.png)
.png)
Fireside Chat
Cube's AI for FP&A: Context, accuracy, & control
Discover how Cube’s AI features are delivering real value today—and what’s next for finance teams.
Cube
☰
Product
Solutions
resources
Company
Details
AI promises to transform finance—but only when it’s built with context, accuracy, and control.
That’s exactly how we’re building products at Cube. Our AI features are purpose-built for the demands of FP&A—responsible, dependable, and aligned with how modern finance teams operate.
Tune in for a fireside chat with Suresh Bala, Cube’s Chief Product & Technology Officer. You’ll learn how our AI features are already making an impact—and what the future holds.
In this 30-minute session, you’ll:
- Learn how Cube is approaching AI differently
- See a deep dive into Smart Variance and some of Cube's other AI features
- Get a sneak peek at Cube's new conversational AI capabilities
- Receive information on how to join the conversational AI beta program
Speakers
Video Transcription
Well, thank you everybody for joining today's webinar on AI. We're really excited to get kicked off here and start this fireside chat.
Just a couple of quick housekeeping items to start. We are going to have time reserved at the end of the fireside chat for a q and a session.
So throughout the session today, please feel free to submit questions using the q and a button within your Zoom toolbar.
And after today's webinar, you're also all going to receive the recording of this session and the q and a in your inbox. And for our cube customers, you're also gonna start receiving some beta links for new AI features, and we'll be sure to share our other resources as well.
With that, I want to introduce our main speaker here today, Suresh Bala.
Suresh is the chief product and technology officer here at Qube. He's located in the Bay Area and came over to Qube with years of experience in the FP and A and product space.
He brought all of that knowledge over here to Cube about a year ago and has been instrumental in really developing our product and enhancing Cube to be the most robust powerful solution that it can be. So I'm really excited to chat with him today and learn all about AI and where we're going with that in Cube.
And my name is Taylor Josephs. I'm a senior product marketing manager here at Cube. I was previously a solutions architect, so I may have given demos to some of you on the call today.
But I'm really looking forward to kicking off this conversation here and diving deep into AI with Suresh.
So without further ado, Suresh, I'm gonna ask you first and foremost, we all have really been hearing a lot about AI, not just in the FP and A space, but across every industry.
So I'm really curious to hear from you. What is Qube's general philosophy on AI, and what makes AI really different in Qube?
Yeah. First of all, thanks, Taylor. Really excited to be here.
And with regards to your question about AI, you're you're spot on. Right? If you just step back and look at what's going on across the board, right, like all of us, we can see that there's an AI revolution that's going on already. Right?
It's pretty remarkable. All of us who have played around with ChargeGPT or Claude or any of these, you know, AIs and large language models, it it's almost impossible not to be impressed with what's going on, what the potential for the technology is. And also a little concern. Right?
Like, there's always this I think pretty much all of us could also see that there are concerns about about that, the technology itself, what it can do, and potentially what the repercussions of the technology are.
So, but the reality is it is going to transform the way all of us works. It's going to transform the way finance and finance teams work.
And at Qube, our approach is really how do we leverage this technology, this this evolution of this technology, which is moving to such a fast pace, but doing it in a way that helps our users, our finance users, our business users kind of maximize their ability to be productive, maximize their ability to be effective in their roles while mitigating the risks. K. So that's gonna be our approach into terms of how we layer in the AI technology within Q. And in particular, the way we are thinking about it is really along three guiding principles. Okay? So the first guiding principle is around context.
What we have on behalf of our customers is really their historical data, their financial models, their scenarios, their what if analysis. There's a lot of rich contextual data that we have for each of our customers that we are stewards for.
And combining that with this AI technology, there is really a potential to fully unlock some some meaningful, some compelling insights.
It helps organizations and our users understand the past better in terms of their business, forecast the future better, and make better business decisions. So there's a lot of potential there. And this context of layering in the the data that the customer has, their their sanitized data that they have with this technology, there's a lot of potential to do that. And that's one aspect, one guiding principle that we're looking at.
The second thing is around accuracy. At the end of the day, we are talking about financial data. We're talking about financial metrics, business metrics, and there is no room. There's no scope for ambiguity or incorrectness or imprecision.
Right? So we have to make sure that the data has to be precise. It has to be, correct, and it has to be explainable. It has to be transparent.
So there should be clear transparency for the user about how the AI systems came up, analyze the information, and presented the data in a way that they can also then, as a user, go back and understand it better and dive into the details and validate that it picked the right things. So that's the second part, accuracy.
And then the third part is control. This goes to the the risk mitigation part of it. Right? At the end of the day, we take our job very seriously that we are stewards of our customer's sensitive financial data, and we have a really high bar in terms of security and control around it.
And that is not that is the same level of rigor is gonna be applied for our finance for this AI systems as well. As we're layering these AI systems into our platform, the same level of rigor will continue to be applied. So that's that's one aspect. And then the second aspect of control is really that we want to make sure that our customers, our users, the finance, the finance teams, and the business teams that use our software, they have complete control about who gets what access to what kinds of data.
So you can very granularly control what data, what information, what capabilities are being exposed to what user. And everything, including this new set of AI technologies, will all honor that. So we want to make sure that there is perfect control that the finance teams have about, about the data and then who gets access to it. So those are the three guiding principles.
Right? As as we at Qube think about how to leverage this this fast moving AI technology, to help our users and our customers use our software better.
The last part I will say is the other aspect of this approach is what we want to do is we want to really layer in these AI technologies, and and infuse it within our platform. And what I mean by that is as users, as finance users, as business users, as you're interacting with Qube, our goal is that we want to provide this artificial intelligence and machine learning capabilities at your fingertips while while you're working at it, while you're actually interacting with it, while you're examining data, while you're understanding what to do next, that is actually at your fingertips. So it isn't gonna be a bolt on that you have to go separately or do some other perf perform some other actions somewhere. The goal is really, can we do it in the context of what whatever job, whatever function that you're that you're working on? And that way, it makes it more natural, more easy to use, and it just makes this entire thing basically work at your service effectively.
Awesome. That was super helpful to hear from you, Suresh. It sounds like overall, the pillars of context, accuracy, and control are what really is setting you apart in a world that seemingly wants to move fast and break things in regards to AI.
So given that philosophy, how are we really setting up our customers for success in adopting AI?
Yeah. So in fact, on that, we're already on our way. Right? So the first thing we did actually, and this was, you know, more than a year back at this stage, was really our focus around this notion around applying artificial intelligence and machine learning actually more in particular in that situation to do what we call smart forecasting. The idea really is our customers have their historical data in cube.
Can we have this machine learning and artificial intelligence technologies understand the historical data, use that to project ahead and to the extent that the past informs the future, be able to project ahead and show you what the your business potentially might look like, based on historical so that was smart forecasting. So that was something that we rolled out a few months back, almost a year back. The next thing we did, which I was really interested about, and I'm sure we'll talk more about this, was around smart variance analysis, which is really something that all finance teams do month in and month out, quarter in and quarter out, which is really to compare what did we say we are going to do versus what actually happened, what's the delta, and how can I, explain the root causes of of the variances?
And this is something that we applied the AI ML. So that was a second second, kind of, for a hint of this, if you will. The third thing that we are actively rolling out, which I'm really excited about, is this notion around ad hoc analyst analytics agent. So the idea really is almost think of it like as as as a junior financial analyst who's at your fingertips, but it's really it's a software.
And you can interact with it, ask questions, and get responses, and and chat with it. So there's a lot of potential there, and we can talk more about that as well. So that's kind of the journey. So we have had a lot of things that we're already putting in place.
We're continuing to build this out.
And as we build this out, the goal is really how do we help our users, our finance users and our business users, automate what can be automated. So that's one aspect of it. Second part of it is provide more insights about their past, help them forecast the future more confidently. And then the last part is how do we help, finance teams and business leaders make decisions more confidently, based on these technologies and based on the underlying data. So have to make data driven decision making a core part of it, and there's a lot of things that these technologies, can do to support that.
Excellent.
So I heard you mentioned a lot of different applications of AI within Qube, and I definitely want to dig into all of these today. But the first piece that I really wanted to focus in on was that topic of using AI to start gaining insights, particularly around variance analysis.
Now I know we recently launched smart analysis in beta in Qube, and I've been playing around with it a little bit and think this would be a really good time to show this off to the group here today. I'm gonna go ahead and share my screen here.
So right now, I'm in my Q web platform, and I'm in this smart analysis section of Q. And what I wanted to do today was really dive deep into the variances between my actuals and my budget.
