If you would like to earn CPE credit for listening to the show, visit earmarkcpe .com slash FPA.
Download the app, take a short quiz, and get your CPE certificate.
Finally, if you enjoy listening to FPA today, please go to your podcast platform of choice, click the subscribe button, and leave a rating and review of the show.
And now, onto the show.
From Data Rails, this is FPNAA Today.
Welcome to FPNAA Today, where we dive deep into the issues that matter most to finance professionals.
I'm your host, Glenn Hopper.
And today, we have an exciting episode lined up for you.
Our guest is Tobias Zwingman, a seasoned data scientist, AI consultant, and author of two groundbreaking books, AI -powered business intelligence and augmented analytics.
Tobias has over 15 years of experience in leveraging data, analytics, and AI to help businesses drive growth and innovation.
Today, we'll be exploring how finance professionals can leverage AI to enhance their decision making, improve forecasts, and transform their organizations into insight -driven powerhouses.
Whether you're just starting your journey with AI, or looking to deepen your expertise, this episode is packed with insights you won't want to miss.
Let's dive in. Tobias, welcome to the show.
Hi, Glenn. Hi. Thanks for having me.
Huge fan of your work.
And we talked about this before the show, how, to me, how ingrained BI is to financial planning and analysis and how important it is.
So I love the work that you're doing.
And I think it's going to really speak to our F .P .N .A.
professionals who are constantly getting better and better, like increasing their kind of their data science and BI chops and finding value in doing that.
So really appreciate you coming on.
Yeah, really excited for this and talking about that.
Tell me about your background, your journey from working in data science roles within corporations to ultimately founding rapid AI and becoming an AI consultant.
Yeah, so it's actually been quite a long journey, I would say.
So originally, when I go back to my bachelor degree, I studied business administration, so I didn't start out with computer science or AI or anything like that.
I was always kind of technical, though, like back in school, building my own websites and tinkering with computers.
But at some point, I felt like I'll have to do more business -related stuff.
And I ended up actually working in the corporate world for more than 10 years in different roles.
And most prominently there in data projects, analytics projects, and ultimately diet data science projects, which was like mainly due to the fact that I did a master degree in IT management and learned R programming, which I found out was a high demand skill then to know how to do R programming.
So yeah, I worked in data science and mostly on like marketing and customer -related tasks, like building share prediction models or next best offer modeling for this whole like customer sales marketing domain there, which was pretty good.
But I also worked on, I would say, more you would call these like digital projects.
For example, back then, I was working for an exhibition organizer, and we developed new digital products that were able to track attendee behavior on physical event spaces.
So that was pretty cool.
And we had really great projects there.
Also had like projects regarding data governance and data strategies, all that thing, these things.
But then COVID happened.
COVID happened in 2020.
And yeah, I was working for that exhibition organizer.
And as you might know, that you know, during the pandemic, there are no exhibitions.
So I found that this might be a really good opportunity to finally make the jump and I stopped my own company.
And that's what I did back then in 2020, starting rapid AI, together with a co founder started out as a as AI as a service business, like we build an AI as a platform where people can easily access AI services.
But meanwhile, it has more pivoted into an AI consulting business, because there's such an high demand out there, not just for people to build their own use cases, but also to learn how to leverage AI for their business.
And that's where I think really like the demand is right now people are excited about AI, which is great, but they are very often not really clear about how to use it most effectively and what to do with it actually.
And so trying to help them figure this out.
And that's what I'm doing right now.
Yeah, I definitely want to come back to the comment you made in your work in marketing.
But before I dive into that, though, you said something else that had me thinking about early on, looking back at the people who were able to use machine learning early on it was the e commerce companies and the SAS company, the companies that were digital in nature, and had access to all this data really
before a lot of non digital, you know, brick and mortar kind of companies did, you know, obviously, those companies will continue.
And whether you're a social media platform, or, or e commerce, or so you have so much data.
And that's why you can have really good AI.
But what we're seeing now is with generative AI and everybody talking about it over the last year and a half or two years, even more so than we were before that, for companies that are not in those super digital and maybe even you know, just non technical industries with where we are today, how do you
see AI transforming businesses?
And do you have any examples from your consulting work around that?
Or a lot of companies think they are, they are not ready for AI.
That's my experience.
Like, they think, oh, we need to figure out like data governance and data strategy, and the whole data topic is so complicated, and so on.
And I would say, at least until two or three years ago, there was really the case like if you did not have a good data structure or data infrastructure, it was really hard to do anything with ml or you know, with AI, there were some AI as a service providers out there, which I also like introduced in my
book, where I say you can use AI, for example, to help you fix data quality issues, or to help you make sense of unstructured data and all these things.
But it was so complicated to do this.
And I wrote AI powered BI before chat, GPT was released, even like back then, I was trying to emphasize the fact that you can use AI, even though you're maybe not ready in terms of like data, data maturity in the whole organization, because there are so many opportunities.
But especially since generative AI, I think there's really no reason for any business not to figure out how to what I call augment or enhance processes with AI.
Because it doesn't matter if you're a SaaS business, if you're a brick and mortar business, they are typical things that are just recurring.
Finance, for example, finance, you know, what I call money data, every company has money data, they need to figure out like, you know, what's the planning process like?
What will the business be next year?
There's so many external factors, but also marketing and sales, like the big second pillar, which in the end, you know, boils down to customer data.
So every business has customer data, they want to figure out like what kind of products to offer to them, like what's the best pricing strategy?
