
Kevin A.Thomas, CFA, is the founder of Omniga and a seasoned strategic finance leader focused on scaling digital businesses. Kevin is passionate about rebuilding the back office to help businesses make smarter decisions, faster. Drew Hyatt is the Co-Founder and CTO of Omniga. He plays a key role in architecting Omniga’s infrastructure, translating complex workflows into scalable technology for finance teams.
In this episode, you will discover:
Why clean books are the foundation for strong financial decisions
The hidden risks in small business bookkeeping and how to fix them
The difference between replacing tasks and supporting human judgment
How to expand from bookkeeping to full finance strategy support
How to structure a finance workflow that scales with your business
Kevin and Drew leave us with a clear message: better finance starts with a stronger foundation. By focusing on clean books and scalable workflows, Omniga is reshaping how small businesses manage finance. It’s not about replacing people, it’s about empowering them to do more with less.
Join hosts Glenn and Paul as they unravel the complexities of AI in finance.
Follow Kevin:
LinkedIn: https://www.linkedin.com/in/kevin-a-thomas/
Company: https://www.linkedin.com/company/omniga-ai/
Follow Drew:
LinkedIn: https://www.linkedin.com/in/drew-a-hyatt/
Follow Glenn:
LinkedIn: https://www.linkedin.com/in/gbhopperiii
Follow Paul:
LinkedIn - https://www.linkedin.com/in/thefpandaguy
Follow QFlow.AI:
Website - https://bit.ly/4i1Ekjg
Future Finance is sponsored by QFlow.ai, the strategic finance platform solving the toughest part of planning and analysis: B2B revenue. Align sales, marketing, and finance, speed up decision-making, and lock in accountability with QFlow.ai.
Stay tuned for a deeper understanding of how AI is shaping the future of finance and what it means for businesses and individuals alike.
In Today’s Episode:
[03:39] – Why They Started Omniga
[07:11] – Drew’s Tech Journey
[09:28] – What "Omniga" Means
[11:26] – Rules vs. Suggestions
[14:33] – How the Workflow Scales
[19:28] – Above the Ledger
[25:40] – From Inputs to Entries
[29:50] – Real-World Use Cases
[35:05] – When to Add FP&A
[37:45] – Closing Fun Segment
Full Show Transcript:
Co-Host: Glenn Hopper (00:41):
Welcome to Future Finance. I am your host, one of them Glenn Hopper, along with my esteemed colleague, the right Reverend, the FP&A guy Senator, Dr. Paul Barnhurst. You
Host: Paul Barnhurst (00:55):
Can just call me Reverend Guy,
Co-Host: Glenn Hopper (00:57):
Right? The FP&A. Reverend, that's actually, that's got a nice ring to it, I think. Put some holiness on your FP&A. Oh boy. I'm waiting
Host: Paul Barnhurst (01:06):
For the
Co-Host: Glenn Hopper (01:06):
Light. What am I doing? I'm doing an intro to the show, Paul, let me keep going. Why aren't you interrupting? Today we're joined by Kevin Thomas and Drew Hyatt. They are the founder and CTO behind Omniga.ai. Omniga sits in a familiar place for anyone who's ever had to clean month end numbers. In a QuickBooks world picture transactions are there, but the work is in everything around them, figuring out what something really was, chasing context when the memo is useless, handling weird one-offs and making sure someone actually looked at the transactions and calls that matter. AMGA is building a layer on top of the general ledger that helps with that review workload. The system can propose how to classify activity, draught entries, but it's built so that a human can review, fix, and approve before anything is booked. The promise is to save time without turning the close into a guessing game. So Kevin Drew, welcome to Future Finance. We're glad to have you here today. Thanks
Guest 2: Drew (02:05):
For having us. Thanks for having us. We really appreciate it. Yeah.
Co-Host: Glenn Hopper (02:07):
I don't know, Paul, you want to kick us off since I feel like I've been rambling for 10 minutes. I'm just wasting every time. No,
Host: Paul Barnhurst (02:12):
I keep, you're doing great. No, a serious note again. So why don't we start here? We'll start with you, Kevin. So you spent years on the corporate finance and FP&A side with that CFA lens. When you look back on the operator seat, what was the recurring pain in the bookkeeping and close layer? It really pushed you to starting omi. How did it kind of come about? What was the impetus there?