Now between those two scenarios, I wanted to focus the scope of this analysis a little bit further. So I tacked on some filters, like just focusing in on quarter one of this year. I also only wanted to investigate variances related to my US entity.
And finally, I only really care about variances that are going over a thousand dollars, so I applied that threshold as well.
So from that point, I clicked analyze here within Qube, and you can see these results are starting to render. So Qube is leveraging that AI functionality.
And now Qube has generated for me my largest variance by amount here at the top, and it's given me really quick insights into what's driving that core variance. We can see it's being driven from platform revenue, mainly in the month of March.
I'm also getting these really nice charts too that are outlining my top five variances, and I've hovered into some of these too to understand a bit more about the makeup of that data.
But the piece that provides the most context for me here down below is this summary portion.
I can see that this AI smart analysis has spun up a quick summary telling me that my quarter one financial performance significantly exceeded my budget. We can see that we had really strong revenue outperformance that sort of surpassed some of the expenses that we might have incurred in that quarter.
And it even took a stab at finding the primary root cause behind these variances, showing that revenue outperformance of two point seven million flowed through to the bottom line with opex increasing only moderately.
And beneath that I'm also seeing some other variance highlights too, but beyond just getting this summary glimpse, I also wanted to dig a bit further into what was making up these numbers from the bottoms up. So what I did next was I toggled over to this report section and this outlined in a lot of detail for me sort of what was driving those top variances.
I first made sure to sort this, so we can flip this ascending or descending. And what I really cared about in this case was digging into revenue. So from this revenue account, I clicked into this, and this actually brought together a drill down for me, outlining the four core GL accounts that made up that total revenue variance. We can see their amounts highlighted here on the right.
Since platform revenue was my largest variance, I wanted to explore this a bit further by digging into that quarter. This is outlining that March had the highest variance of that quarter. So from there, I dug a bit into products, finding that my enterprise product generated about five hundred and sixty thousand dollars worth of that variance and even going all the way down to markets. And I'm getting this really key highlight here of what's driving those variances.
So I found this to be super valuable, but I would love to hear from you, Suresh, kind of your thoughts around how this could add value to our customers.
Yeah. Yeah. But this is a really good example of of combining the best of the data that's available and these AI and machine learning technologies to to really develop insights. Right?
Like, I I really like this because there is something that, you know, every finance team does. Right? You do variance analysis all the time. And to the extent that the system can take a first stab at really developing, insights into it and then allowing the the the user to to review it.
But then she'll also be able to examine further. Just like you showed here, we could then examine it further, dive into the details, understand that, get get get a little bit more appreciation of what the data is that's supporting some of the inferences that AI came up with. That's that's that I think is is the best combination. But, like, that's what makes sure that as a finance user, as a business user, you're getting confidence about the the analysis that's going on, and you are feeling more and more confident about, the root causes that are being identified.
So so this is a really good example of of where we could go with these kinds of technologies.
And I'm really I'm really excited to see the adoption around this. Would love to hear feedback from our customers and users as they use this more and more.
Yeah. This is this is really, really cool.
That's great. Thanks, Suresh. And, yeah, speaking of feedback, I want all of our customers on the call today to know that what I just showed here is actually available in beta in all of your cubes right now. So you don't have to install anything. You don't have to add extra applications or programs, but we can share out a resource to you after today's webinar so that you can start trying this out for yourselves. And, again, we would love to receive any feedback from you as you're starting to test out this feature.
Awesome.
Well, another use case that I wanted to dig into is beyond just leveraging AI in Qube to start analyzing our existing datasets, I wanna understand a bit more from you, Suresh, about how we can use AI to start generating new data, like maybe forecasts or plans.
Yeah. No. That that's a great point. But this is this goes to, again, the use case based, thinking that we are applying. That another classic example of what finance teams do oftentimes is on a monthly basis, on a quarterly basis, they forecast the rest of the year, maybe, or they forecast, you know, sliding window of the next eighteen months or whatever the case might be.
And the idea is, can we and this is what I was describing earlier. Can we have the system? Can we have machine learning examine the historical data, identify things like growth and seasonality and outliers and, like, try and learn how to ignore outliers, and use that as a way to project ahead. Right?
Like, you can almost imagine that this is a way by which the software comes up with or the system comes up with, hey. If the history is any indication, this is where you will end up end of this year, end of next year, whatever the the time horizon might be. And then this enables finance teams and and business teams to be able to say, okay. This is what the system is suggesting.
But, of course, the system does not may not know some of the actions that the business might be considering to do, you know.
So there might be other business judgment layers that might be layered on top of it. But it becomes a good way to do a sanity check, right, to do actually to be able to review it. It could also become a good way to have a starting point. Like, this could be the starting point based on historical data, and then you can layer on top of it other intelligence that you have about the kinds of actions that that the business is planning to take. So to me, yeah, there's a lot of potential there, of letting of letting the system, do some of the heavy lifting, apply these AI ML technologies, combine that with the data like we talked about earlier, and then use that as a starting point to to both analyze and, in this example, forecast ahead. And then then then be able to then, let the business users apply and the finance users apply the the strategy and the, and the decision making on top of that.
Excellent. That's super exciting.
You know, something that I've been hearing a ton of, and I'm sure everyone on this call has been hearing this too, is the concept of conversational or plain language AI. So, Suresh, I'm curious to hear from you how the product team is thinking about this in Qube and how conversational AI might start developing in the future.
Yeah. Yeah. No. This is definitely something that's moving fast and but seems to have a lot of potential.
So we are pretty excited about the potential and the possibilities around this. Some of you might have heard of this term called agent e k I. That's becoming a new thing that's coming into the market. And the idea really is, we all are now getting more and more comfortable and familiar with Chad GPT and Claude and all these systems which do things called LLMs or large language models.
The next generation of that is can you then use that to actually perform actions on your behalf based on natural language instructions that you provide to these engines?
So so there's a lot of potential there, in terms of how we could do it in our context. So we are very excited about how we could help our business users, our finance users, kind of interact with these agents, give it give it instructions, then have it do the actual heavy lifting and come back with the results.
So there's definitely a lot of a lot of potential there that that, that we could do. And you and I know about this. We recently are just rolling out. It's still on beta mode, but we are really rolling out our first notion of our AI agents.
It's gonna be really focused around ad hoc analysis. So the idea really is that you can ask it a question, get get responses, and interact with it in natural language. But it's doing the same thing that we talked about earlier. It's combining the best of the artificial intelligence technologies with the data and the context of the specific business, of the specific, company and and and then combining that tool to provide meaningful insights.
That's amazing.
Yeah. And I know we do currently have that pilot program running for, you know, that conversational AI in Slack and in Teams. I'm a Slack user myself, so I've been trying this out, and I wanted to share this with the group as well today. I think that this feature really speaks well to those pillars of context, accuracy, and control because this is really a clear picture into my numbers and into, you know, how this AI chatbot has answered my questions.
To walk you through what I asked, about an hour ago, I asked what was my revenue per employee in q one of twenty twenty five? Now this chatbot only took a couple of seconds to respond to me, and it gave me a really descriptive answer.
It first let me know that it's gonna help me calculate that revenue by retrieving both my revenue data for quarter one and my employee headcount information.
Now it first needed to find out the number of employees, and it was able to calculate this here down below, and it showed me all of the math and the insights that it applied to get there. It showed me that my total revenue for q one was about twenty one million dollars. My employee head count was ninety four doll ninety four, and it also gave me an option to calculate this based off of my FTE count, which was ninety three point five.
Now from there, it showed all of the math explaining that the revenue per employee was that twenty one million divided by ninety four, giving me about two hundred and twenty six thousand dollars per employee. And it also calculated that based off of FTE count, which was roughly two twenty seven.
From there, it gave me sort of insights into what that metric means. So it's showing me that, you know, per employee, we're getting this really high revenue amount, which is primarily driven by our platform revenue with additional contributions from support and implementation revenue.
As I start scrolling down, it even got more specific in a cube context telling me the dimensions that it used to get this number. It was filtered on my actuals, my global company, total departments, and markets, and it even spun up some data that was available for me to download. So what I was thinking here is that maybe I could download this headcount component or this revenue component, bring that into Excel. I could even start drilling into the details using cube from there. But Suresh, would love to hear from you sort of any other thoughts or insights you have on how this could bring value.
Yeah. No. This was such a this is such a good example. Right? So the one aspect of it is you started off in the natural language conversation.
Right? You just asked it a question. You can almost imagine the situation is there is a meeting going on and a senior exec asks the finance person or finance leader in the in the room, hey. You know, what is our revenue per employee?