What's kind of the best timing in order to release products?
What's the best way to communicate with customers?
And with generative AI, there have been like huge opportunities that can be unlocked just by using it, for example, for personal productivity reasons.
I mean, there are different studies showing that people that use generative AI effectively can raise their productivity by something like 20 to 40%, depending on which study you read, which is like massive.
If you're running a big business and you have a 20 % productivity increase, that's huge.
So I think there are different stages of use cases.
And there are those use cases where you need this super high level of data quality.
But there are other use cases where it's pretty easy just to get started.
I mean, it can be just as simple as like being faster reading and writing emails or you know, it's being faster to generate reports, or just being more thoughtful about putting together a strategy paper or things like that.
It's super, I wouldn't say trivial, like everyday tasks for every knowledge worker out there.
Yeah, I've heard, and Ethan Malik talks about this a lot as well, that with generative AI, it's not like if we were going to roll out a machine learning platform and a big data tool in our company, it would be done from IT with whoever the other involved parties are, and it would be mandated and dictated,
this is what we're going to use, this is how we're going to use it.
But where people are seeing the best results with generative AI is when they put it out into the hands of the workers directly with the guardrails of don't upload proprietary information here, or knowing, okay, we're on an open AI enterprise account, or we're doing it in Azure or whatever it is that they
know, as long as they have guidelines for the employees to use, and they give them some basic training, and put the generative AI out there in front of them, people are finding ways to use this that's in their own benefit.
So that's, I think those efficiency gains are coming from companies that are actually giving employees the tools directly and not telling them use it to do this, but showing them how to use it and letting them figure it out themselves.
Yeah, absolutely. I completely agree with that.
And also the whole training aspect around that you can't just roll it out and just say, hey, we bought co -pilot licenses for all employees, just go figure it out.
And there might be some people who can figure it out, and just because maybe they have used it in their private life, now they are using this for business as well.
But most people won't really care, they don't understand what to do with it, or maybe use it in the way that's actually counterproductive and decreasing their output or generating worse output because they are just prompting chat activity to say something like, write a business plan for me for the next
five years. And of course, it will produce something and it will write your business plan.
But very often, this business plan doesn't make sense, so it's not realign to your business.
So we have to be careful about that.
But I agree. This has been a huge democratization of AI, also in terms of taking AI and data away from the pure IT departments and just putting that more into the hands of business users.
There have been trends before generative AI, but generative AI has been a completely different new level in terms of accessibility for these products.
We see people interacting with AI that have never had anything to do with data or machine learning before.
And they are now chat activity power users.
I'm dropping this on you when I have not even dug into this.
But this morning, I was seeing a lot of updates on CoPilot.
I don't know if you've had a chance to look at anything yet, but all the headlines are like, did CoPilot just get good?
I know there's CoPilot pages, there's a PowerPoint narrative builder.
Yeah, I've seen that.
And honestly, I was surprised Microsoft is releasing that so quickly right now.
I think they are maybe on the pressure because the previous CoPilot was not ideal, so they were high expectations.
But honestly, I'm really getting Microsoft a second chance for this, because I think it's such a huge opportunity for AI augmentation, because there are two main factors.
The first factor is we have the most widespread analytics tool in the world.
Excel has, what, like 700 million users globally?
Something like that by far the most popular analytics tool.
And then we have Python, which is the best understood language for AI.
So every AI model that you have is literally using Python as their default way of expressing themselves programmatically.
So now these two things are coming together.
Native support for Python in Excel and CoPilot support in Excel, which is a force, joining forces of two super trends.
And if Microsoft really kills that one with their new CoPilot integration, which I still have to test out, if they make this really good, it will unlock so many people to do so much more advanced stuff with their data, because they can now literally ask for these things.
And then the machine or Excel will be able to actually also do it.
Before, Excel was just able to tell you, OK, you need to do this and this and this and run this formula and so on.
But now with Python, it's a whole new level.
And I'm really looking forward to that.
But I remain a little bit skeptical because I've seen the previous CoPilot announcements.
So let's see how they do this one.
But I would really appreciate it if it's an amazing integration, but let's see how that works out.
But yeah, it's like a total killer use case.
Yeah, it really is.
And what I've thought from the beginning with generative AI is we've been talking about democratization of data for decades now, right?
It's like we need everybody we need to tear down the silos and get everybody access to the data so they can use it.
But what generative AI does, you know, the barrier to data science used to be you had to be able to do Python or be advanced BI tools and all that.
And now if Microsoft can successfully integrate in CoPilot and let people just with natural language be able to access the Python and tie it in to excel the tool that they're already familiar with and already know this gets really interesting.
And so it's not just democratization of data, it's democratization of data science.
I tell you what, I think that's my personal opinion.
This will be a huge catalyst for data quality and also like the acceptance of data in general.
Because what we saw before is that everyone was talking about like we need higher data quality, we need better data, we need like less data silos and so on.
But then people were like, Okay, but why?
What for? Well, because then you can use this spreadsheet and this BI tool and run this forecast and so on.
But the problem is, and that's also what I'm like discussing in the newest book, augmented analytics.
Very few people actually doing that.
So we assume that everyone in the company wants to open up an BI dashboard and like see what's the sales forecast for next week and so on.
But turns out a lot of people don't want to do this.
But the what a lot of people do want to do in business is to make better decisions and to get questions answered.
