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, on to the show.
From Data Rails, this is FPNA Today.
Welcome to FPNA Today.
I'm your host, Glenn Hopper.
Today on the show, I'm joined by Tom Hinkle.
Tom is a senior leader in data and analytics who's built his career at the intersection of finance, technology and strategy.
He also has a budding YouTube channel called Tom's Data World, which includes instructional videos on Excel, SQL and now AI.
He's worked extensively with tools like Excel and SQL, as well as Tableau and Power BI, and he brings a deep background in data governance, compliance and building systems that actually get used by business teams.
Tom recently stepped into a new role at TIAA, where he's helping drive productivity through sanctioned tools, while continuing his work on data quality and governance.
He's passionate about making messy data usable, about bridging the gap between what systems say and what users trust, and about helping finance teams unlock real value from their analytics.
We're going to talk about how to approach data governance in the real world, the evolving role of FPA and what the rise of AI means for the skills that matter most in finance.
Tom, welcome to the show.
Hey, Glenn.
Great.
Thanks for the intro.
I'm glad to be here.
Yeah, I've been excited about this.
We talked about having you on a couple months ago, and I think we both got busy.
And honestly, I thought we had already recorded it.
So I'm happy that we are finally getting this in front of the microphone and everything, because I always love your insights and background.
And I love the work that you do and how neatly that folds into what we do on the FP&A side.
I think a lot of our listeners are probably already familiar with you but for those who aren't, tell us about your background and focus and kind of what your work looks like today and how you've sort of built your career on these past experiences and data and analytics and working with finance.
All right.
Sounds great.
I'll try to keep it a little short here.
But, you know, just start with my classical education.
I went to NC State University and majored in computer science there.
First couple of years, I was more of a developer, kind of in an IT role.
And then right around 9-11, I
I moved positions and I actually moved into working in the business units.
And that's where I got much more into, got a little bit away from, like just you know, low level programming, got into more in data analytics and did a lot, lot more SQL.
So living in Charlotte, North Carolina, as long as I have there's like eight banks that are headquartered here.
So I've worked at several of them and just spent more time in the financial community.
And then within those companies, it's been nice to move around.
I've worked in asset management capacity and operations and marketing and finance.
So it's kind of neat to take Less is from one place and move them to another and see how you can plug them in there.
So today, right now, I'm working in one of those financial centers in a marketing capacity.
We're looking at link clicks and emails opened and how many people we converted and all that good stuff.
All right.
And I guess we I guess we probably won't have time to talk about NC State basketball.
And I think you're like third or fourth in the ACC right now.
So it's great.
Yeah.
Unfortunately, for many years, I love my alma mater.
But for many years, it's just tough to watch the Wolfpack.
I mean, but the but the last couple, especially when they made that final four run a couple of years ago.
That was just fantastic.
And the way they ended up, I didn't even feel that bad when they got knocked out.
I'm like, you guys showed up and you were a force to be reckoned with.
Yeah.
So being in Tennessee, I'm a Titans fan, a Grizzlies fan, and a Memphis Tigers fan.
And it's been...
It's a rough year for all those teams.
So yeah, so it's actually.
I've got so much time back back in my days now from not having to watch sports because they've been so bad.
So I don't know.
I guess there's a silver lining in every cloud.
But I guess I'm way off topic there.
It's Friday afternoon when we're recording this, but I'm just I've already checked out for Super Bowl weekend.
More more seriously, and topic at hand um, I love talking to you because our conversations always go into data.
This was my experience.
I always thought of myself as a data-driven data-forward, CFO and even coming up through FPA.
It was very hard to get people to put money and resources towards data governance and all that.
And you said before that data governance is the least sexy but most important part of analytics.
If you could peel that back and dive in a little bit, I'd love to hear your thoughts on that.
You know, today, a lot of times we talk about AI and that's like the buzzword everybody uses.
But you know if, without getting too deep in the weeds with AI, you know it has the machine learning, the artificial intelligence component, but that's all built on a foundation of data.
And so it's very sounds, very common sense-y and very simple.
But if the data is wrong, it doesn't matter how advanced your model is, the model's going to be wrong.
And And it's also just not fun work.
We'll talk more about data quality and data governance and what you have to do there.
But basically, you're looking at the data and saying like, okay, how many things are wrong in here?
Do I have the right identifiers for each of these types of rows here?
Are the dates formatted the right way?
I mean we've all had to deal with date files where half the dates come in as text and half of them are dates.
And then You don't realize until you're presenting it to your leaders that it's wrong.
And so it's just not exciting to go through a big data set and clean it and make sure the data is valid.
But everything else fails if that step doesn't take place.
Yeah, and we've been talking about data governance and data cleanup.
And I think it goes back to when we first saw the power of machine learning, and we've been talking about this for a decade.
Heck, we've been saying all the time, we've been talking about digital transformation for 30 years and I think There have been incremental gains and COVID was certainly a big one to get people more into the modern data stack and systems and all that.
But I work with enterprise clients every day who still have messy data.
And I think about it in FP&A, when we receive the data, how much cleanup we still have to do.
And we're downstream.
A lot of times we don't even have access to it.
And I think and I'm thinking, you know, everything is with a lens towards FPA, but it starts upstream too.
So this could be at the finance level or even at the data governance level.
But When a company has messy data across multiple systems, where do they start with data governance?
Because these projects are so big and they're expensive and it's Hard sometimes to tie a concrete ROI to it.
We know the work needs to be done, but it can be expensive and it's a mess.