Guest 1: Kevin (02:39):
The pain points I noticed weren't really limited to bookkeeping in close, but in the lower and middle markets, it's become obvious that that's where you have to start. There's a few signals I think that have made that clear to me. I think first is just the market itself. I mean, if you look at the size of the bookkeeping and client accounting services market compared to say FP&A or strategic finance, it's considerably larger. Second, there's very consistent hiring pattern in small businesses where typically the first finance hire is almost always an accountant or a controller. Then eventually maybe they bring in A CFO and it's only much later if ever, that a dedicated FP&A function comes into play. And then third, everything else in finance kind of depends on having clean tax ready financial statements. So when you get into that below roughly $30 million in revenue, the ledger usually isn't a stable source of truth.
(03:43):
And it's not because people are careless, but it's because controls and ownership around the books is pretty thin. So I watched a client build an investor presentation off of a QBO file recently, where prior bookkeepers had run owner contributions through revenue, or at least their prior bookkeepers never raised the flag to it. That was kind of the moment it clicked for me that the foundation really was the problem. And so if the end goal is scalable FP&A or CFO level insight for smaller businesses, I don't think we start by building better models for every single client out there. I think you start by fixing the layer that feeds every model and then that meant redesigning bookkeeping and closed first, not necessarily to move faster, but to increase granularity at the source and to kind of embed that review memory and control directly into the ledger so that contacts compounds over time instead of being recreated every month.
Host: Paul Barnhurst (04:45):
So I see it, the pain point is you wanted to help the business be better at what it's doing, the operations, having that information to make decisions, but in order to address that, it starts with how do we make sure the data is cleaner, earlier and better? As much as I love to run my contributions through revenue, I know it's not a great way to, how about you, Glen? Are you still running yours through revenue? Last I checked you work?
Co-Host: Glenn Hopper (05:08):
Yeah, but I do the entry backwards so that they bump revenue for me and then more money in my pocket. It's a huge success
Host: Paul Barnhurst (05:16):
Story. You're going for the lower tax liability. I get
Co-Host: Glenn Hopper (05:20):
It. So I love Kevin and Drew, the combo that you guys bring with the domain expertise, seeing the problem, and then Drew on the tech side being able to take what you've visualised problem you've identified and do under-the-hood magic to make it work. And I'm sure there's a collaborative work that goes on around that too. But Drew, I know a little bit about your background, but for our audience, maybe give us the 15-second version of what you were building before Omniga, especially relevant to this audience, the FinTech and integrations work, and then maybe bring it forward a little bit when you're building AI into accounting workflows. Actually, I'm going to make that two questions.
Guest 2: Drew (06:03):
Okay.
Co-Host: Glenn Hopper (06:03):
Let's talk about your background, and then I want to jump into the AI side because I think it will move smoothly, but I don't want to gloss over any of that.
Guest 2: Drew (06:11):
Oh, absolutely. It'll dovetail nicely. I'm not going to give everybody my full history, but as of recent work in two work streams was built on technology. So construction finance was a founding member of the team there to facilitate payments rails within the business. So working with payment service providers to move money between parties, collecting compliance documentation for construction projects, typically between general contractors and their subs, being able to manage that stuff. Also integrating with QuickBooks, Sage, inta, Procore, things like that. That was about three and a half years there kind of building that kind of a startup within the startup. It's a little bit bigger. So anybody who needed to move money at builds, they had to come through my team, if that makes sense. And then as of late before I started working with Kevin, and we go way back, we're college roommates. I was kind of being his fiduciary for free and just kind the kindness of my heart from the tech STEM point, but my last gig with peer supply, which was kind of medical supply chain optimization and layering AI into that workflow specifically with LLMs.
(07:18):
So kind of how that process works is going to do the traditional blocking and tackling of data ingress and just being able to use kind of illuminous amounts of data. But once you get disruptions in the supply chain, that's where AI comes into play. Think about 10 different positions trying to figure out a replacement for something they need during a procedure and injecting LLM into that email thread or that conversation thread. And also giving an access to that kind of Illumina data set to find replacements and find areas to optimise the supply chain so that we had a near-death experience at that startup and decided to hook up with Kevin and be his fingers here with ga. I guess that kind dovetail into how we're using and leveraging AI in kind of this capacity, if that makes sense.