So if we find now that this new interaction layer where people can just ask questions, for example, in a team's chat to a chat board or within Excel about questions they deeply care about, and they find out that everything is there, except for the actual data that is now needed to answer this question,
there's a much higher pull and much higher need and businesses in order to say, Hey, we need to figure out this data quality topic, because I want to have the right answers in my chat board.
So I think that by the same way that, you know, AI got democratized, and more people are using tools like chat GPT, more people will now also like have a higher demand for a we I want to have this chat board answer my question reliably, or I want to have this forecast being actually like valid because of the underlying
data being correct.
So by opening just that up to more people, I think this will give the whole data ecosystem much higher pull effect.
At least that's what I'm expecting.
But let's see how how we speak into it.
Yeah, and you know, this goes back to what it's your mention of your start being in marketing and something we talked about before the show to that marketing because of the data that they had, and to my mind, FP and a people have been we're the original business analysts, you know, going back 20, you
know, a million years ago, when I started in finance, I mean, I did that business analysis part of what we do was very important.
But then when machine learning came along, sales and marketing was using it way more than finance was.
And I think it's the nature of how much data they had and, and early on, and also being you know, the sexy department with the top line revenue and everything versus us just in the back office.
But now, you know, kind of what we were talking about with co pilot, and with the democratization of data science, what do you think that finance professionals who are now getting more access to AI and but what do you think finance professionals can learn from other departments like marketing, or even
operations who's been using machine learning a lot, but when it comes to adopting AI and machine learning?
So two aspects, I think from AI, what they can definitely learn is the whole like communication storytelling aspect of it.
Because what you can now do, like really easily with tools like chat GPT is like take your, your report, your forecast, your your dashboard, and just turn it into a like a comprehensive data story where you say, Okay, this is the key insight.
That's what we need to do.
It's just 1234. Like these are the arguments for this, which a lot of people like kind of like struggle to do.
Marketing is often the other way around, like they have their opinion first, and then they collect data to support this opinion.
So with the finance people, they often know the truth of what's currently happening, but they need to find ways to communicate that effective.
That's number one. The second one is to just help them like do lots of these tedious processing's automatically are much faster.
So for example, if you have to consolidate different reports into one big table, or if you have like messy data, and you want to bring that into a tidy data format, these are all things where you, like you could theoretically sit down and write an Excel macro or write some Python code that is kind of automating
that for you. But most people don't want to do this, or they don't have the skill to do it.
And now with AI, they can just ask you to say, Hey, like I have this report, that's what I'm going to do.
This is what's happening every month.
How can I do this? And then it will just give you some suggestions on how to automate this process and just make the work so much easier.
And then ultimately, the third aspect is of course predictive analytics, which like we can also talk separately about that.
But I think this is a whole different field in terms of like, not only understanding what happened, but trying to predict and forecast what will happen in the future, not only considering the data that you have, but also external factors, external market trends, or also like actual data sets that you
would not have access to normally as a like a controller or FPNAF professional.
Yeah, let's take into that a little bit more because on the predictive and prescriptive analytics, that's moving beyond.
And I think that even small companies right now, from a finance perspective, you'd say they have good descriptive analytics.
They can look at the GL and they can say, this is where we are.
And this is, you could even say, this is the trend we're on and all that.
And you can sort of do some basic forecasting.
But the promise of machine learning and predictive analytics and ultimately prescriptive, but the promise is that you can use this historical information.
And like you said, start factoring in macroeconomic conditions and external factors and add new features to make the model more accurate.
But for finance teams who are just starting to use AI machine learning to enhance their forecasting accuracy, are there specific tools or techniques like kind of somebody that's first moving into this sort of predictive era?
What advice would you have for them?
Yeah, so I would say, let's talk about forecasting, the actual tool that you are using is still the same.
You're doing a regression analysis, for example.
It can come in different variants or factors, but let's say you're trying to predict the budget or sales or whatever using a simple regression analysis.
Maybe you're already doing that right now in Excel using the advanced or data tools in there.
Now, what you can do with machine learning is you can still double down on your regression analysis, but the way you get there is a different one.
The paradigm you get there is a different one.
Normally, you would figure out, okay, what kind of variables do I need to integrate in order to have good forecasts?
And what are the parameters?
And so on to make it all statistically significant and minimize the p -values and stuff.
With a machine learning approach, you basically say, hey, I don't care about any, or I don't really prioritize explainability of my model so much.
I just want to make sure that the prediction is on point.
And please tell me, what are the three, four, or five main factors that are driving this prediction?
And for example, with Python and Excel, you now also have access to the whole ecosystem of machine learning and Python, which is like SKlearn, scikit -learn for the main machine learning framework in Python, which is also well understood by every large language more loud there.
So if you just ask it how to implement regression analysis using SKlearn, it's just two lines of code, and it can help you do this.
The good thing about this is that you can, A, get more data.
You can just literally make your table wider, get more variables and factor in more external factors, for example, for different observations.
And you will also be able to leverage tools from interpretable machine learning, which tells you which factors are driving your predictions.
For example, if, in plain English, what happens?
If this value goes up, which other values are also going up?
Which kind of effects or patterns do you see in your data?
And this gives you a pretty good idea of what are the key drivers, what are the key influences in your data set for forecasting future trends or forecasting future financial metrics.
There are so many fancy terms in there, and it all looks so scary for a lot of people, but it's actually not that complicated and not that complex, especially if you have the support or the help of an AI system built into Excel to do this.