What is it this year? What is it last year? And has it improved or not?
You don't need to now have it be predefined anymore. Right? Like, the beauty of this is it's it's smart enough to first understand the intent. It's then smart enough to interrogate cube data to actually come up with the results. And like you like you described, then it's also able to explain to you how it arrived at these numbers. So you, as a finance user or as a business user, you have confidence that this is following some logic that you can understand, that you are comfortable with, and, hence, you are confident about sharing this with your stakeholders, with your audiences.
So there's a lot of power in bringing the best of, AIs and these notions around LLMs, which is the natural language part, with the robustness of Qube, its data, its structure, its ability to perform these kinds of analysis. And bringing the two together, there's just so much of, so much of potential here, which is pretty exciting.
The other thing to think about this is that you just open Slack and you just started to type on this, which is which is pretty awesome. Right? Like, the the this also meets the other goal that we have that we want to be able to meet our users where they are. So if finance if your finance teams and your business teams live in Slack, then, hey, you know, you could just interact with this in Slack. Similarly, we're also doing it for Teams, so that could become a third second part.
The goal is really to expand the ability to interact with us in this kind of a national language wherever you are, wherever you as a user are, whether you are on Slack or Teams or you're on our browser, that web based presence. Or if you're on your spreadsheet, especially if you're on your Google or or Google Sheet or Excel. The idea here is, can you interact with it, ask it these kinds of interactive dynamic, just natural language fluid questions, and be able to get the results back. The other thing that you pointed out, which I'm I'm really happy about as well, is that it also provides you with the supporting data.
So the table below that you showed here, it's really cool because it also tells you, hey. Here's the data that I got from Qube to support the core Qube engine to support the analysis that I just did. So that also builds just more confidence and more understanding about, okay, this is how this AI agent actually tell you what the result that I was asking for. And, hence, I can be more confident about sharing this with with my stakeholders.
So really yeah. So really excited about the potential here. Maybe the last thing I will say about this is, this is just the beginning. Right? I can imagine that this one we can we can and then we are actively working on this. So this is the notion about ad hoc analysis. So how do you actually just ask questions and get responses back?
The next part of it with something that I alluded to earlier as well is this notion of actually performing actions. You could actually also give it instructions about to perform some actions. So we're really looking we're really looking to invest in strengthening that kind of muscle where it can actually also perform actions. So things like change my models, spin up a new scenario, you know, map some data from a source system, a ERP system into cube and be able to do all of that in a in a in a way where you can describe the intent as a user and have this agent run and do the actual work for you. That does that make sense?
Absolutely. Yeah. Really, really exciting things on the way. And I just wanted to remind everyone as well that this conversational AI chatbot is currently in beta, and we have a pilot program running as well. So if you're interested in that, we'll send out links after today's webinar where our customers can sign up and try this out for themselves.
Awesome.
Well, Suresh, I know you've kind of touched on this throughout this session, but I'll just ask it again to sort of wrap things up today with our fireside chat.
Where is Qube really going next when it comes to AI? You know, we've talked about the features that we've released that are in beta, but I'm really curious to know what else is on the road map.
Yeah. Yeah. No. Like like like I was saying earlier as well, there is a there is just so much of potential here.
Our our general thought and this is something we would love to partner with our customers on because this is definitely something that's fast moving. It's evolving quickly.
But a general thought is what we can do with the with the core cube model, the richness and the sophistication of a cube model, the data, the the financial model that sits on top of the data is really build, you can think of it like a junior FP and A analyst. A junior analyst that can perform the task. It still needs supervision, so it still needs our the the the folks on this call. Right?
Whether you're finance or business users, you're still providing the guidance. You're still providing that, the targeting. You're probably also still reviewing the results. But this can do a lot of the heavy lifting.
This can do a lot of the analysis. This can actually sift through a lot of data to come up with some of the insights that that the business would actually need.
And and it can really perform this notion of a junior financial analyst that can actually do a lot of this work. So this notion around agentic AI combined with the context of, combined the context of the data that I talked about earlier, there's just a lot of potential for for this to to take on. The use cases that we are exploring are things like how do we make data management, easier. Right?
Like, can we ask these agents to actually, you know, get the data in a shape that it can be used, effectively for analysis? Can it actually perform the analysis? Can it come up with insights as it looks at historical data? Can it help you predict the future?
So we talked about that a little bit. How can it help you get to generate first better forecast so you can actually look ahead?
And then the last part that we're also exploring is can it also help, do what I consider as prescriptive analytics? And what I mean by that is can it actually prescribes given a set of objective functions, if you will, like the goals that you want to, have, can it actually come up with here are three options that you have, And then you can decide which option to pick and and why and so on.
So there's a lot of potential here about leveraging the best of this fast moving technology, combined with the robustness of how Qube manages and organizes the data and, the models on top of it, to make a meaningful impact, to make our finance users, our customers, our finance users, our business users be that much more productive. Right? Like, so this this thing can do a lot of the the grunt work, if you will, that frees up our users to actually become more strategic, to become the key advisers to their business partners, to to help drive more data driven, and more, kind of insightful decisions for the business based on the, the underlying kind of work that that these agents can can do and do value, value.
So that it's really exciting. It's exciting what what can be done. I'm looking forward to it. And then the last part I would ask for, as a parting thought, and, Taylor, you briefly mentioned this as well, is, we look at this as a journey.
We look at this as a partnership with our customers, with our users. We would love to partner with you on this. So there is just so much interesting things that we could go do, and we would love to partner with you. We would love to brainstorm with you.
We would love to develop these, get feedback, iterate quickly, learn quickly, so that we are building tools that are that are making you more effective, more efficient, better at your job, and hopefully freeing you up to do more strategic work and having this do all of the the legwork, if you will, to support that kind of strategic financial analysis that that you'll want to do.
Awesome. Thank you so much, Suresh. I feel like I learned so much from you today, and it was really awesome to hear all of your insights.
I'm sure we have tons of questions in the q and a, so we'll definitely leave some time at the end of this session to get to those. But before diving into that, I just wanted to share these resources again that we plan on sending out after today's webinar.
First resource is for our existing customers. We are going to share out a link for you to sign up for that pilot program so that you can start playing around with that conversational AI chatbot, the one that I just showed earlier.
Secondly, we'll also be sending out a link for you all to access the smart analysis feature within your existing cubes. So So, again, this isn't a program that you have to sign up for. Super easy access link. You're gonna be able to start playing around with your existing data. And, again, please share any and all feedback that you have with us.
And lastly is for any prospective customers or folks who are still checking out Qube. We will also be sharing out a link for you to book a custom demo with us so that you can start exploring other topics, whether it's AI related or more broad FP and A topics with us and learning more about how Qube could fit into your business.
So with all of that being said, I'd love to move into the q and a session now, and we can start answering any questions that folks have had.
So, Alyssa, I'll have you start reading some out.
Thank you, Taylor. Hi, everyone. My name is Alyssa. I'm on the marketing team here, and I'm gonna help out with the q and a today.
So first, thanks so much, Taylor and Suresh, for such an insightful conversation. This has been really exciting. We've had quite a few questions come in. So I'll throw them your way, and you can decide which one you would like to answer.
So the first one we've got, for the smart analysis, can you do the same for current month versus prior month or prior year?
Hey. It's a it's a great question, and this is something that we want to do next. So right now, it is doing it's comparing versions of of your plans, but you could compare actual versus plan.
But it's a really good question. Like, the next iteration of this will really be to do period over period comparisons and kind of the same kinds of insights.
Yes. So definitely on the on the on the road map. So great question. Yep. Makes perfect sense.
Nice.
Alright. Next question. Can smart analysis do a detailed variance analysis with commentary based on a full trial balance or detailed p and l?
Yes. It can. So so I don't know if you if you saw that when, Taylor was showing. You can decide you can define the context that you wanted to analyze. And you can say, hey. It's gonna be my entire p and l, or it could be my entire balance sheet or cash flows or any specific aspect. Could you also be much more precise?
And for each of that, as long as you provide the context and you provide the two scenarios in this case, right, it could be actuals versus plan or actuals versus forecast or what have you.
Yeah. That's all it is. Right? And what it's gonna do is it's really gonna make two passes at it. There's gonna be not to get too technical on you all, but, there is gonna be a machine learning pass where it's actually gonna do a comparison and try to establish what are the actual true variances and which is the highest contributor. And then the second part is then it applies the artificial intelligence on top of it to actually kind of develop insights and then communicate those insights in a way that's easily readable and and, explainable.