Co-Host: Glenn Hopper (08:10):
And I guess when we were talking, so wait, I was saying Omni niga, but is it Omni GA as in omnipresent, Omni ga? Walk us through that for a second. I don't dunno, Kevin or Drew, whichever of you prefers to do that, but explain that. And then I want to dig into the AI workflows.
Guest 1: Kevin (08:28):
So the name comes from Omni GNA, kind of the end game of what we're building here is we want to build a kind of fractional back office, the full GNA stat of the p and l, so to speak, where you can really kind of rely on us for administrative duties end to end. And so that's where the name Omni GNA comes from Omniga. But yeah, drew and I kind of go back and forth in how we refer to it.
Host: Paul Barnhurst (08:57):
So Glenn, what should we refer to it as going forward? Should we do omni GA or the other one?
Guest 1: Kevin (09:02):
Yeah, let's go with Omni ga. Do you like Omni ga? Hey, hey, it is up for a vote on this call. We can make it official right now.
Host: Paul Barnhurst (09:12):
Ones it going
Guest 1: Kevin (09:13):
To be,
Host: Paul Barnhurst (09:14):
Should we leave it up to ai? We'll ask three tools and which there
Co-Host: Glenn Hopper (09:20):
You go,
Guest 1: Kevin (09:20):
One
Host: Paul Barnhurst (09:20):
Selects two is the winner or all three for that matter, but the majority.
Co-Host: Glenn Hopper (09:24):
So I guess where I was going to go ultimately with that is I was fascinated by, because this is the problem that anyone who's trying to bring AI into accounting and finance right now is facing, and I know you guys are dealing with this too, because a lot can be handled with basic rule and some things are better in a deterministic programmatic. This is how we do steps one through 13, but there are areas where generative ai, the probabilistic nature of it, the sort of natural language of it and everything are more valuable than tools we've had before. So as you guys are thinking about accounting workflows, how do you approach what you keep deterministic, whether it's classical machine learning or just rule-based code where you have standard outputs versus what you are willing to let a model suggest and provide insight obviously with that human review?
Guest 2: Drew (10:26):
Yeah, I'll speak to that. I don't want to get too technical, but we use kind of a mixed mash of traditional web dev blocking and tackling, right? Relational database, non-relational traditional APIs. Front end got 10 years of experience in that. But where the AI comes into play polymorphic relationship at the database layer. So you've got traditional tables to house your facts layer, but then you layer in the LLM artefact and generative AI in a way that's permission G. So depending on whatever first class citizen you need assistance with in the accounting workflow, the generative ai, whether it be LLM or whether it be predictive and more concise, it's generating artefacts to help you along your flow, kind of knocking out that busy work and letting you just kind of arbitrate over. Most people read faster than they can type. So you permission gate what the AI can output and what can it do on the system, if that makes sense. We get into a little bit further detail on the nuts and bolts of that, but each layer in the data ingestion pipeline and then decision tree, if that makes sense.
Co-Host: Glenn Hopper (11:35):
Yeah, and I bet that seems like that would be really important in building customer confidence and trust, and most of your users aren't going to understand the difference, but when it's giving them the answer that they expected and limiting potential for hallucination, and if you turn over the wrong task to generative ai, I guess that would be the problem.
Guest 2: Drew (11:57):
Yeah, we don't let it do too much now, but going to give a more concrete example. Say you've got the shoebox data coming in, you've got a layer of OCR and just a kind of simple deterministic. What piece of data is this? Can it be traditionally matched with relational db? Do we have an exact date, an exact amount, exact vendor customer reference? And then where that fails, you kind of bring in the generative UI for light suggestions, classifications easy, one click kind of pass through arbitrates over whatever you're interacting with at the time. Yeah,
Co-Host: Glenn Hopper (12:33):
More like a partner than a replacement.
Guest 2: Drew (12:36):
Absolutely. Yep. And you get a good feedback loop too, that helps you build the context as you go along, informs things later down the line, reporting decision making. I'm going to suggest some rules based on patterns I've seen in the past. You let the AI review in the background, not necessarily user facing, but goes through your traditional dataset and suggests what can be automated away for you. If you're tracking with me there.