And I think there will be some really great insights that people can draw from that, not only for predicting future variables, but also from getting insights of what are actually the drivers here in my data set for certain financial metrics, like sales or costs or whatever.
Yeah, and it's going to be really interesting to see how much Excel's capabilities are expanded as more and more people, even if they don't learn Python, if they're able to do it through co -pilot and it's happening under the hood and they don't even know exactly how it's working, it's going to open
up a lot of possibilities in Excel.
But one of the big things that you've talked about is Power BI.
And I know a lot of FP &A departments have been using Power BI or Tableau and other reporting, even data rails and other tools out there that give them flexibility in their forecasting and the way they look at the data.
Looking now at the Microsoft universe, how can finance professionals leverage Excel and Power BI to enhance their data analysis capabilities?
If someone, if they haven't used Power BI before, or even if they had, but maybe they're just using it, just using the dashboard of it.
In my opinion, you can use tools like Power BI from two different paradigms.
And paradigm number one, which is most commonly used by financial departments, is for reporting and self -service BI purposes, which means you essentially pre -calculate any possible combination in your metric and you allow some slices and dices in your data so people can drill down and get some insights
that they need. This is the go -to way where they hope that either themselves or someone else will do the actual work of digging down and figuring out what are the details here and what are the key insights.
Paradigm number two, which I encourage people to try out more often, is just to load data into Power BI in essentially just one big table, tidy data, which means every column is one variable and every row is one observation.
For example, if you look at financial statements or if you look at different departments, you would have either different departments listed line by line and then different metrics like, for example, sales, cost, or whatever listed by columns or for example by products or customers, whatever.
But that's the kind of structure you would need.
And now if you have that structure and you add more data, which is more variable to the table, it could be that you end up with a table that has 300 columns and 50 ,000 rows.
And with modern tools like Power BI, this is like absolutely no problem to handle this amount of data.
So you can just load it in there.
And once you have that, you have now a table, which is super easily accessible for computers, for AI tools, but also for just simple machine learning algorithms.
And if you do this, you can build in tools in Power BI, for example, like the Key Influencer tool or the Decomposition Tree tool, which I explained in my book, AI Power BI, that allow you to drill down into your data automatically.
So instead of doing all the slices and dices by yourself, you will just tell the algorithm or you would just tell the system, this case Power BI, hey, this is my data.
I'm interested in, for example, the sales metric.
If this goes up, which factors are influencing this variable?
And if we just run the calculations and show you the top 10 or top five metrics that are influencing this variable.
Or for example, take the Decomposition Tree where you say, okay, this is my high level KPI, for example, total sum of revenue.
And then you want to break that up by different contributing factors, but you want to do it in a way where the most important factors are split at the first level.
So for example, the first could be by product, by region, or by country, or by sales department or whatever.
But the AI will figure out what is the best split to make at the beginning to have these greedy splits to see, okay, who's driving most of what and then go deeper and deeper, totally automatically.
And this is a feature which has been in Power BI for, I don't know, five years or something like that.
It's been there for a very long time, but people are not really using it because mostly they are still operating in that paradigm one mindset where they think, okay, we need to deliver all the knobs and buttons so people can do these calculations by themselves instead of thinking, hey, maybe we just
try to integrate our different data sets into one big table, and then allow Power BI to get these insights, to help them get these insights much faster.
So that's what I do to definitely recommend.
I don't know if Tableau has a similar functionality, to be honest.
Maybe they have, but Power BI definitely is a pretty good tool for doing this.
Yeah, I think if we were in academia, this would be frowned upon.
I guess it's almost p -hacking or data dredging.
But to me, when I get all the data consolidated like that, I like to start looking for correlations.
And not that I'm jumping to, that it's automatically causation.
But if this isn't an academic research paper, this is trying to find some insight into your business.
And once you have that view of everything you're going through, you're looking for these correlations and trying to find new information about it, and I think is kind of a tangent there.
But as you consolidate all this information, it's easier to start trying to find those correlations.
Yeah, absolutely. And in the end, very often, you're not looking for these 0 .2 percentage increasing points, but you're looking for the high level insights.
What's the top three drivers that are contributing that our revenue will be declining in the next three years?
This is super important business questions and tools like this can help you figure this out in an instant, given you have the data and if you have it in the right format.
But I think, especially in FP &A, there's such a big opportunity for getting these insights, just from the project that I've worked in, where most people stumble upon is that they try to aggregate data first and then load it into the BI tool because they are afraid the BI tool can't handle all the data.
Whereas, what you actually need to do is load data in a very granular form.
Ideally, every sale by every customer, or at least aggregated by weekly sale or daily sale, depending on what the size of your business is.
But at least getting to a level of granularity, where you're really maxing out the possibilities of your BI tool.
Because what I see often is that these BI tools are not really operating at their maximum level of capacity.
Sometimes they're working at their maximum level of capacity in terms of understanding, from a human point of view, understanding the complexity of the data set.
Because what people will have in the end is 100 different tables with different aggregates.
You have an aggregate table for this value, an aggregate table for this value.
You need to figure out how these things correlate or can be brought together.
This is what really makes it really very complex.
Again, this is not how you would bring that out as a self -service BI solution tool like all your users.
This is more something where you as a solo analyst or business analyst sit down and want to get the insights for certain questions fast.
This is where you can leverage these BI tools especially with Power BI Desktop, which just runs on your computer.
If you have a pretty decent modern computer, it'll have a good amount of storage to process that data.
