How to close a $3.25M EBITDA gap to the board's target

Hi, everyone. Not sure if anyone's actually watching. I hope so. We're trying to be social influencers. Oh my God, social media influencers. That's how much a boomer I am. Guys, we're really excited to have this. Hope someone out there is checking us out. Eugene and I love, FP&A, love finance.

We've spent our whole career in the space. And what's been really fascinating in the last, I want to say one to two years, is the speed that we can move now. And you might be saying it's AI, but AI is actually part of it. But I think the tradition kind of experience of knowing what's good, what's right, decisions, design, things like that, is also a key aspect of it. Otherwise, you hear the term everyone's saying so much.

It's just AI slop. It's so easy to create a slop now, but it takes experience and real customer discussions to find the real solutions. And Eugene and I, we've spent, I don't know, our whole career seeing projects with clients thinking that solutions take six months, nine months. That's just not a reality anymore.

We truly believe a lot of solutions we can have an answer within a week. And that's why we have this MCC Finance Speedrun. I'll let Eugene kind of talk a little bit about it. And then we're excited to showcase what we have today. This was a topic that we hear often, which is about kind of what if scenario and then goal seeking towards certain KPS or metrics that you need to hit for your budget. But Eugene?

Yeah, absolutely. So a little bit about this, what this form is, is there are a lot of different challenges out there. And in the last 10, 11 years in the space is, we've seen all types of variations of it. But at the end of the day, it's really boils down to a few main key problems. And one thing that really excites me is that the same concept or same methodology can be applied to a lot of these problems. So that's why we have this session is that how we can take some of the learning in the last 10, 11 years, how can we convince them or instill them into problems that we've seen before and with the help of the latest set of technology and a little bit, even with agentic AI now, how can we really help the customer digest those problems and come up with a solution in a matter of days.

So this is what this session is truly about. And today we're going to dive into a little bit as Charlie mentioned, into a really specific problem where you have an expectations from the board, but your bottom up approach simply doesn't support it. So what do we do? How do we model this type of situation? And what are some of the tools we have at our disposal?

Absolutely. And as everyone sees, we're actually in our office today. We do have a professional recording studio, but it's our first time trying to make a little bit casual, trying to have a little bit fun with it. And we're going to do it live from our office. So everyone audience, like Eugene mentioned, like what we talked about, the whole goal is to address a specific solution or problem every week.

And not just talk about the theoreticals. I know there's a lot of kind of presentation, podcasts, writing about that. And I do think thought leadership is critical. But I also know at the end of the day, sometimes you just need to get some, get something done.

And we want to show you practical advice on how to do this. All this will be shared, shareable, accessible from our website as well. So please visit us after like, I think that's what all the folks are saying, like, comment, repost, subscribe. I don't know if the subscription link them, but subscribe.

But all this stuff will be available on our website for you as well.

So what is the problem that we're trying to solve? The problem here is you see this every year with your budget cycle. You start with the bottoms up planning because you want your budget owners to really think through what their department's changes look like. And do a bottoms up zero base approach where they explain why do I need to have this higher? Why do I have to have this project? Why do I have to have expenses?

And then you go through this and your EBITDA, your budget owners send you everything, compile it and it's your EBITDA and it's nine million. Well, the board, investors, depending on your owner, depending on the company.

Well, I've never seen the case with the first iteration of the budget is spot on. So the board wants us to push this up and we need to get us to about 12 million for our EBITDA. So how do we do this? Now historically, you basically send the budget back out, get everyone involved, have some meetings, lots and lots of business partner sessions.

You go top down, bottoms up, trying to meet in the middle approvals and that takes a long time. I will share a solution of what that looks like today. Now, it's a combination AI process, best practices, but collaboration, signing off accountability is still what's critical, especially in this day and age of AI work. So much stuff can be produced so quickly and AI is trying to make you think that is, I don't know about trying to think that makes it sound like it's conscience, but cautious, but who is ultimately signing off on the numbers, right?

So I think that governance and a sign up, it's more and ever in terms of how critical it is. But you didn't, any kind of past experiences in terms of what you see in this space and we'll have folks address this so that we can talk about what things look like in the past before we talk about what things can look like in the future.