And then the third part is the one that Taylor showed you apart, which is also very powerful is you can then explore further. You can then say, okay. Now that I see it, I want to understand why did you end up saying that this is the biggest variance and be actually be able to look at those numbers, drill down into the details, and recursively do so. Right? You could keep drilling down as much as you want. And with and the result being this, that you get a better insight as to why the software picked the thing that it picked, and gives you more confidence to communicate to your stakeholders as well.
Great.
And just a quick reminder to everyone before we continue with the questions. You will be receiving this recording after the webinar in your inbox. So if you do need to jump, don't worry. You'll receive all the questions and answers following the webinar.
So next question. You mentioned layering other intelligence on top of AI forecast. I've noticed the current AI for forecasting model does not capture any current circumstances like China tariffs nor integrate them into the future forecast. Is this something we'll need to layer in afterwards, or are there any plans to integrate some of these external variables in future releases?
Hey. This is something, I would love to partner on. So we have a lot of ideas and parts on how we could do it, how we could layer in external variables, layer in events that that are either happened or are going to happen, the, like, the whole the the tariffs and all that stuff, and how to kind of correlate that to some kind of a forecast that that we have. We have a lot of ideas there.
There's a lot of work to do as well. So I would love to partner with the there's just to kind of discuss and brainstorm a little bit on how we could do that.
In general, how these technologies work and they work well is if there is historical context to it. So if there is a way that the machine learning engine can look at the past and say that, oh, every time this external event or external incident happened, this was the impact on the business, then it can learn from that and you can apply it in the future and say, okay. If there's another event like that were to happen, this is going to be the impact on the business. So that's that's what it's good at.
So some of these China's that is specifically, you know, it isn't obvious that it'll be able to do a really good job there. But some of the other things like seasonality. Right? Like, hey. You know, storms happen and, that impacts retail businesses like that kind of a thing. There are the the core mechanics exist.
We have a lot of ideas and thoughts about how we would, kind of surface that, but this is these are still early days. Would love to partner with with you all as we as we, dive further into it.
Awesome.
Next question. Where did Slackbot get the budget info originally?
Good question.
So the way this works is so in in the Slackbot so so effectively, what what we have is a cube application, a cube app within the Slack bot. So you you've install it into Slack like you would many other many other applications. So another one would be, like, if you use Google, you know, Google Drive as its own Slack app, if you will. So they can there's a Qube app. You you you connect with the Qube app. There is a way to tell which Qube environment you want to link it to, so basically associated with your Qube login effectively.
And then that becomes the bridge. So next time when you ask it a question, the first layer of it is the AI part, which will just say, okay. I understand what the user is asking. I'm gonna translate that into a way that that, that can be that can be mined, right, that can be used from a data mining point of view. It's also going to apply the context of the user who's asking the question. So you cannot ask questions for which you don't have answers to, or at least we won't let you get answers to it. You need to be authorized to actually get that data.
And then it it once it knows the connection, it's able to then translate that into a query on our data that's sitting in our system, in our multidimensional system, and it's able to respond back. So, this is part of the reason why I am at least personally so excited about it. It's the ability to have this, you know, natural language conversation that gets translated into something that becomes a specific question that you can ask in the software, get the response, and then it it translates it back into natural language. So it's a very fluid conversation, and there's a lot of potential there to to really make it powerful, insightful, and, commonly available.
Right? You can imagine that this is something that you can roll out. You can you can have your executive team access it. You can have your business counterparts access it, because it's gonna make sure that it still controls what access you have, but you still have this fluid conversation back and forth, like, just like you would with, let's say, an FP and a analyst in the room.
Nice.
Next question we have here. When you get information from the Slack slash cube AI tool, does it link to the source data anywhere in case you wanna look further into the source of the information?
Right now, no. I so maybe maybe the way I would explain this is right now, the conversation you have really is with your data setting in cube.
It doesn't directly go against the source system. Right? Like, you're they're not trying to then relay the question back to the source system right now, but definitely a possibility. You can almost imagine that it will do some of these interactions and some of these trade downs. One of the use cases that we were thinking about is how do we let the user describe what they want to dive into the details and actually effectively do that drill down analysis where we are actually doing the drill down and going into the details.
Again, another candidate for for great discussion and debate and and for us to think more about it.
So right now, it's gonna directly talk against newer data sitting in queue interacting with it to begin. And then I would love to get the feedback on how that is working and then also brainstorming on what else we could do there.
For sure.
Right. These next two questions are kind of related, so I'll ask them both at the same time. First, is there an option to install the chatbot with Teams?
And is Kyub an app I need to download in the Teams app store?
So the Teams is coming up next. So the team is active our team, not to overload the term, but our team is actively working on making it available on Microsoft Teams. So that's that's happening as we speak. So you should see, a beta program that we run on on all of this on Teams as well. So we will get back to you, on Teams very in in short order.
The the important thing as well is to remember that this is all in beta mode. And the reason that we call it in beta mode is really, you we all know this. This technology is moving fast. It is, it's it's evolving very quickly and it's improving very quickly, but there are still things that we want to test out. So part of it would be that you will participate in the beta program with us.
We would look at it collectively. We'll look at it together. We'll make sure that it is doing the right things, and all of us will build build build conference around this. But, yeah, that's that's the journey that we are on. Slack is already in the beta program, and Teams is coming up next. I'm in very short order.
Great. Alright. We've got quite a few questions here. So I think for time's sake, we're probably gonna have to answer a few of them post webinar, but I'll send two more your way. So this next one, it's nice having BVA analysis components, but is there a capability for the AI to analyze and flag missing accruals during the accounting close process to avoid accrual based variances?
Yeah. No. This is something that we've been thinking about.
This problem about missing data, and then and, hence, kind of alerting it is, is a very both it's an interesting problem and it's it's a solvable problem. So there is something that we want to be able to do with this, effectively, being able to be able to compare what the source data because at the end of the day, for a lot of these accrual based accounting, we we depend on the source data. Right? It's a ERP data that's sending us all your general ledger data that's sending us.
That becomes kind of the basis for it. Our ability to detect and anomalies within that to say, hey. You were sending us this information. Suddenly, it missed or disappeared last month.
Did you really intend to or not? So this is something that we are looking to explore and investigate.
So that'll be one aspect of it. Can I can I detect both absence and presence of data that I didn't expect? So so build a baseline and then identify that, hey. I didn't expect undetected.
And then the second part is then proactively inform you of it. So the second half of what is now, can can I raise the hand and say, as this data agent, the I raise my hand and say, hey. It looks like there's something here that you need to take a look at. So there's a lot of potential here. This is something that we are looking at actively. Again, another one where we'd love to partner with you all, in terms of specific examples or use cases that we could chase down.
But, yeah, I think I think that that's definitely a a big part of the power of the tool, and, something that we're actively looking at.
Awesome. Alright. Last question we'll answer live today, and then everything else will be sent via email, so be sure to keep an eye on your inbox.
Can you tell us more about the approach towards the opportunity of prescriptive AI that you mentioned at the very end where your tool is providing us with options to take action. For example, options to reallocate budget to cover upward forecast trends above budget for unplanned spend.
Yeah. No. This is this is where it can really be amazing what is possible. Now to be fair, the preliminary analysis that we have done, it it may not all be there yet.
But directionally, this is where it's going. Right? Like, the the the goal is, as a finance leader or as a business leader, I just describe what is the objective that that I want. And what I'm what am I trying to do here?
What what's the goal As as I look ahead over the next, what am I, twelve months, twelve months and eighteen months, and the system comes back to you with, hey. Here are three options. You've got your you know, well, but like you've said, hey. You know, we are we're running hard on our budget.
What options do we have on, on on kind of course correcting? And then the system comes back to you with two options.
It is it is combining a lot of what we talked about, which is the AI aspect of it. It's the combining with the machine learning that we talked about. It's also other aspects like optimization engines, which is yet another piece of technology that we'll we'll have to consider layering on top of it. So there's a lot of potential to do it.
We haven't yet, made a lot of progress on it, but this is something that I can imagine seeing that as we make more progress again, this will be something another one, where we'd love to partner with you and figure out some specific examples and use cases so we could actually go and and refine this engine further.
But this notion of prescriptive AI, prescriptive intelligence is is very compelling.
It's obviously will be very meaningful from a business outcome point of view. And, technically, as well, there is now a set of technologies that are available. So this is definitely on our journey, definitely on the road map, something very exciting that, we could collectively build. We could collectively do this together.