If you aggregate everything, I guess you lose the trees for the forest when you aggregate everything and you're just working on the aggregated data versus, I think, maybe the mentality of thinking of if I'm using something very specific, a random forest decision tree or gradient booster, whatever, I've
got this sort of explainable model, but you could go even deeper and say, well, if we were able to use deep learning, the more data we have, we don't know what features are important.
If the more data, the more granular data we have, we load that into the model.
By not collapsing it into these augmented groups, you can actually find real information.
These are all the things that you can do, but these things, of course, all come with a cost because there's costs of maintaining this data.
There's cost of moving this data.
There's cost of just making this data accessible to the analyst.
In the end, you need to have a good use case or a business case for using this data or for working with this data.
If there are more users because of the democratization of AI and the easier interface of running these queries or just asking a question, then there will be also a higher potential for unlocking or for bringing this data in because there will be more people benefiting from it.
Simple example, currently, without generative AI, we need to build this predictive model and put it into the dashboard to have these forecasts and so on, which ultimately limits the scope of this tool to people working in the BI tool.
If we now take that, for example, and integrate it into a Teams chat board where you can just ask, hey, what's my sales forecast for next year?
Then the next question will be, oh, why is that so bad?
Actually, it just gives you the answer based on the exact same analysis that you would do in the BI tool, but just trying that under the hood and just processing or formatting that information in a way it gets transmitted or communicated to you in a more natural way.
That's such a big breakthrough.
I think we are not really leveraging that yet.
But if we do it in that way, I think people working with data and data analysts will have such a higher leverage inside their organization.
To be able to get to that point, you really have to kind of reskill and upskill everybody across the organization.
I know you've talked about the need for more data literacy within organizations.
I'm wondering for companies or for departments who feel like they've got a little bit of a gap to get across there, what steps would you say that companies or individuals could take to close that gap and ensure that they've positioned themselves to use these AI tools?
There are four stages where a company can be in.
The first stage is really your company is not doing anything with data at all.
You're just sleeping over the whole data thing, which is not so many people out there.
The second stage where a lot of companies are already in is what I call data active.
Data active means they are using data to make decisions and to influence their reporting or to make better business decisions.
They are actively trying to seek data in order to drive their business forward.
A lot of people are there.
Every time you have an internal BI tool or doing some trends analysis, that's basically the data active stage.
The next stage is data progressive.
Data progressive means you're more doing these predictive analytics things.
Maybe you have a data center of analytics center of excellence in your organizations.
To get to this level, like data progressive, you can use upskilling and getting more people data literate, because this essentially means that you have a certain number of people in your organization, maybe 5%, maybe 10%.
If you're lucky, maybe 20 % or 25 % of quote -unquote data professionals or highly data literate people that are ideally spread across different departments that can help these departments to get better insights, to make faster decisions, better decisions, and so on and so forth.
The ultimate level, which is data fluent, which means every business process in your organization is infused by insights and uses data in order to get this competitive edge.
I don't think you can get to this stage with pure education and pure data literacy because this involves everyone in the organization to become more aware and more thoughtful about data.
It turns out, at least that's what I've seen and also working with larger organizations that are in this journey for the past almost decade.
Not everyone in an organization wants to do this.
There are so many people that really don't care about data.
They don't want to open a spreadsheet.
They think that even an Excel table is just too much hassle.
And even thinking about running their own forecasts is just too much.
What we need to do in order to reach this group of people is not to try them upskill more and more.
We can upscale them to a certain level in terms of that data is critical and relevant and you have to be thoughtful about your decision making and so on and so forth, but not on the tooling side.
For these people, we need to bring the whole tooling side closer into their workflows.
And this is what augmented analytics is all about, trying to integrate and infuse insights into the work that they are doing on a daily basis.
Simple example, if you're a sales rep and you want to know which customer to call first because maybe you have a customer list of like 100 customers and who are you going to call first, you're not logging into a BI system to figure out what's the highest opportunity right now.
What you do is maybe you log in into your CRM system and you want to have this insight into your CRM system, ideally in your dashboard personalized to your specific account or to your specific user profile.
And this is where you can bring in these augmented analytics tools to bring these insights to users so they don't have to go somewhere to get the information.
But this information is just neatly integrated into the workflows that people are doing.
Or for example, it can also be more for in the FP &A realm.
For example, imagine you have a CFO who is thinking about, okay, what could be different scenarios for our top level growth for next year?
Very often, they don't want to go somewhere to a different tool or to a Jupyter notebook that is running a forecast analysis or so on that.
They just want to have this solved in the same tool on the same interface where they are just right now consuming this report.
And this might be Excel.
And so we need to bring these advanced insights or advanced analytics and these insights into this Excel spreadsheet.
And this is why I'm so excited about like the Python integration and Copilot and all these things, because it allows us to tap into this really well -proven ecosystem with very advanced technologies and very advanced models.
But here's the thing.
You can't really leapfrog or shortcut to this last level.
You need to have a maturity, which is at least data progressive or data active that has some guardrails around data governance, that has some guardrails around a data analytics center of excellence that is monitoring what is being done in the organization.
That's only the final stage.
But you have to go through these stages before and make sure that you have all the groundwork in place.
Because otherwise, what you will do is you will infuse wrong or potentially misleading insights into the workflows.
And then you don't really have a net positive but a net negative impact on your business.
And you don't want to have that.
Yeah. And actually, let's dive in because your second book is on augmented analytics.