Yeah. And you knew it perfectly because in the past, this was a lot of meetings through Zoom, through Teams, Slack, wherever it is that you communicate from. And the communication is one half the aspect. The other half is that you also be bombarded with what we call Excel hell, where someone sends you the latest budget, they call it budget version, version two.

And then latest, they'll say, oh crap, I feel I have an hour later. Essentially, we forgot something, now there's version 2.1. You scroll through your email, you find the latest version of it, and then you accidentally, you actually load it in the wrong version of budget. And then a month later, two weeks later, you realize, it's too late.

You're already committed to the board on a budget that was not the latest version. So definitely the human component is the communication aspect of it. But there's also the technical component of how do you get everyone on board on the same time using at least some of the same set of source of truth. So that's one of the things that we see quite a lot in our customers.

So we'll use an illustrative fictitious kind of example today with real templates, real data. And what we have is a company that has around 700 people, five entities, 10 department owners, really that gets you to kind of 50 different budget owners. So if you think about that, right, all those different entities, different business owners, different departments that requires tens, if not hundreds of different files over multiple round, it takes multiple weeks, you have Excel files going everywhere. Eugene just said it himself that Excel hell. It's everyone's favorite tool.

But also everyone's most terrified or hated tool, right? Everyone loves it. But in the collaborative process, it's it's chaotic, right? It says there's no security, there's no controls, file files are going everywhere. Yes, you might have a SharePoint, but then you have different versions, you're running around collecting different files.

And that's truly what it used to look like, and still looks like for a lot of companies, right? So and I don't want to hate on Excel, because I truly do love Excel

myself, I've spent, I don't know, 17 years in finance ops data now, and everything kind of ends up back in Excel. Even when I was doing my machine learning masters, where we were modeling things in Python, guess what, I was basically myself and as well as other data scientists, we were checking the numbers back in Excel, Eugene, you saw this when we were working with the larger client master that required machine learning.

Guess what we took the weights from the Python model, and we went back to Excel. Yes, we actually just had this actually just today, we built some elaborate dashboard for very important customer of ours, turns out the really elaborate fancy dashboards were not using that often, well, not using as much as we wish it was, because all the users were saying, hey, can I get this in Excel? That way, that is, even though all the dashboards, all the fancy colors and whatnot, customer that our customers want to do something else to it.

And that's just not something not as flexible when you have an interactive dashboard. So a lot of this still go back to Excel. Yeah, and ultimately, like there are a ton of different softwares that's trying to replace Excel. We've seen this over the years, we've worked with many of them in the past. I think there's a time and place for everything.

But I still love Excel. Because if you think about it, at the end of the day, if you're doing a fundraising, you're not giving access to an investor to a specific software, you're giving them to a financial model in Excel, Google Sheets. If you're doing, I don't know, if you're sending a bank covenants report, if you're into the close, a lot of kind of close processes, you're keeping records in Excel, right? It's just, Excel is fundamental to open finance and accounting, which is why we don't think the our hypothesis is you want to keep Excel in the loop, but you want to build the process, you want to build technology, you want to build SOPs, and naturally want to build AI into this whole package.

But I don't think you can replace Excel. So one of the things here is we're working with Excel, we talked about those 10s, these 100s, these files looking back and forth.

I'll talk about the next iteration of how we can improve this. So we love a tool called Vena. They are an Excel first tool. Given the database, we're allowed to manage all the semantic layers that has workflow auditability. It does a lot and makes that process a lot easier. But what we think the magic sauce now is we combine Excel, we combine a platform like Vena, and then we leverage AI. Now that AI I think was a bit more, what's the word, a bit more flexible. You can use your AI of choice.

You can use Claude, you can use GPT, you can use co -pilot, you can use Claude. I don't think it really matters. Heck, you can even have your own server run your own local model if you want. But that's our vision in terms of what it looks like. This is what it used to be. This is like, so how we were talking about how do we give it control? How do we make a structure scalable?