Host: Paul Barnhurst (13:02):
This question is towards you, Kevin. You've said the model is one, fractional CFO, running a portfolio appliance to the platform. So I'd love for you to kind of describe the operating system. Who on the team does the first pass, who receives it, who's it escalated to? What's the role of the CFO kind, how you see that working, make sure they keep the books consistent, especially when you have five clients and you go to 10 or 20 or 30. So maybe walk through a little bit of that.
Guest 1: Kevin (13:33):
You're absolutely right. I mean, I think the kind of unit of scale is the fractional CFO to however many clients, but it could just as easily be a bookkeeper potentially. Basically the operating system is designed to kind of shrink the surface area of human judgement as volume increases without eliminating judgement altogether. So the first pass is kind of the automation pass and it's very staff level by design where you've got your documents, your bank feeds, your exported files from your revenue sources or whatever source it may be. They come in, they get normalised, and then after normalisation, the system determines how they need to be accrued, classified, and it applies a score of confidence against all of those decisions. High confidence items that conform to the existing rules they can flow through automatically. If they're lower confidence items, they go to a review queue and they get handled by staff or kind of a central function that is there to review these low confidence items, not A CFO.
(14:51):
Typically then certain things are designed to be escalated. So a new vendor comes in or there's an unusual pattern, something has a material variance to something that the system identified as a match to it. Something breaks a rule that maybe we have as kind of a guardrail for a given client. So that gets flagged automatically and then summarised for a controller layer that step above the staff layer. So I think what kind of allows us to go to scale or scale from five to 30 clients without quality degrading or really just kind of the hard control gates where nothing posts without a rule or a user configured confidence threshold or human review periods are locked centrally, so nothing's getting put in prior periods, which is a very common thing I've seen across the QBO world. Every override and escalation gets logged by the system. So it kind of feeds into the pattern data Drew was talking about earlier.
(16:01):
For an AI to raise concerns or common occurrences, context carries forward. So the system remembers how items were handled before and proposes those treatments again. So your judgment's getting reused instead of the user having to recreate judgments over and over and over again. There'll be a surface area for chat, but you can never post from chat. Its role is really to prepare and to provide context and support review, but not execution. Then under the hood, the system will constantly be reconciling against the banks and other source systems. So yeah. Then lastly I think is kind of produce production of work papers. So after it does a reconciliation on a monthly basis, there's kind of an audit artefact that's placed out there for you to prove that this reconciliation has been done and that things were good. So then lastly, you get to the CFO level work, and that's where the work looks completely different, but it feeds off of everything that's happened up until this point. So now the CFO is at reviewing transactions or just looking at parent level accounts. OPEX went up X percent. They're actually able to review dimensional insights inventory by location or customer growth by channel margin shifts by product. The data's already reconciled, it's queryable and it's exportable. CFO's job can really be about interpretation and not assembly. So yeah, I think that's how we intend to kind of build scale.
Host: Paul Barnhurst (17:36):
Got it. No, thank you for sharing that. Glenn, what are your thoughts?
Co-Host: Glenn Hopper (17:39):
Yeah, so I think as you go through what GA does, I think we need to be clear on that Omni GA is ledger agnostic. It sits on top of QuickBooks or whatever system. And this approach is super interesting to me because right now so many people are going after mid cap and above last count, Paul, I don't know where we are numbers wise, but in December, I think over half a billion dollars had been invested in new ERPs that are targeting the mid-market and above companies. This is something, but you think about how big the market is with small businesses that are under 30 million, under 10 million that could really use something like this
Host: Paul Barnhurst (18:28):
Because it economy runs on small businesses, even though all the attention is often given to the
Co-Host: Glenn Hopper (18:34):
Large. And if you think about the problems that small businesses are having, a lot of times maybe somebody's spouse is doing bookkeeping in their spare time or they have a fractional accounting team on some level or maybe a fractional CFO, but not really having a team focused on this and with a layer that's provided by this platform, I think it gives some more insights. But I guess when we're talking before the show, you both referenced ledger agnostic and I think it's important to understand where this platform sits. And I guess my question to you would be why sit above QuickBooks instead of becoming the new, trying to go head to head with QuickBooks and really the value proposition and what problem you're solving for your customers and what gets harder when you try to make that reliable across different businesses from your perspective, obviously to the customers, you want them all to have a similar experience, but as you're building this out, a lot of difference between a services company and a manufacturing company or whatever the case is there
Host: Paul Barnhurst (19:41):
Ever feel like you go to market teams and finance speak different languages? This misalignment is a breeding ground for failure in pairing the predictive power of forecasts and delaying decisions that drive efficient growth. It's not for lack of trying, but getting all the data in one place doesn't mean you've gotten everyone on the same page. Meet QFlow.ai, the strategic finance platform purpose-built to solve the toughest part of planning and analysis of B2B revenue. Q flow quickly integrates key data from your go-to-market stack and accounting platform, then handles all the data prep and normalization. Under the hood, it automatically assembles your go-to-market stats, makes segmented scenario planning a breeze, and closes the planning loop. Create air-tight alignment, improve decision latency, and ensure accountability across the team.