So I want to dive in and unpack that a little more.
So what does augmented analytics look like?
I guess, one, to the person who's the beneficiary of it and two, to the person who's putting it together.
When you talk about...
So you've got a whole book on it.
So what's the idea of augmented analytics and how a company could move towards applying and integrating that?
Yeah, so augmented analytics.
And I have to say that because this is actually...
The whole book is actually a big learning story from my co -author, Willy Weber, who has been driving an analytics transformation in a very large insurance company here in Europe over the past couple of years.
And where they did exactly that.
They faced this ceiling of saying, we can't upskill everyone, but we still need to make sure that everyone in the organization is using these insights in order to make better decisions.
Because especially in insurance or financial businesses, it's all about interpreting and making sense of data the right way.
That's management, risk assessment, making sure that you have the right pricing and everything.
So there's lots of data involved.
And the big learning that we came up with was to say, okay, we need to give people an analytical leverage.
What do we mean by that?
Analytical leverage means that people are able to draw insights from data without being qualified to do so.
So for example, they are in a certain workflow and they just get context -relevant insights for making a business decision.
There are so many places in business where you have these like context -relevant augmentations where either by looking at different examples or by looking at certain forecasts and giving an explanation of why that forecast might happen.
And just bringing these in into the actual workflows where people are in.
Chatbots are a different example where we say you can augment Microsoft team channels, for example, with a data analyst chatbot where people can just ask this chatbot while they are, for example, in the conversation with a colleague over a team meeting.
Instead of like leaving the meeting and then logging into a BI tool, which no one is ever doing, but instead relying on some gut feeling, you could just bring in the chatbot and just ask this chatbot some targeted question right within the conversations that you're currently having with your either
department head or with another stakeholder.
Just get the insights from there.
So these are all examples of this spirit of augmentation.
And again, it's all about bringing people this analytical leverage to ultimately make better decisions and build better business outcomes.
FP &A today is brought to you by Data Rails, the world's number one FP &A solution.
Data Rails is the artificial intelligence -powered financial planning and analysis platform built for Excel users.
That's right. You can stay in Excel, but instead of facing hell for every budget, month -end close, or forecast, you can enjoy a paradise of data consolidation, advanced visualization, reporting, and AI capabilities.
Plus game -changing insights giving you instant answers and your story created in seconds.
Find out why more than a thousand finance teams use Data Rails to uncover their company's real story.
Don't replace Excel, embrace Excel.
Learn more at datarails .com.
The past several months I've probably had I've had several data based guests on and you know to the point where some of our listeners are probably saying, am I listening to data analytics today or FP &A today?
But it's this is very important to me and FP &A remains crucial and we'll keep talking about that and all the components from the reporting to the storytelling and everything we do around it.
But with generative AI and with increased automation, I think where FP &A professionals we used in early in my career, some of my proudest moments were these super complicated models I built in Excel.
It's like, look at all this.
Look at this massive spreadsheet that took a decoder wheel to understand how everything happened.
But you know, these very complex models but as that gets more and more automated, the way that we're going to add value, like you said, if you have a generative AI tool and you can just go into Microsoft Teams and ask it a question just like you maybe used to.
It's the same question maybe a couple of years ago you would have asked your head of FP &A.
Now you can just ask a chat bot, well, how is that head of FP &A going to provide value?
And I think it's going to be about learning to harness and use that data more and more and to be fully integrated into it.
So I would say whether I'm someone who's been an FP &A analyst for a long time or if I'm very early in my career, if I've just been, I am an Excel warrior and I can do anything in Excel.
And now, though, looking at the potential and the increased amount of data and kind of how more and more people, it won't be a superpower when a brand new Excel user can go in and in natural language ask Excel to do something that he used to do.
So I think by integrating and kind of rescaling and broadening that horizon into something like Power BI, I mean, what do you think for someone who is that they're super high level at Excel, but they want to make that transition and broaden to Power BI, how can they sort of make this transition in the smoothest
way to expand their skill set?
I would say the most important thing right now is to understand the data structure that Power BI needs to be most effective.
I briefly discussed that before.
This is like the tidy data format.
And maybe we can put a link or so in the show notes because lots of datasets that I see in Excel are not tidy.
They might be clean, but they're not tidy in this data analytics sense because maybe they have merged table cells or there's some text going vertically somewhere.
These are all things that are super hard for computers and especially Power BI to make sense of.
What you want to have is a flat table or multiple flat tables where every row is one observation and every variable is one column and every cell has one single value.
If this is the case, then this is tidy data and this makes it super easy for Power BI or any other analytics tool.
Python was able to work with this data.
And I think for lots of people, the big aha moment will be to realize how to bring their untidy data into this tidy format.
And for this, actually, Power Query is actually the big unlock because it helps you to run these data transformations and for example, have this kind of messy format but always bring it into this tidy format automatically and just rerun the whole pipeline.
Everything just gets updated.
Because once you have your data in a flat table, just as simple as saving it to an Excel file and just opening this Excel file with Microsoft Power BI, that's it.
And then you just use some of these pre -built tools like, for example, the Key Influencer tool or you just use the built -in co -pilot to ask it questions.
There's the Natural Language Q &A tool inside Power BI to ask questions right away to your data.
So once you have your clean data in tidy shape, everything is good.
What I've seen mostly in the financial domain is that data is pretty clean because people really care about this data to be accurate, but it's not tidy.
It's not the right format for analytics to really make sense of.