And we'll showcase you a little bit what that looks like. And I'll flip to the Vena platform. And this is not really a pitch for Vena. It's more to talk about what are the options. There are other platforms similar to this. You can get similar results. There are pros and cons between different platforms.

It might be a different time and place for each of them. But ultimately what it is, is we want to have the semantic layer to control the knowledge, definitions, all based in those one place. And it's collaborative. Meaning if you're accessing through a report template, they always have the same definitions. And what I'll do is I'll grab one of our hiring templates.

You can see my dog teaching in the back. And I will take one of these files. And again, this is where I was going back to that love Excel I was talking about.

The idea is I truly believe accounting and finance, you want to stay with Excel. But at the same time, you don't want to use Excel as a database. That's where so many of us fail. We all use Excel as a database. You take your raw data, you export it into data tabs. You share the files with other folks. Now we have local versions like Excel, local databases for different reports and templates. And the worst off is Excel is a terrible database.

Over the years has been interesting. Microsoft added data queries, power queries into Excel to have some database-like functionality. But ultimately it's not a database. So what better it does here is it gives the ability to kind of work with that template. But then we can select and reuse this template over and over again for different departments, different entities, right? And this is what I was mentioning earlier.

In that scenario, the fictitious scenario is talking, but we can have hundreds of different templates. Right now I can go to operations. So I can slice and dice. So in the normal Excel world, this will be one Excel file. And then if I switch to a different, let's go to commercial, that would be a different. Notice how it shrinks, meaning here, it controls the data, the fact data, the financial data, any operation data, but also the structural data as well.

And ultimately it also gives security as well. So if you're not able to see any data, you're not going to be able to access it. That's where you kind of go from Excel to the next level where scalable is structured. You have that collaborative semantic layer definitions, calculations, and you're not redefining these templates, copy them over and over again.

You allow the ability to kind of use the same template, but we just toggle a switch. In this case, I'm going to go to all departments a little bit easier to kind of go take a look at all the data and go through that planning process. Now I have a template here.

This is a fictitious kind of HVAC plumbing company that we created. We do actually work with a lot of companies in this space, a lot of people of HVACs. And we have kind of direct technicians. We have office hires as well. And you can see here, and again, team or folks, I'm not going through a full kind of demonstration and everything. There's different other templates for drivers, start date, hires, trend transfers, things like that. It's all about how do we make the reasons to actually get to the budget that we're looking for. So if you remember earlier, we were talking about how we need to go from the 9 million to 12 million. So this is a big target, big switch that we have to do.

Now, in this next phase that we're talking about where we're using a platform like Vanna to kind of scale your Excel template. Well, it absolutely makes it easier. You're sitting in one template, you're going between different kind of departments, different entities.

And it does make it traceable. Any sort of changes that we have here, we can follow it. We have audit trail. We can see who made what change. We can sign off on it. In fact, one of the big pieces that we're doing here is to have that approval process. So you make the proposals, someone changes it, someone signs off it.

And that way you can have the full control in it. But that can still take weeks, if not months, depending on the size of the company. This is where we want to highlight that new world where we can combine Excel, our favorite UI UX, a platform that allows us to scale, like governance control, auditability, repeatability, semantic definitions, but then let's add intelligence, artificial intelligence into this.

And what we've done is we've connected this into Claude. Now I'm using Claude, it could be GPT, it could be co -pilot. Essentially, we have a MCP into Vanna that has the data brought in. As you can see here, I can actually have it work with me, understands what the employee asks, existing pays, the different scenarios we have.

It even gives me questions, right? What are the different scenarios? What are the different questions that I should actually walk through with the business partner? And the idea is they can work with us to create, to understand what are the projects that are important? What are the entities? And again, this is the one I highlighted, you didn't talk about this earlier.

This is the initial kind of analysis. This is not the final sign off. Ultimately, we are still importing the decision -making, that accountability, that making that decision, that digest this intelligence that's given to me.

So what I've done here was with the Claude platform, it's created these decisions, it's worked with me. I'm demoing kind of just a few pages, but I saw there actually for about a couple of hours kind of discussing with other best practices, where can we adjust the pay with kind of minimal impact on the company. And that's where you can start having different scenarios.