Guest 1: Kevin (20:49):
Yeah, yeah. So I guess when we say ledger agnostic, basically we're trying to be deliberate about the problem we're choosing to solve. We sit above QuickBooks online, not because we don't think ledgers matter, but because for our customers the ledger isn't necessarily the bottleneck. The bottleneck is everything around it. It's the review, it's the controls, it's the context, it's the reconciliations and the multi-client workflow. So QuickBooks is a really good system of record. It handles compliance, it handles tax workflows, has an enormous ecosystem of add-ons and additional bolted on capability. Replacing that would basically force customers to migrate data. Restrain teams take on risk without actually solving their day-to-day problems that make finance hard at scale. So by sitting above the ledger, we solve a different set of problems for customers. We let firms and fractional CFOs across many clients, work across many clients without being locked into a single system.
(21:57):
If one potentially a fractional CFO has client with QuickBooks and another client with zero or some other kind of well-documented API system, we can support all of those ledgers. We create a consistent review control and evidence layer When the underlying ledgers differ, we reduce switching costs. Customers can change ledgers without losing process history or institutional memory, and we allow finance teams to scale judgement and oversight without rebuilding workflows for each individual client. So that being said, being ledger agnostic makes things a lot harder technically, which I'll let Drew speak to a lot of those difficult technical implementations, but the difficulty is the point, right? I mean if we required being the system of record, we'd just be another accounting platform competing on features. But by staying ledger agnostic, we're able to focus more on orchestration and workflows and making finance work reliably scale across tools, clients and teams. And for a multi-client professional especially, that's the difference between managing a portfolio of clients versus being trapped inside of a single system for each client.
Co-Host: Glenn Hopper (23:16):
As you're talking about the workflows, I'm picturing now I'm trying to picture, drew, maybe you can walk me through this a little bit, but it's funny, everybody, Paul and I, we talk about AI all the time, but a lot of times when people say ai, they mean automation. And I know a lot of what you guys are doing really is this automation layer and it's designing these workflows and transforming data in a way that's useful to the clients. I know there's a generative AI layer like you talked about Drew, but I'm trying to picture when you take that messy shoebox data, CSVs exports or seats, partial descriptions, help us visualise what that actual workflow is from ingestion all the way to, I don't know if it's a review queue or whatever it is, all the way to posted entries and where you see the bottleneck or the problem that people that your customers have where accuracy breaks down and these things get confused and they do stuff like post their owner contributions to revenue or things that are wrong and have gone wrong forever. And how much of that's an AI layer? I don't know, I may be asking too much of a question here, but if you, I'm just trying to picture that workflow and where rules and AI step in and how it looks to an end customer.
Guest 2: Drew (24:40):
Oh, absolutely. No, that all makes sense. It's not too hard of a complex question there. So a lot of it is traditional patterns that I've learned what works with traditional web dev, no AI involved at all, and external integrations like bi-directional sync. So for in this use case, first step is let's connect your ledger. That's as easy as logging in, clicking a button and logging in and your data's getting piped into Omniga immediately. And then kind of the enrichment ingestion AI layer comes later in that process. Same thing with loose document shoebox stuff, your bank connections connect with cloud and just getting that data seated perculating a task for somebody non-technical, right? But 10 years experience, we know how to do that. And once you've done one API integration, you've pretty much done one because really all you're doing is keeping the data in sync and then kind of mapping it to our first class system, more or less our data structures, which are pretty flexible.