And what people often do is just to make a copy of this data and then another copy and then look at this from this angle and from that angle, maybe pivot it here and there just to answer different questions.
But what you want to do is to bring it into this tidy format and then use that as your central repository to draw any analytical query you have to this special table.
And I think this is very often the missing puzzle piece of bringing Excel tables to Power BI because what a lot of people are doing is they just take their Excel reporting file and then upload this to Power BI or even upload it to chat GPT and then they are wondering why it can't work with this data or why
the results are just so underwhelming.
Well, because your data is not in the right shape.
And I think it only makes sense if you look at how data, especially in FP &A, is supposed to look like in the end because if you look at the balance sheet, that's kind of like a single source of truth.
And then from there, you just go deeper and deeper into all these different realms.
And also if you think about business cases and scenario planning and so on, these are not formats that are made for machine or algorithmic based data analysis.
So you have to bring it to this format.
I definitely recommend learning our query in order to have these transformations correct.
And for this, actually a chat GPT can be super helpful.
For example, that's also one thing that I teach in my courses.
If you have an Excel spreadsheet, just take a screenshot of that spreadsheet, paste the screenshot into chat GPT and ask it, how can I make this data tidy using Power Query?
You will get a step -by -step process of what to do and which step to do in Power Query in order to make this data tidy.
You don't actually have to understand to a full level what tidy data is.
But at least by knowing that you need to bring it into this format, and this is the keyword to tell chat GPT to give you these instructions, you will be far ahead.
Yeah, I've actually found some super quick ways to use chat GPT for data cleaning.
And this is a super simple one, but it's funny.
So first off, if you're just uploading a Microsoft Excel document into chat GPT, it really would much prefer a CSV file that doesn't have all the meta information and all that.
But then you were talking earlier about columns being your features and rows being your observations.
Well, if you think about in finance, and so all the financial statements are the rows are your accounts.
So those are actually the features and then the dates going across the x -axis.
That's your observation.
So even just having chat GPT transpose that and flip it to a way that makes more sense for a machine learning model.
And that's something it can do really quick.
And it can do things like if you're going to be doing some kind of predictive information or whatever, if you have categorical variables and numbers, it'll make sure everything's in the right format.
So chat GPT has saved a bunch of time for that data cleanup.
And I would imagine as you get that into Power BI or any of the tools, but I guess with Power Query you could do the same thing very quickly.
But yeah, but I think what most people do is they try to fix this like messed up data structure in Power BI through just different reports.
And by trying to figure out what is the measure or what is the actual dimension here and just turning a measure into a dimension and then just messing things up and it gets super complicated.
So just try to load it into the right format and then just run the short thing.
And then the next level would be once you have a tidy data set or a tidy table, just take another tidy table and figure out what's the link between those two.
This could be a date stamp or this could be a customer ID or this could be any other point of reference, because this would be like in data modeling terms, your primary and foreign key.
But this is essentially the connection where you can just combine both data sets without actually merging them in one big table.
You can just say, okay, this is my customer table and this is my sales table.
And once there's an ID here and there's an ID there, and then in Power BI, you can just define that there's a relationship.
You don't actually need to bring it into one table.
But this is how you step by step build your analytical data set by figuring what's the most important thing that I'm looking at here.
For example, it could be sales transactions, could be cost, and it could be cost distributed from different departments, whatever.
And then you think about for every row, what's the different or what's the other dimension, the other column I want to see in there.
And if you have that in the different table, you just need to figure out what's the missing link between those two.
And that's how you step by step add complexity to it and also unlock deeper insights.
If there are deeper insights, sometimes you figure out that just by adding more data, there are actually more insights to it that can also happen.
So, well, okay. I know we're running long on time, but I do want to hit one more thing before we get to the personal and fun questions.
Everything that you're working on in the analytics space and the books and teaching around it and the consulting that you're doing, I mean, we're seeing that companies across the board, everything is getting more data driven.
Certainly, finance is getting more data driven.
And as we're looking at making, and also, more is getting automated now with generative AI, it's going to be quicker to do these reports.
If we're not spending all of our time just putting together these models and these reports and doing our presentations, where is finance going to add the value?
And I guess what I mean is, what skills do FP &A folks need to really focus on right now as more and more is automated and we're relying more and more on data?
How can you kind of future proof yourself and put yourself in a position where you're riding this wave rather than paddling behind trying to catch up?
It sounds stupid, but you have to get hands -on with AI.
And the reason I'm saying this is, right now, AI has what's often called, also by either more like a jacked frontier, which means there are certain things that AI can do exceptionally well.
Like, for example, making a prediction based on historical data set with like 5 ,000 columns and 100 ,000 rows and figure out what's the most important thing.
It can do that in a fraction of seconds.
But there are other things where it really struggles to find an answer, things that we would normally think, well, it's actually so simple to give that question.
Even chat GPT, for until two weeks or so, struggled with doing very basic mathematics or very basic mathematical operations.
And this is just because of the way of how these models work.
But the point is that just looking at it, from an external point of view, you would say, wait, if you do these advanced forecasting, why can't you do basic mathematics in high school grade?
It just makes sense.
So it's quite hard to explain why this is the case.
There are reasons for it, but it takes some time to explain.
And the best thing really to figure out how it works and what does not work and what works is really just to get hands on and try it out.
Because this frontier, this jagged frontier is also moving.
We just saw a recent release by OpenAI, where we now have higher mathematical capabilities in these models.