So one of the things I want to highlight is if I go back to the reports, and this is another reason where a platform kind of helps is if you are doing this locally with Excel, you can actually inject some intelligence. But now you're just getting AI to create even more Excel files. And everyone knows here, the speed of creation is faster than ever now.

So you can see here, this is the original bottoms budget, bottoms up budget that we talked about was about 9 million. You can see here it come down here. So our top line was about 110 million. This business, we ended up about 9 million EBITDA. And what you can see here is with the help.

So you can do this manually, creating different scenarios different. But what we've done is we had AI kind of pre -seed different scenarios top down, different adjustments, things like that. And I think that's the scenario we want to look at. I think it's this one actually.

We're working with that AI advisor going back and forth, and you can kind of create brand new scenarios, not different files. And again, we're talking about hundreds of files, we're talking about all the different budget owners, things like that. How do we take these exact same files, but then we add a new slice of data, a new version of that scenario, get to 12 million.

But the thing here is you can't sign off on it. We do actually have to go back to the templates, review the reasoning, right, talk about capacity, different options. Some of this is AI driven, some of it's your own commentary.

As you go through the decisions, are we going to approve this? Are we going to reject this? Are we going to defer this? This is some hiring. You can also have strategic projects as well. So I'll come back here. I have other templates. We can do this for any sort of expenses, budget, revenue, debt, and the idea of planning, anything that is skill -based.

And even I think just Charlie mentioned this, as you go through different scenarios, actually use AI to explain why some things have changed as well too, what the difference is between two of these scenarios. And that's exactly right. So here at the template, we're actually looking at strategic projects. And naturally, these have dependent schedules that talk about the investments needed, the headcount needed, the ROI, so that you can create an absolute case about why you have these projects.

But ultimately, we have to hit that budget, that target of $12 million. We've got to work backwards. So the AI has created this scenario for me. I have different explanations. Now, this is really useful because historically, you have to sit there and torture the data. You have to look at vendors, you have to look at cost, you have to look at customers, you have to look at operational data, desk schedules, et cetera, et cetera. This is where AI can start going through that.

Naturally, you have to test and check the numbers. We actually did another session with the Chartered Institute of Management Accountants a couple of weeks ago, where we shared kind of best strategies to test the AI figures. Different sessions at different times. But the idea is to analyze the explanations, confirm. Again, make those decisions on it, sign off on it.

This way, you understand what are the decisions that we are making, what's approving, what's not approving yet. And then once you toggle these, a platform like Vena, these are the calculations are now mechanical, deterministic. What happens is once we have those decisions kind of creative, you have audit trail around them. You can kind of see who made what changes.

But ultimately, when I go back to that report, I could start with the scenario that the AI has created for us. But ultimately, we can go to the finalized kind of decisions that we had as well. So after the executive signing offs decision maker, we can kind of see what it looks like as well. Oh, it looks like I didn't finalize the decisions making there. But the idea is once we have to sign off, we have the approvals we have to save.

We can then take the AI suggestions, decisions we made, which projects which hiring are approved, align that into calculation, and then produce a final budget. And one thing to sort of add here is that as you go through this approval process, definitely we want the human component to come in and basically bring back everything. But one thing that will find it very helpful is to actually ask AI to

validate itself as well, too. Can you green light yourself? More than often not, we have had situations where AI, as Charlie mentioned, just gave us a number that was incorrect. But AI is also capable of catching that error themselves. You might not be able to get that on the first prompt, but if you ask it, just take an extra 30 seconds out of your day to say, AI generated results, can you run validation on it? More often not, it will actually find its own nuances or own edge cases that is not seen, not hitting correctly. One example is that we actually have a customer where they are applying for properties and the AI created revenue for that property, but then taking the consideration that there's actually a field where the property is closing for the rest of the year, but because of the AI logic, they generate revenue, but it was able to catch it after second prompt, after second iteration.