(25:42):
Then that's where AI comes in and starts to eliminate the busy work. That whole task of the data movement and consolidation is hard enough, but then you start to take action on the data, use traditional tools like off the shelf open source OCR, kind of get a subset, use predictive AI to see, oh, what kind of data structure is this, if it's loose or unstructured. And then you start to layer in matching auto classification or we start to seen it break down a little bit is what you bump up to that context window. So we're having to put some guardrails on kind of the LLM integration that we have needs to be very targeted kind of at the transaction or entity level and then giving it the proper context. Try to do matching classification on a bulk set of things, entities. That's where it starts to break down.
(26:36):
It's really bad at matching sql, just does the job better in that case for us, traditional kind of blockade tackling there. And then the sky's the limit in that polymorphic pattern I spoke about previously. Anything that the LLM can think up and you can define a structure for if you need a structured output, it's very, very good once you kind of give the guardrails to what the context window is. And it's chain logic too. You're not going to one shot things like this. You reach for some of the off the shelf open source tools to help chain that and build your algorithm. So it's a hybrid, hybrid of traditional and then injecting AI where appropriate, whether after the OCR layer to a more complex structured output for bank statements or invoices. We've nailed that, which is great. Kind of the standard documents that are loose leaf AI really shines there, which if you get good OCR, which is not ai, but the name entity recognition and structured output of those things, that's highly important. We've nailed that already. I'm pretty excited about having shipped that puts a big problem out of our way, but hopefully I wasn't too long-winded there. I could probably talk more on the details, but I'll put a pin in that for a second. Yeah,
Co-Host: Glenn Hopper (27:54):
No, honestly, this is the kind of stuff I'm talking about every day. And Paul, you and I have talked about this on a show before and offline AMGA is solving it for small businesses, but this is the same problem that every finance department in the world is having right now. It is they're being told by their CEO or board or investors or whoever, you need to do some ai. But the problem of doing AI is not knowing where to apply it, how to apply it. And so solving for, if we go with that bigger umbrella of ai, it can be, let's just say we're going to automate stuff that we used to have to do manually and we'll put AI in where it makes sense, but we're not trying to just sprinkle AI across. But I don't know, Paul, I know we talked to a lot of midcap and enterprise level customers.
Host: Paul Barnhurst (28:50):
I think there's a couple of things here, particularly your comment. We've all heard it, you just got to get AI out there, go implement ai. The reality is the CEO, the CFO, whoever's saying that doesn't care if it's ai, if it's automation, if it's machine learning, if it's gen ai, if it leads to better insights, if it leads to cost savings, if it leads to benefit, do they really care? And outside of they want to say they had the AI win, they don't, right? And so I think you talk about a lot of times we hear gen AI so much, but the reality is these solutions need to be holistic technology and companies are taking advantage. Interesting. Kind of bringing this all back, talking to a lot of companies, I've talked to a company today, I think it was this today, I lose track. I talked to so many it feels like, but they kind of build a financial, an AI financial analyst focused at the midmarket and the guy's like, yeah, we realise it's kind of a new category problem.
(29:45):
It's not a planning tool, it's not a new Yorkie, it's not the treasury platform. Are you going to get financed to spend more money goes right now we're really very heavily into, it's your AI solution. And he's like, the reason is is everybody's been given AI budget in two years. It'll be like, well, that's really an analyst, that's a new category. Do we want to spend money on it right now? It's an AI solution that you can spend your budget on. And so I just think it's so interesting. I know I'm getting a little off track here, but just watching this and everybody uses the term ai, but at the end of the day, it's about solving problems and using technology and AI has enabled us to do more. But so often, as I think Drew said, and Kevin is right, deterministic, maybe you're route to go or automation or RPA, it's, it's really about the holistic solution. I ran there for a minute, so I could tell Kevin nodding. Any thoughts there?
Guest 1: Kevin (30:40):
Yeah, well no, I completely agree. I think it's one of those things, garbage in, garbage out, and I think that's kind of the approach we've taken with starting at bookkeeping. Make that initial data layer as clean and structured and high quality as possible and then figure out how AI can exponentially improve your operation from there. But you can't just kind throw AI at the wall and hope for the best.
Host: Paul Barnhurst (31:15):
I'm trying
Guest 1: Kevin (31:16):
Kidding. We all a little bit. A little bit anyway.