Awesome. Exciting to hear. Well, thank you, Suresh and Taylor, so much for hosting such a wonderful webinar. Very insightful. And thank you to everyone who submitted questions today.
As we've mentioned a few times before, remember to keep an eye on your inbox. We're gonna be sending you the recording, the q and a, and the beta links that Taylor mentioned earlier in the webinar.
And, yeah, thank you so much for joining. We hope you have a wonderful day and are just excited about these new features as we are.
We'll see you next time.
Bye. Thank you all.
Thanks, everyone.
Just a couple of quick housekeeping items to start. We are going to have time reserved at the end of the fireside chat for a q and a session.
So throughout the session today, please feel free to submit questions using the q and a button within your Zoom toolbar.
And after today's webinar, you're also all going to receive the recording of this session and the q and a in your inbox. And for our cube customers, you're also gonna start receiving some beta links for new AI features, and we'll be sure to share our other resources as well.
With that, I want to introduce our main speaker here today, Suresh Bala.
Suresh is the chief product and technology officer here at Qube. He's located in the Bay Area and came over to Qube with years of experience in the FP and A and product space.
He brought all of that knowledge over here to Cube about a year ago and has been instrumental in really developing our product and enhancing Cube to be the most robust powerful solution that it can be. So I'm really excited to chat with him today and learn all about AI and where we're going with that in Cube.
And my name is Taylor Josephs. I'm a senior product marketing manager here at Cube. I was previously a solutions architect, so I may have given demos to some of you on the call today.
But I'm really looking forward to kicking off this conversation here and diving deep into AI with Suresh.
So without further ado, Suresh, I'm gonna ask you first and foremost, we all have really been hearing a lot about AI, not just in the FP and A space, but across every industry.
So I'm really curious to hear from you. What is Qube's general philosophy on AI, and what makes AI really different in Qube?
Yeah. First of all, thanks, Taylor. Really excited to be here.
And with regards to your question about AI, you're you're spot on. Right? If you just step back and look at what's going on across the board, right, like all of us, we can see that there's an AI revolution that's going on already. Right?
It's pretty remarkable. All of us who have played around with ChargeGPT or Claude or any of these, you know, AIs and large language models, it it's almost impossible not to be impressed with what's going on, what the potential for the technology is. And also a little concern. Right?
Like, there's always this I think pretty much all of us could also see that there are concerns about about that, the technology itself, what it can do, and potentially what the repercussions of the technology are.
So, but the reality is it is going to transform the way all of us works. It's going to transform the way finance and finance teams work.
And at Qube, our approach is really how do we leverage this technology, this this evolution of this technology, which is moving to such a fast pace, but doing it in a way that helps our users, our finance users, our business users kind of maximize their ability to be productive, maximize their ability to be effective in their roles while mitigating the risks. K. So that's gonna be our approach into terms of how we layer in the AI technology within Q. And in particular, the way we are thinking about it is really along three guiding principles. Okay? So the first guiding principle is around context.
What we have on behalf of our customers is really their historical data, their financial models, their scenarios, their what if analysis. There's a lot of rich contextual data that we have for each of our customers that we are stewards for.
And combining that with this AI technology, there is really a potential to fully unlock some some meaningful, some compelling insights.
It helps organizations and our users understand the past better in terms of their business, forecast the future better, and make better business decisions. So there's a lot of potential there. And this context of layering in the the data that the customer has, their their sanitized data that they have with this technology, there's a lot of potential to do that. And that's one aspect, one guiding principle that we're looking at.
The second thing is around accuracy. At the end of the day, we are talking about financial data. We're talking about financial metrics, business metrics, and there is no room. There's no scope for ambiguity or incorrectness or imprecision.
Right? So we have to make sure that the data has to be precise. It has to be, correct, and it has to be explainable. It has to be transparent.
So there should be clear transparency for the user about how the AI systems came up, analyze the information, and presented the data in a way that they can also then, as a user, go back and understand it better and dive into the details and validate that it picked the right things. So that's the second part, accuracy.
And then the third part is control. This goes to the the risk mitigation part of it. Right? At the end of the day, we take our job very seriously that we are stewards of our customer's sensitive financial data, and we have a really high bar in terms of security and control around it.
And that is not that is the same level of rigor is gonna be applied for our finance for this AI systems as well. As we're layering these AI systems into our platform, the same level of rigor will continue to be applied. So that's that's one aspect. And then the second aspect of control is really that we want to make sure that our customers, our users, the finance, the finance teams, and the business teams that use our software, they have complete control about who gets what access to what kinds of data.
So you can very granularly control what data, what information, what capabilities are being exposed to what user. And everything, including this new set of AI technologies, will all honor that. So we want to make sure that there is perfect control that the finance teams have about, about the data and then who gets access to it. So those are the three guiding principles.
Right? As as we at Qube think about how to leverage this this fast moving AI technology, to help our users and our customers use our software better.
The last part I will say is the other aspect of this approach is what we want to do is we want to really layer in these AI technologies, and and infuse it within our platform. And what I mean by that is as users, as finance users, as business users, as you're interacting with Qube, our goal is that we want to provide this artificial intelligence and machine learning capabilities at your fingertips while while you're working at it, while you're actually interacting with it, while you're examining data, while you're understanding what to do next, that is actually at your fingertips. So it isn't gonna be a bolt on that you have to go separately or do some other perf perform some other actions somewhere. The goal is really, can we do it in the context of what whatever job, whatever function that you're that you're working on? And that way, it makes it more natural, more easy to use, and it just makes this entire thing basically work at your service effectively.
Awesome. That was super helpful to hear from you, Suresh. It sounds like overall, the pillars of context, accuracy, and control are what really is setting you apart in a world that seemingly wants to move fast and break things in regards to AI.
So given that philosophy, how are we really setting up our customers for success in adopting AI?
Yeah. So in fact, on that, we're already on our way. Right? So the first thing we did actually, and this was, you know, more than a year back at this stage, was really our focus around this notion around applying artificial intelligence and machine learning actually more in particular in that situation to do what we call smart forecasting. The idea really is our customers have their historical data in cube.
Can we have this machine learning and artificial intelligence technologies understand the historical data, use that to project ahead and to the extent that the past informs the future, be able to project ahead and show you what the your business potentially might look like, based on historical so that was smart forecasting. So that was something that we rolled out a few months back, almost a year back. The next thing we did, which I was really interested about, and I'm sure we'll talk more about this, was around smart variance analysis, which is really something that all finance teams do month in and month out, quarter in and quarter out, which is really to compare what did we say we are going to do versus what actually happened, what's the delta, and how can I, explain the root causes of of the variances?
And this is something that we applied the AI ML. So that was a second second, kind of, for a hint of this, if you will. The third thing that we are actively rolling out, which I'm really excited about, is this notion around ad hoc analyst analytics agent. So the idea really is almost think of it like as as as a junior financial analyst who's at your fingertips, but it's really it's a software.
And you can interact with it, ask questions, and get responses, and and chat with it. So there's a lot of potential there, and we can talk more about that as well. So that's kind of the journey. So we have had a lot of things that we're already putting in place.
We're continuing to build this out.
And as we build this out, the goal is really how do we help our users, our finance users and our business users, automate what can be automated. So that's one aspect of it. Second part of it is provide more insights about their past, help them forecast the future more confidently. And then the last part is how do we help, finance teams and business leaders make decisions more confidently, based on these technologies and based on the underlying data. So have to make data driven decision making a core part of it, and there's a lot of things that these technologies, can do to support that.
Excellent.
So I heard you mentioned a lot of different applications of AI within Qube, and I definitely want to dig into all of these today. But the first piece that I really wanted to focus in on was that topic of using AI to start gaining insights, particularly around variance analysis.
Now I know we recently launched smart analysis in beta in Qube, and I've been playing around with it a little bit and think this would be a really good time to show this off to the group here today. I'm gonna go ahead and share my screen here.
So right now, I'm in my Q web platform, and I'm in this smart analysis section of Q. And what I wanted to do today was really dive deep into the variances between my actuals and my budget.
Now between those two scenarios, I wanted to focus the scope of this analysis a little bit further. So I tacked on some filters, like just focusing in on quarter one of this year. I also only wanted to investigate variances related to my US entity.
And finally, I only really care about variances that are going over a thousand dollars, so I applied that threshold as well.
So from that point, I clicked analyze here within Qube, and you can see these results are starting to render. So Qube is leveraging that AI functionality.
And now Qube has generated for me my largest variance by amount here at the top, and it's given me really quick insights into what's driving that core variance. We can see it's being driven from platform revenue, mainly in the month of March.