And who knows what's coming in six months.
So you need to be kind of on the pulse and figure out regarding the tasks and workflows that are relevant for your daily job, which of these are inside the frontier of AI.
Because that area is where you need to leverage AI in order to do this for you or do it together with you.
And what are still areas that are kind of considered outside of the frontier of AI, where you need to still double down on your own human reasoning and your own human perception of these things to drive the needle?
This is how you really can future -proof yourself.
I don't really believe in enrolling in super long bootcamps and studies and all that, because you will learn a ton of things.
I mean, it's great for learning purposes.
And if you want to pivot careers and so on, that's totally fine.
But I think for most people, it's very hard in the end to say, okay, what will make an impact on me and what won't?
Just the best way is to reflect what are you doing every day?
What's your typical working day looks like?
What are things that are coming in?
What are things that are super annoying?
And then just trying out how AI can actually help you here with these things and figure out how does the frontier look like for your current job?
And make no mistakes.
There are jobs where literally all of the tasks of what people do are inside the AI frontier.
I recently read an article of professional translators in the European EU Parliament.
They don't worry about their jobs because they're employed by the EU.
They can literally retire there.
It's not about job security, but they are literally saying, well, we're pretty sure in four years from now our job will be non -existent anymore, because translation is inside the frontier of AI.
And especially with regards to all the things that are being discussed on the political level, because these are terminologies that are well known.
And AI is pretty good at exalting that.
Because it has been trained on 20 years of translation material from all these folks.
So these are completely inside the frontier, and it will be clear that their job might not be existed five years from now.
For FP &A, I think there's more nuance to it.
There might be parts that are inside the frontier, but there are parts which are outside the frontier.
And there's no way to answer that in a single sentence, what is inside, what is outside.
You have to figure it out for yourself.
And I think this is why getting hands on is so important, because the experience just adds up over time.
You start learning, and then you start picking up and figuring out, oh, well, this works.
Oh, this works if I do it like this, and so on and so forth.
So it has to be this continuous journey.
So just allocating at some time, one hour, two hours every week, just dabbling around, seeing how these things work, and then also integrating that in your daily workflows.
Yeah. And the added benefit of that is, as you start integrating it and learning more, that becomes your superpower, and you start to be able to be more efficient, have better models and all that.
So you're keeping up with the technology, but you're also making yourself more valuable by being able to use it.
Absolutely. Yeah. 100%.
Yeah. Well, we are getting close to end of time, but we always wrap up with a couple of questions, because we just dove straight into our nerd stuff and got into all the analytics and everything.
But before we let you off the hook here, one question we always love to ask the guest is, what's something that not many people know about you?
Kind of some inside baseball in your life?
I know I typically don't share much personal things also online.
But what I do in my free time, whenever I'm not on the computer and doing all the AI and data analytics, it's just basically spending time with my family.
I have three kids under the age of 10, and lots of people ask me how I can still stay sane with running a business and having family life and everything together, because if you have these nights, and I really like father of young one will know that where you work, you sleep like four hours or five
hours. If you're lucky, you can really get crazy.
I'm behind that stage right now, but yeah, that's something that I still need to juggle.
And actually, I was thinking about also writing a little bit more about how it is to be a parent and working full -time and doing all these things, and also trying to keep up with everything that's coming here.
Because if you think about it, there are so many questions attached to that.
My kids are now going to school, they are learning handwriting and calculating in their head and everything like that.
And I'm like, oh, gosh, should I intervene?
Or should I just let everything happen?
And just watch them know for the best in the end?
Because honestly, I tell you, I do think our education system is absolutely it's not prepared for what's coming.
I mean, there will be a big impact in the next two years, but the impact for the next 10 years, it's like no one is prepared for that.
And so I'm really wondering, okay, what does the future look like, especially for the young generation right now?
So yeah, that's something that a lot of people know about me.
I'm thinking about that quite often, actually.
Yeah. Yeah, on the education front, what happens when you can completely outsource thought?
I mean, it's kind of moving up the chain now, but what does that do to education, when every bit of your homework you could outsource to chat GPT or whatever the latest LLM is?
Yeah, so that's a whole other episode and we are late in it.
So I guess I should get with it.
But I hear you on that and it's going to be interesting to see the transformation there.
Okay, everybody's favorite question, we always have to ask, what is your favorite Excel function and why?
I think my new favorite Excel function will be the Python function and Excel from my own run Python code.
Yeah, that's you know, it's funny is as that as co -pilot and Python is more and more integrated, it's going to be like, you know, we used to remember everybody's phone numbers and like even memorizing passwords for our logins and everything.
I feel like at some point, we're going to forget the names of these functions because we're just it's going to be whatever we call them in natural language.
Yeah, absolutely. And then Python is going to go do it.
So nested ifs and index match and all that or like, well, we remember what that's called.
Yeah. Okay, last question.
How can our listeners connect with you?
How can you find out more about your books?
And I know you've got another one coming out.
What's the best way for people to reach out and stay in touch?
Yeah, I think the best way probably is to connect with me on LinkedIn, I'm pretty active there.
But also subscribe to my newsletter to be a Sigma .com.
You will be redirected to my newsletter.
So feel free to subscribe that I share some AI related insights every week on Friday.
So yeah, sign up if you like and keep in touch there would be great to buy us.
Thank you so much. Really enjoyed having you on the show.
Thank you so much for having me.
It was a great pleasure.
Transcribed by https://otter .ai