Host: Paul Barnhurst (31:21):
Yeah, I mean I totally agree. You don't have clean data. If you haven't thought about it, you're going to be one of those that we spent a bunch of money we have to show for. And we're hearing lots of stories of that. Everybody's piloting something right now. Most people are having the level of success they want. What will lay down the road? It depends. We'll see. But it's exciting nonetheless. But like you said, you got to put the guardrails to really be thinking about this.
Guest 2: Drew (31:51):
I'll add onto that too. We haven't released this, we're just been testing this internally, but we're getting very close to that foundational contextual data being in place. Where that generative AI does shine is we kind of let it loose in the background to review the entire state context per client, and then you start to get real actionable insights immediately or with a little lag, right? If you've got a 2, 3, 4, or five years of history, it takes a little bit of time for the AI to crawl that and understand it. But we're producing weekly actionable C-level information from the clean books that we're doing, which is cool to see. And then people say ai, that's probably what they think, right? I've got, let me run chat GPT on my books and tell me what to do.
Host: Paul Barnhurst (32:44):
Yeah, so many are looking for AI to be able to crunch the numbers and give insights that they can then action. And what's funny is I think that first year at least gen ai, more people are using it to help be creative. We all thought it would free us up to be more creative, but often you're using it to help write the article, to generate ideas, to help you build the training. And over time, I think we're seeing more and more of what you're talking about where let's figure out the ways to let it loose on the right dataset in the right environment where it can start to return meaningful information that we can then action because that's what finance wants.
(33:23):
So as we kind of zoom out and wrap up here, I know we're kind of near the end of our time. We have our AI generated section next that we'll share with you. I have a little fun here, but before we do that, as we zoom out, you described the end game is having that fractional finance department as a service. Let's talk about software as a service. This is more of that, but you've also said clean books come first. How do you decide when it's time to add the FP&A outputs? Like the cash forecasting, the budgeting, all the planning stuff. And so I'd love to get a little thoughts and just keep it, how do you make sure you're always keeping that foundational layer clean but adding in all that kind of FP&A into this layer?
Guest 1: Kevin (34:05):
It's a good point. I think our end game and our entry point are intentionally different. The end game for us is, as I mentioned earlier, it's Omni GNA, right? Our namesake Omniga an operating system for how small and mid-market businesses can run their back office from end to end. But our near term wedge is fractional finance, and we'd even sequenced that deliberately from bookkeeping first and then to controlled finance. And then all the way up to kind of the CFO level insight. We don't unlock FP&A until the accounting signals say the foundation is state. So closes are predictable, reclassifications are trending down, reconciliations are tying out, review volumes dropping instead of spiking each month. And once those signals are there, I think we kind of take the same control discipline that we took into the accounting layer and we move it into the planning layer is an easy add. And so yeah, I think the sequence is intentional. It's fractional finance because it forces rigour and then FP&A once the data earns it, and then omni GNA over time as those principles extend across the rest of the back office. But yeah, I mean I think our whole point is to make strategy compounding possible by fixing the foundation underneath it first.
Host: Paul Barnhurst (35:30):
Yeah, I mean, Glenn, when you've come into a company, I know you've done fractional services, how often have you started with strategy
Co-Host: Glenn Hopper (35:37):
On day one? Yeah, never. Yeah. How often have you started with cleaning up data? You do know what a chart of accounts is, right? Exactly. Exactly.
Host: Paul Barnhurst (35:46):
It starts with the data and it starts with the clean data because we every company, I mean, I know there's some things, shouldn't admit this being an f, p and A and accounting person, but there's so a few things I do with my books for simplicity. I'm like, I'm small, they're in material. I'm not having 'em fix 'em. And so I get it and I think, okay, someone who doesn't know FP&A, how many things like that are they doing on a compounding level that have no idea about accounting or bookkeeper just doesn't want to deal with it? Mine I could fix, and there are a few thousand dollars here and there, but they're not material to my business. And so I think, okay, if I'm doing that, how bad is it for the average person?
Guest 1: Kevin (36:22):
Yeah. Well I've had the chance to see it firsthand over the last eight months or so and there's folks running contributions through revenue.
Host: Paul Barnhurst (36:33):
Like I said, Glenn and I are going to start that. So if we can grow our businesses, alright, so we have this section we do, Glenn and I do, and Glenn, who do you want to take? We'll each do one question. You want Kevin or Drew?