I'm also getting these really nice charts too that are outlining my top five variances, and I've hovered into some of these too to understand a bit more about the makeup of that data.
But the piece that provides the most context for me here down below is this summary portion.
I can see that this AI smart analysis has spun up a quick summary telling me that my quarter one financial performance significantly exceeded my budget. We can see that we had really strong revenue outperformance that sort of surpassed some of the expenses that we might have incurred in that quarter.
And it even took a stab at finding the primary root cause behind these variances, showing that revenue outperformance of two point seven million flowed through to the bottom line with opex increasing only moderately.
And beneath that I'm also seeing some other variance highlights too, but beyond just getting this summary glimpse, I also wanted to dig a bit further into what was making up these numbers from the bottoms up. So what I did next was I toggled over to this report section and this outlined in a lot of detail for me sort of what was driving those top variances.
I first made sure to sort this, so we can flip this ascending or descending. And what I really cared about in this case was digging into revenue. So from this revenue account, I clicked into this, and this actually brought together a drill down for me, outlining the four core GL accounts that made up that total revenue variance. We can see their amounts highlighted here on the right.
Since platform revenue was my largest variance, I wanted to explore this a bit further by digging into that quarter. This is outlining that March had the highest variance of that quarter. So from there, I dug a bit into products, finding that my enterprise product generated about five hundred and sixty thousand dollars worth of that variance and even going all the way down to markets. And I'm getting this really key highlight here of what's driving those variances.
So I found this to be super valuable, but I would love to hear from you, Suresh, kind of your thoughts around how this could add value to our customers.
Yeah. Yeah. But this is a really good example of of combining the best of the data that's available and these AI and machine learning technologies to to really develop insights. Right?
Like, I I really like this because there is something that, you know, every finance team does. Right? You do variance analysis all the time. And to the extent that the system can take a first stab at really developing, insights into it and then allowing the the the user to to review it.
But then she'll also be able to examine further. Just like you showed here, we could then examine it further, dive into the details, understand that, get get get a little bit more appreciation of what the data is that's supporting some of the inferences that AI came up with. That's that's that I think is is the best combination. But, like, that's what makes sure that as a finance user, as a business user, you're getting confidence about the the analysis that's going on, and you are feeling more and more confident about, the root causes that are being identified.
So so this is a really good example of of where we could go with these kinds of technologies.
And I'm really I'm really excited to see the adoption around this. Would love to hear feedback from our customers and users as they use this more and more.
Yeah. This is this is really, really cool.
That's great. Thanks, Suresh. And, yeah, speaking of feedback, I want all of our customers on the call today to know that what I just showed here is actually available in beta in all of your cubes right now. So you don't have to install anything. You don't have to add extra applications or programs, but we can share out a resource to you after today's webinar so that you can start trying this out for yourselves. And, again, we would love to receive any feedback from you as you're starting to test out this feature.
Awesome.
Well, another use case that I wanted to dig into is beyond just leveraging AI in Qube to start analyzing our existing datasets, I wanna understand a bit more from you, Suresh, about how we can use AI to start generating new data, like maybe forecasts or plans.
Yeah. No. That that's a great point. But this is this goes to, again, the use case based, thinking that we are applying. That another classic example of what finance teams do oftentimes is on a monthly basis, on a quarterly basis, they forecast the rest of the year, maybe, or they forecast, you know, sliding window of the next eighteen months or whatever the case might be.
And the idea is, can we and this is what I was describing earlier. Can we have the system? Can we have machine learning examine the historical data, identify things like growth and seasonality and outliers and, like, try and learn how to ignore outliers, and use that as a way to project ahead. Right?
Like, you can almost imagine that this is a way by which the software comes up with or the system comes up with, hey. If the history is any indication, this is where you will end up end of this year, end of next year, whatever the the time horizon might be. And then this enables finance teams and and business teams to be able to say, okay. This is what the system is suggesting.
But, of course, the system does not may not know some of the actions that the business might be considering to do, you know.
So there might be other business judgment layers that might be layered on top of it. But it becomes a good way to do a sanity check, right, to do actually to be able to review it. It could also become a good way to have a starting point. Like, this could be the starting point based on historical data, and then you can layer on top of it other intelligence that you have about the kinds of actions that that the business is planning to take. So to me, yeah, there's a lot of potential there, of letting of letting the system, do some of the heavy lifting, apply these AI ML technologies, combine that with the data like we talked about earlier, and then use that as a starting point to to both analyze and, in this example, forecast ahead. And then then then be able to then, let the business users apply and the finance users apply the the strategy and the, and the decision making on top of that.
Excellent. That's super exciting.
You know, something that I've been hearing a ton of, and I'm sure everyone on this call has been hearing this too, is the concept of conversational or plain language AI. So, Suresh, I'm curious to hear from you how the product team is thinking about this in Qube and how conversational AI might start developing in the future.
Yeah. Yeah. No. This is definitely something that's moving fast and but seems to have a lot of potential.
So we are pretty excited about the potential and the possibilities around this. Some of you might have heard of this term called agent e k I. That's becoming a new thing that's coming into the market. And the idea really is, we all are now getting more and more comfortable and familiar with Chad GPT and Claude and all these systems which do things called LLMs or large language models.
The next generation of that is can you then use that to actually perform actions on your behalf based on natural language instructions that you provide to these engines?
So so there's a lot of potential there, in terms of how we could do it in our context. So we are very excited about how we could help our business users, our finance users, kind of interact with these agents, give it give it instructions, then have it do the actual heavy lifting and come back with the results.
So there's definitely a lot of a lot of potential there that that, that we could do. And you and I know about this. We recently are just rolling out. It's still on beta mode, but we are really rolling out our first notion of our AI agents.
It's gonna be really focused around ad hoc analysis. So the idea really is that you can ask it a question, get get responses, and interact with it in natural language. But it's doing the same thing that we talked about earlier. It's combining the best of the artificial intelligence technologies with the data and the context of the specific business, of the specific, company and and and then combining that tool to provide meaningful insights.
That's amazing.
Yeah. And I know we do currently have that pilot program running for, you know, that conversational AI in Slack and in Teams. I'm a Slack user myself, so I've been trying this out, and I wanted to share this with the group as well today. I think that this feature really speaks well to those pillars of context, accuracy, and control because this is really a clear picture into my numbers and into, you know, how this AI chatbot has answered my questions.
To walk you through what I asked, about an hour ago, I asked what was my revenue per employee in q one of twenty twenty five? Now this chatbot only took a couple of seconds to respond to me, and it gave me a really descriptive answer.
It first let me know that it's gonna help me calculate that revenue by retrieving both my revenue data for quarter one and my employee headcount information.
Now it first needed to find out the number of employees, and it was able to calculate this here down below, and it showed me all of the math and the insights that it applied to get there. It showed me that my total revenue for q one was about twenty one million dollars. My employee head count was ninety four doll ninety four, and it also gave me an option to calculate this based off of my FTE count, which was ninety three point five.
Now from there, it showed all of the math explaining that the revenue per employee was that twenty one million divided by ninety four, giving me about two hundred and twenty six thousand dollars per employee. And it also calculated that based off of FTE count, which was roughly two twenty seven.
From there, it gave me sort of insights into what that metric means. So it's showing me that, you know, per employee, we're getting this really high revenue amount, which is primarily driven by our platform revenue with additional contributions from support and implementation revenue.
As I start scrolling down, it even got more specific in a cube context telling me the dimensions that it used to get this number. It was filtered on my actuals, my global company, total departments, and markets, and it even spun up some data that was available for me to download. So what I was thinking here is that maybe I could download this headcount component or this revenue component, bring that into Excel. I could even start drilling into the details using cube from there. But Suresh, would love to hear from you sort of any other thoughts or insights you have on how this could bring value.
Yeah. No. This was such a this is such a good example. Right? So the one aspect of it is you started off in the natural language conversation.
Right? You just asked it a question. You can almost imagine the situation is there is a meeting going on and a senior exec asks the finance person or finance leader in the in the room, hey. You know, what is our revenue per employee?
What is it this year? What is it last year? And has it improved or not?
You don't need to now have it be predefined anymore. Right? Like, the beauty of this is it's it's smart enough to first understand the intent. It's then smart enough to interrogate cube data to actually come up with the results. And like you like you described, then it's also able to explain to you how it arrived at these numbers. So you, as a finance user or as a business user, you have confidence that this is following some logic that you can understand, that you are comfortable with, and, hence, you are confident about sharing this with your stakeholders, with your audiences.