So ask AI to check its own work. Yeah. And for all the folks that's a little bit newer to this was creating prompts, skills, routines, agents now, right? One thing to think about is like, how do you actually QA things yourself? Take that knowledge, document it. You can create a QA agent that can help you do your work as well. So every time you do these work, it helps you check.

Now again, AI is non-deterministic. You can't trust it a hundred percent, but it's like having a really strong analyst check your work before you do a final check of it, right? And this is kind of where we're going from that initial budget towards the end. We have a nice kind of waterfall diagram, but I want to highlight all this is really easy with a platform like that, because you have all the data, but it's even easier with Vena plus AI, because you have all the data sitting with the database, which you can tap into with your AI of choice.

So Claude, GPT, Copilot, Grok, et cetera. You can tunnel in to the MCP and you can read the data, you can analyze it and create a scenario to adjustments for you. And then you can ultimately make that decision, right? So for everyone that's kind of watching to summarize today's session, you have

the initial budget process, all the way that you do today, where this bottoms up to your base, top down, et cetera. So often it makes it what that looks like. That gives you a starting point. Never have the AI do that initial budget. You still need to have people involved in terms of how they think about next year, the nuances for the project. Because remember, all this is driven off data, but there's a lot of tribal knowledge that's sits inside each of the budget owners teams that the AI is not going to have context for.

You build up the budget, you have that budget within a platform so that AI can get that data in one accuracy, it's accurate, you can see it will make the changes. It has a contextual layer about the financials, about the company, the definitions, the calendar year, I don't know, the org structure.

Have AI review that data, suggest different drafts, different options. And this is where you can go back to the budget owners, revise the lines. And again, so different cell files or different artifacts, we can do this in a controlled fashion. And then with that approval process, the execs can then essentially settle down on decisions.

And once you have the decisions, who's going to be approved, what's going to be approved, what's going to be rejected, what's going to be deferred. You now have the moving pieces or the components for the platform to deterministically calculate what that final budget looks like, along as the speed of the final budget so that you can get faster CFO sign off and board approved.

And hopefully, our wish for kind of the Office of Finance, go home a little bit earlier, go home on the weekends, and finish your budget cycle a little bit earlier. And all things related to this is applicable for your forecast as well as for the companies that's only budgeting once a year.

This is going to make it that much easier to start doing a rolling forecast so that you can be more agile in this crazy, crazy world of ours. That's kind of my bit. Eugene, I'll let you kind of close it off.

Yeah, absolutely. So thank you, Charlie, for walking us through. As you can see, here's one example of what we can do with combination of your raw data, whether that's coming up from your ERP, your subledgers, your revenue data, your Salesforce pipeline, a lot of different data that you can bring in to a single unified platform.

Then that becomes your semantic layer, or what we call is just basically a layer of context with clean data parsed out the way that you need it, the right dimensionality. So now, AI can then put the layer on top to basically give you an extra layer that you can understand the data, have it have conversation with it, and even make some judgments on top of it to make changes to your forecast or goal seek to where you want to go.

And then finally, having the human layer component of it. So this is one example of what we can do. And for the next couple of weeks, or as long as we want to run this, is that we also start showing these in a lot of different examples. One of the things that we're going to start hitting in maybe the next couple of weeks, or sometime early next month, is how can we do this in month-end close?

For reconciliation or consolidation, we use the same methodology and same concepts for different use cases. So stay tuned for a lot more of those. We're really excited to showcase what some of the stuff that we're cooking up and what the data is offering that we have on our side and how we can speed run your problems.

Yeah, everyone, we're cooking. Get excited. Like I said earlier, throw your hardest problems at us. We'll guarantee we'll solve it within a single week. And all this will be shared with the you as well, any sort of data sets so that you can try this yourself. I did this on Mac. It doesn't matter.

You can do this on Windows as well. Join us, work with us, throw your hardest problems with us. And proof is in the pudding, right? Let us show you what it looks like. And we're happy to kind of deploy some of these solutions for you for free if we select you as well. Also, a lot of you kind of watching me, we have some sick merch, right? Happy to kind of share this with us as well. with you as well. Thank you so much.

Ciao. Bye.