Co-Host: Glenn Hopper (36:45):
I feel like I need to make an excuse or an apology from CHATT PT. So
Host: Paul Barnhurst (36:52):
Yeah, I always going to get there. They aren't pretty loudly. Go for it.
Co-Host: Glenn Hopper (36:55):
So every week when we do an episode, we feed your LinkedIn profile and show notes and everything into an AI and we say, come up with 25 off the wall questions and we kind of bounce around different models and I don't know, so chat GPT 5.2 is to blame for these. These are the worst questions we've ever got. So what we normally do is, Paul and I have different approaches to it and you can either pick a random number between one and 25, or Paul has a random number generator there that he'll pick one for you. Or I always just go back to the AI and say, just pick one of these and I'll ask it. But these questions are so bad, I'm embarra just
Host: Paul Barnhurst (37:38):
These instead of actually asking them just so people could see how bad they are, who votes for that?
Co-Host: Glenn Hopper (37:46):
And if you guys want to answer one jump in.
Host: Paul Barnhurst (37:49):
I'll read a first couple and then you can read a couples, we'll switch off. We'll see what you guys think and you're welcome to answer some. So let's maybe take five minutes or so and have a little fun here. So if Omni GA were a restaurant, what would be on the menu and what would you call the house special. Alright. All right. You guys could think about that. Number two.
Guest 2: Drew (38:09):
Yeah.
Host: Paul Barnhurst (38:10):
What's the most harmless looking transaction description that immediately makes you want to click into the details? Yeah, Kevin's a look right there. Go ahead. Go ahead. Read a couple. Glenn, you're up.
Co-Host: Glenn Hopper (38:24):
You know what I think I did here? I think the problem is I gave it too much context because I was researching your website and going through it and it knows everything about what you guys do normally it's just like the LinkedIn profile. So I think too much context made it really boring and weirdly specific. Let's see. If you could hand every small business owner one laminated card, what would be the three rules if you had to teach a golden retriever to do bookkeeping? This is when I said that's not funny. Be funny. It said, if you had to keep a golden retriever to do bookkeeping, what's the first task you'd to sign? If you had to pick a walkup song before explaining AI to a sceptical accountant, what would it be? I dunno. There might be a funny answer there, but to spring it on you live, I don't know, I'm, I mean
Guest 2: Drew (39:18):
That's an easy one. The
Co-Host: Glenn Hopper (39:20):
Bring
Guest 2: Drew (39:21):
Smooth operator by Saade.
Co-Host: Glenn Hopper (39:24):
Nice, nice good charade reference. And
Guest 2: Drew (39:26):
Then I'll throw one, I answer to the transaction one, the most concerning transaction description. So it is tangentially related, but Plaid seems to classify me as Hyatt Hotels. That's my last name. So the enrichment data that Plaid puts on there, I apparently I'm the Hyatt hotels, but I've never received any pay downs for that.
Host: Paul Barnhurst (39:53):
Here's another pretty good one. I'll do one more. One or two. Let's see if the AI could speak out loud in the review queue. What's one line you'd want it to say when it's on? Sure.
Co-Host: Glenn Hopper (40:04):
Oh hell no. That's probably the closer right there. I don't know. I think the lesson for us, Paul, is
Host: Paul Barnhurst (40:14):
I a good closer. I agree. So as you could see, AI blew it on the question. The golden retriever retriever one was just, wow, if you can come up with a good answer for that, we'll have you back. Just kidding all. Alright. No, Kevin, drew, thank you so much for joining us. Glenn, thank you for setting this up. We enjoy chatting, exciting to see as you built that layer and what automation and AI allow us to do exciting times. And we look forward to hearing your answers to these great AI generating questions when you come up with them. So Glenn, any last thoughts?
Co-Host: Glenn Hopper (40:52):
It's interesting. We sometimes forget the small market. And I love that there are people like you guys out there that are focused on them because I think a lot of times because of budgets and limited data and all that, SMBs are kind of being left out of the technological wave. So if there's an off the shelf software solution that helps them do finance and accounting better, that's pretty exciting.
Guest 1: Kevin (41:17):
Yep. Thank you. Thank you guys for having us. We're really excited to be on here.
Host: Paul Barnhurst (41:22):
Thanks for listening to the Future Finance Show. And thanks to our sponsor, QFlow.ai. If you enjoyed this episode, please leave a rating and review on your podcast platform of choice, and may your robot overlords be with you.