So there's a lot of power in bringing the best of, AIs and these notions around LLMs, which is the natural language part, with the robustness of Qube, its data, its structure, its ability to perform these kinds of analysis. And bringing the two together, there's just so much of, so much of potential here, which is pretty exciting.
The other thing to think about this is that you just open Slack and you just started to type on this, which is which is pretty awesome. Right? Like, the the this also meets the other goal that we have that we want to be able to meet our users where they are. So if finance if your finance teams and your business teams live in Slack, then, hey, you know, you could just interact with this in Slack. Similarly, we're also doing it for Teams, so that could become a third second part.
The goal is really to expand the ability to interact with us in this kind of a national language wherever you are, wherever you as a user are, whether you are on Slack or Teams or you're on our browser, that web based presence. Or if you're on your spreadsheet, especially if you're on your Google or or Google Sheet or Excel. The idea here is, can you interact with it, ask it these kinds of interactive dynamic, just natural language fluid questions, and be able to get the results back. The other thing that you pointed out, which I'm I'm really happy about as well, is that it also provides you with the supporting data.
So the table below that you showed here, it's really cool because it also tells you, hey. Here's the data that I got from Qube to support the core Qube engine to support the analysis that I just did. So that also builds just more confidence and more understanding about, okay, this is how this AI agent actually tell you what the result that I was asking for. And, hence, I can be more confident about sharing this with with my stakeholders.
So really yeah. So really excited about the potential here. Maybe the last thing I will say about this is, this is just the beginning. Right? I can imagine that this one we can we can and then we are actively working on this. So this is the notion about ad hoc analysis. So how do you actually just ask questions and get responses back?
The next part of it with something that I alluded to earlier as well is this notion of actually performing actions. You could actually also give it instructions about to perform some actions. So we're really looking we're really looking to invest in strengthening that kind of muscle where it can actually also perform actions. So things like change my models, spin up a new scenario, you know, map some data from a source system, a ERP system into cube and be able to do all of that in a in a in a way where you can describe the intent as a user and have this agent run and do the actual work for you. That does that make sense?
Absolutely. Yeah. Really, really exciting things on the way. And I just wanted to remind everyone as well that this conversational AI chatbot is currently in beta, and we have a pilot program running as well. So if you're interested in that, we'll send out links after today's webinar where our customers can sign up and try this out for themselves.
Awesome.
Well, Suresh, I know you've kind of touched on this throughout this session, but I'll just ask it again to sort of wrap things up today with our fireside chat.
Where is Qube really going next when it comes to AI? You know, we've talked about the features that we've released that are in beta, but I'm really curious to know what else is on the road map.
Yeah. Yeah. No. Like like like I was saying earlier as well, there is a there is just so much of potential here.
Our our general thought and this is something we would love to partner with our customers on because this is definitely something that's fast moving. It's evolving quickly.
But a general thought is what we can do with the with the core cube model, the richness and the sophistication of a cube model, the data, the the financial model that sits on top of the data is really build, you can think of it like a junior FP and A analyst. A junior analyst that can perform the task. It still needs supervision, so it still needs our the the the folks on this call. Right?
Whether you're finance or business users, you're still providing the guidance. You're still providing that, the targeting. You're probably also still reviewing the results. But this can do a lot of the heavy lifting.
This can do a lot of the analysis. This can actually sift through a lot of data to come up with some of the insights that that the business would actually need.
And and it can really perform this notion of a junior financial analyst that can actually do a lot of this work. So this notion around agentic AI combined with the context of, combined the context of the data that I talked about earlier, there's just a lot of potential for for this to to take on. The use cases that we are exploring are things like how do we make data management, easier. Right?
Like, can we ask these agents to actually, you know, get the data in a shape that it can be used, effectively for analysis? Can it actually perform the analysis? Can it come up with insights as it looks at historical data? Can it help you predict the future?
So we talked about that a little bit. How can it help you get to generate first better forecast so you can actually look ahead?
And then the last part that we're also exploring is can it also help, do what I consider as prescriptive analytics? And what I mean by that is can it actually prescribes given a set of objective functions, if you will, like the goals that you want to, have, can it actually come up with here are three options that you have, And then you can decide which option to pick and and why and so on.
So there's a lot of potential here about leveraging the best of this fast moving technology, combined with the robustness of how Qube manages and organizes the data and, the models on top of it, to make a meaningful impact, to make our finance users, our customers, our finance users, our business users be that much more productive. Right? Like, so this this thing can do a lot of the the grunt work, if you will, that frees up our users to actually become more strategic, to become the key advisers to their business partners, to to help drive more data driven, and more, kind of insightful decisions for the business based on the, the underlying kind of work that that these agents can can do and do value, value.
So that it's really exciting. It's exciting what what can be done. I'm looking forward to it. And then the last part I would ask for, as a parting thought, and, Taylor, you briefly mentioned this as well, is, we look at this as a journey.
We look at this as a partnership with our customers, with our users. We would love to partner with you on this. So there is just so much interesting things that we could go do, and we would love to partner with you. We would love to brainstorm with you.
We would love to develop these, get feedback, iterate quickly, learn quickly, so that we are building tools that are that are making you more effective, more efficient, better at your job, and hopefully freeing you up to do more strategic work and having this do all of the the legwork, if you will, to support that kind of strategic financial analysis that that you'll want to do.
Awesome. Thank you so much, Suresh. I feel like I learned so much from you today, and it was really awesome to hear all of your insights.
I'm sure we have tons of questions in the q and a, so we'll definitely leave some time at the end of this session to get to those. But before diving into that, I just wanted to share these resources again that we plan on sending out after today's webinar.
First resource is for our existing customers. We are going to share out a link for you to sign up for that pilot program so that you can start playing around with that conversational AI chatbot, the one that I just showed earlier.
Secondly, we'll also be sending out a link for you all to access the smart analysis feature within your existing cubes. So So, again, this isn't a program that you have to sign up for. Super easy access link. You're gonna be able to start playing around with your existing data. And, again, please share any and all feedback that you have with us.
And lastly is for any prospective customers or folks who are still checking out Qube. We will also be sharing out a link for you to book a custom demo with us so that you can start exploring other topics, whether it's AI related or more broad FP and A topics with us and learning more about how Qube could fit into your business.
So with all of that being said, I'd love to move into the q and a session now, and we can start answering any questions that folks have had.
So, Alyssa, I'll have you start reading some out.
Thank you, Taylor. Hi, everyone. My name is Alyssa. I'm on the marketing team here, and I'm gonna help out with the q and a today.
So first, thanks so much, Taylor and Suresh, for such an insightful conversation. This has been really exciting. We've had quite a few questions come in. So I'll throw them your way, and you can decide which one you would like to answer.
So the first one we've got, for the smart analysis, can you do the same for current month versus prior month or prior year?
Hey. It's a it's a great question, and this is something that we want to do next. So right now, it is doing it's comparing versions of of your plans, but you could compare actual versus plan.
But it's a really good question. Like, the next iteration of this will really be to do period over period comparisons and kind of the same kinds of insights.
Yes. So definitely on the on the on the road map. So great question. Yep. Makes perfect sense.
Nice.
Alright. Next question. Can smart analysis do a detailed variance analysis with commentary based on a full trial balance or detailed p and l?
Yes. It can. So so I don't know if you if you saw that when, Taylor was showing. You can decide you can define the context that you wanted to analyze. And you can say, hey. It's gonna be my entire p and l, or it could be my entire balance sheet or cash flows or any specific aspect. Could you also be much more precise?
And for each of that, as long as you provide the context and you provide the two scenarios in this case, right, it could be actuals versus plan or actuals versus forecast or what have you.
Yeah. That's all it is. Right? And what it's gonna do is it's really gonna make two passes at it. There's gonna be not to get too technical on you all, but, there is gonna be a machine learning pass where it's actually gonna do a comparison and try to establish what are the actual true variances and which is the highest contributor. And then the second part is then it applies the artificial intelligence on top of it to actually kind of develop insights and then communicate those insights in a way that's easily readable and and, explainable.
And then the third part is the one that Taylor showed you apart, which is also very powerful is you can then explore further. You can then say, okay. Now that I see it, I want to understand why did you end up saying that this is the biggest variance and be actually be able to look at those numbers, drill down into the details, and recursively do so. Right? You could keep drilling down as much as you want. And with and the result being this, that you get a better insight as to why the software picked the thing that it picked, and gives you more confidence to communicate to your stakeholders as well.
Great.
And just a quick reminder to everyone before we continue with the questions. You will be receiving this recording after the webinar in your inbox. So if you do need to jump, don't worry. You'll receive all the questions and answers following the webinar.
So next question. You mentioned layering other intelligence on top of AI forecast. I've noticed the current AI for forecasting model does not capture any current circumstances like China tariffs nor integrate them into the future forecast. Is this something we'll need to layer in afterwards, or are there any plans to integrate some of these external variables in future releases?
Hey. This is something, I would love to partner on. So we have a lot of ideas and parts on how we could do it, how we could layer in external variables, layer in events that that are either happened or are going to happen, the, like, the whole the the tariffs and all that stuff, and how to kind of correlate that to some kind of a forecast that that we have. We have a lot of ideas there.
There's a lot of work to do as well. So I would love to partner with the there's just to kind of discuss and brainstorm a little bit on how we could do that.
In general, how these technologies work and they work well is if there is historical context to it. So if there is a way that the machine learning engine can look at the past and say that, oh, every time this external event or external incident happened, this was the impact on the business, then it can learn from that and you can apply it in the future and say, okay. If there's another event like that were to happen, this is going to be the impact on the business. So that's that's what it's good at.
So some of these China's that is specifically, you know, it isn't obvious that it'll be able to do a really good job there. But some of the other things like seasonality. Right? Like, hey. You know, storms happen and, that impacts retail businesses like that kind of a thing. There are the the core mechanics exist.
We have a lot of ideas and thoughts about how we would, kind of surface that, but this is these are still early days. Would love to partner with with you all as we as we, dive further into it.
Awesome.
Next question. Where did Slackbot get the budget info originally?
Good question.
So the way this works is so in in the Slackbot so so effectively, what what we have is a cube application, a cube app within the Slack bot. So you you've install it into Slack like you would many other many other applications. So another one would be, like, if you use Google, you know, Google Drive as its own Slack app, if you will. So they can there's a Qube app. You you you connect with the Qube app. There is a way to tell which Qube environment you want to link it to, so basically associated with your Qube login effectively.
And then that becomes the bridge. So next time when you ask it a question, the first layer of it is the AI part, which will just say, okay. I understand what the user is asking. I'm gonna translate that into a way that that, that can be that can be mined, right, that can be used from a data mining point of view. It's also going to apply the context of the user who's asking the question. So you cannot ask questions for which you don't have answers to, or at least we won't let you get answers to it. You need to be authorized to actually get that data.
And then it it once it knows the connection, it's able to then translate that into a query on our data that's sitting in our system, in our multidimensional system, and it's able to respond back. So, this is part of the reason why I am at least personally so excited about it. It's the ability to have this, you know, natural language conversation that gets translated into something that becomes a specific question that you can ask in the software, get the response, and then it it translates it back into natural language. So it's a very fluid conversation, and there's a lot of potential there to to really make it powerful, insightful, and, commonly available.
Right? You can imagine that this is something that you can roll out. You can you can have your executive team access it. You can have your business counterparts access it, because it's gonna make sure that it still controls what access you have, but you still have this fluid conversation back and forth, like, just like you would with, let's say, an FP and a analyst in the room.
Nice.
Next question we have here. When you get information from the Slack slash cube AI tool, does it link to the source data anywhere in case you wanna look further into the source of the information?
Right now, no. I so maybe maybe the way I would explain this is right now, the conversation you have really is with your data setting in cube.
It doesn't directly go against the source system. Right? Like, you're they're not trying to then relay the question back to the source system right now, but definitely a possibility. You can almost imagine that it will do some of these interactions and some of these trade downs. One of the use cases that we were thinking about is how do we let the user describe what they want to dive into the details and actually effectively do that drill down analysis where we are actually doing the drill down and going into the details.
Again, another candidate for for great discussion and debate and and for us to think more about it.
So right now, it's gonna directly talk against newer data sitting in queue interacting with it to begin. And then I would love to get the feedback on how that is working and then also brainstorming on what else we could do there.
For sure.
Right. These next two questions are kind of related, so I'll ask them both at the same time. First, is there an option to install the chatbot with Teams?
And is Kyub an app I need to download in the Teams app store?
So the Teams is coming up next. So the team is active our team, not to overload the term, but our team is actively working on making it available on Microsoft Teams. So that's that's happening as we speak. So you should see, a beta program that we run on on all of this on Teams as well. So we will get back to you, on Teams very in in short order.
The the important thing as well is to remember that this is all in beta mode. And the reason that we call it in beta mode is really, you we all know this. This technology is moving fast. It is, it's it's evolving very quickly and it's improving very quickly, but there are still things that we want to test out. So part of it would be that you will participate in the beta program with us.
We would look at it collectively. We'll look at it together. We'll make sure that it is doing the right things, and all of us will build build build conference around this. But, yeah, that's that's the journey that we are on. Slack is already in the beta program, and Teams is coming up next. I'm in very short order.
Great. Alright. We've got quite a few questions here. So I think for time's sake, we're probably gonna have to answer a few of them post webinar, but I'll send two more your way. So this next one, it's nice having BVA analysis components, but is there a capability for the AI to analyze and flag missing accruals during the accounting close process to avoid accrual based variances?
Yeah. No. This is something that we've been thinking about.
This problem about missing data, and then and, hence, kind of alerting it is, is a very both it's an interesting problem and it's it's a solvable problem. So there is something that we want to be able to do with this, effectively, being able to be able to compare what the source data because at the end of the day, for a lot of these accrual based accounting, we we depend on the source data. Right? It's a ERP data that's sending us all your general ledger data that's sending us.
That becomes kind of the basis for it. Our ability to detect and anomalies within that to say, hey. You were sending us this information. Suddenly, it missed or disappeared last month.
Did you really intend to or not? So this is something that we are looking to explore and investigate.
So that'll be one aspect of it. Can I can I detect both absence and presence of data that I didn't expect? So so build a baseline and then identify that, hey. I didn't expect undetected.
And then the second part is then proactively inform you of it. So the second half of what is now, can can I raise the hand and say, as this data agent, the I raise my hand and say, hey. It looks like there's something here that you need to take a look at. So there's a lot of potential here. This is something that we are looking at actively. Again, another one where we'd love to partner with you all, in terms of specific examples or use cases that we could chase down.
But, yeah, I think I think that that's definitely a a big part of the power of the tool, and, something that we're actively looking at.
Awesome. Alright. Last question we'll answer live today, and then everything else will be sent via email, so be sure to keep an eye on your inbox.
Can you tell us more about the approach towards the opportunity of prescriptive AI that you mentioned at the very end where your tool is providing us with options to take action. For example, options to reallocate budget to cover upward forecast trends above budget for unplanned spend.
Yeah. No. This is this is where it can really be amazing what is possible. Now to be fair, the preliminary analysis that we have done, it it may not all be there yet.
But directionally, this is where it's going. Right? Like, the the the goal is, as a finance leader or as a business leader, I just describe what is the objective that that I want. And what I'm what am I trying to do here?
What what's the goal As as I look ahead over the next, what am I, twelve months, twelve months and eighteen months, and the system comes back to you with, hey. Here are three options. You've got your you know, well, but like you've said, hey. You know, we are we're running hard on our budget.
What options do we have on, on on kind of course correcting? And then the system comes back to you with two options.
It is it is combining a lot of what we talked about, which is the AI aspect of it. It's the combining with the machine learning that we talked about. It's also other aspects like optimization engines, which is yet another piece of technology that we'll we'll have to consider layering on top of it. So there's a lot of potential to do it.
We haven't yet, made a lot of progress on it, but this is something that I can imagine seeing that as we make more progress again, this will be something another one, where we'd love to partner with you and figure out some specific examples and use cases so we could actually go and and refine this engine further.
But this notion of prescriptive AI, prescriptive intelligence is is very compelling.
It's obviously will be very meaningful from a business outcome point of view. And, technically, as well, there is now a set of technologies that are available. So this is definitely on our journey, definitely on the road map, something very exciting that, we could collectively build. We could collectively do this together.
Awesome. Exciting to hear. Well, thank you, Suresh and Taylor, so much for hosting such a wonderful webinar. Very insightful. And thank you to everyone who submitted questions today.
As we've mentioned a few times before, remember to keep an eye on your inbox. We're gonna be sending you the recording, the q and a, and the beta links that Taylor mentioned earlier in the webinar.
And, yeah, thank you so much for joining. We hope you have a wonderful day and are just excited about these new features as we are.
We'll see you next time.
Bye. Thank you all.
Thanks, everyone.