And so it's like the can keeps getting kicked down the road.
But are there some things like if you know your data is still a mess, what's step one and where do you go from there for trying to get a handle on data governance?
Well, you got to go back to the age old quote and I'll attribute it to Ronald Reagan, although somebody else probably said it too.
Is you know the trust?
But verify thing.
And I have a good example here, and then I'll get back into how you bring this up at the enterprise.
But I was putting together some analysis on a brokerage group and we're putting together a dashboard to present to the head, the leader of brokerage and all their directors.
I took the data on faith and I put it together and got in the meeting and one of the directors points out to me hey, that data is wrong.
We've got way more transactions than that.
And what?
I'm pretty confident in my ability to put things together and to make the right visuals and to do that.
It won't get into the rest of that meeting, but later when I went back and checked, We were pulling brokerage data from two different systems.
One of the systems, the identifier for it was the actual word brokerage.
And I found out in the other system that wasn't quite as big, it was BRK was the identifier.
And so that's a very simple example, but it happens all the time.
So I think...
Like I said, nobody really wants to, that's the stuff that everybody just assumes is already done.
Why do you have to do anything?
The data should be good, right?
Should is the operative word, but you have to demonstrate why there's a need for this data governance.
For leaders to start feeling like it's something worthwhile, versus spending money on making a brand new five-year forecasting model with the latest greatest AI and seasonality charts and all this good stuff.
Like, hey, if our base data isn't there, It's no good.
It's not easy, but you have to identify some problem areas too.
Another thing those data people have to contend with and this might even be another show is that so many early growing data practices.
They're so gung-ho on wanting to give real-time data to every customer of the data.
I will be a big proponent.
Every customer doesn't need real-time data.
If you're dealing with a phone center where you're looking at hold times, you need it to be up-to-date every second.
But if you're looking at operations tasks over a month or two, you can update the data once a week.
Again, that's another conversation.
But what we found is some of our data was coming back.
We were realizing stuff was wrong and things were getting loaded.
And it was really bad because the business partners were the ones telling us, the data team.
So one of the things I suggest that we implemented is, instead of it being real time, we sold them on, we're going to have this data be one day behind, so that we would have a day to proof and make sure that all the data feeds came in right.
They all looked good.
Then when it went up, we wouldn't have these complaints from the business users.
As you talk about that, I think I'm having PTSD of some of the data cleanup.
Because I think the one you mentioned, datetime, that's the bane of everyone working with data.
It's the bane of our existence, but addresses.
One of my favorite things to do when I go in with a new client that they're bragging on how clean their data is.
I love to sort or get a count of states and then I'll show back to them, you have 56 states here.
Last time I checked, there were only 50.
But there's just so many examples of that and the different abbreviations and all that and it is a mess.
And everybody wants that flashy dashboard, like you were saying.
But there have to be pillars of solid data strategy beyond the reporting and dashboards.
You have to go upstream and address the issues issue itself.
And for you, I mean do you have, are there key pillars that you say these are the foundations of your data governance and data practice?
Yes, I have.
I have some that I've kind of come up with over the years.
I might refer to my little slide over here, but there's basically two sides of the house of data governance in my view.
One is the data management.
Some people call it master data management.
And that's where that's really most effective across the entire enterprise.
That's where you've got a data glossary and you've got like metric names and fact names and they're consistent across different things.
And that's really challenging because every silo or department has their own way to name things.
And sometimes people may use the same name for different metrics across departments.
So that's really important.
And that does take a cross-company effort.
If that's done the right way.
It's a lot of pain upfront, but then when people start seeing the value, incrementally adding those new metrics in, they'll just incorporate the data stewards and the data governance teams in that process and it's easy.
The other side of data governance, I think, is the data quality and the controls.
These are things like on the very simplest level, we've all seen this in IT.
This is pretty much covered, like hey, my data coming in.
I've got data from five different systems coming in.
Just let me know if the data came in from each of these systems today.
That's like bare bones level.
But then you have to go to the next level.
OK for my Siebel data in the last couple of months.
It looks like we're averaging 50 to 60000 rows of data every day.
So let me put a control in there and just set a threshold saying if I get less than 45000 or more than 65000, send an alert to the data team.
It may not be wrong, but that's outside of a normal tolerance of data that we come in, because we're real good about just identifying the data source came in, but not if we get the right amount of data.
Like what if 10 rows comes in?
That's wrong, but that first check doesn't catch it because, hey, the data came over there.
The data quality and controls are something that are a little bit more department-centric, because they are just managing your pipelines and your data flows coming in.
But some of those controls are, you know, comparing data from one system to another, period over period, day over day, month over month.
What's our volumes like?
Completeness, you know, filling out required fields and things like that.
Yeah, and I do like that extra layer of thinking about the data, because you could have your crime job, that is, pulling the data once every night or once every week or whatever, and something fails in that job and you don't get the data.
But if you don't have any checks there, how long is it going to be?
It could fail over and over.
If you're saying well, the data loaded, but you're not getting any warning that something's out of the norm.
I like that a lot and I think that's the kind of thinking you need to have around your data strategy.
I'm glad you pointed that out.
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 1,500 companies use Data Rails to uncover their company's real story.
Don't replace Excel, embrace Excel.
Learn more at datarails.com.
When you and I talked about this first, I think we were talking about data projects and we were talking about change management in general.
However, since we last spoke, there's been so much publicity and press around failed AI implementation.
We can talk about this as data projects or as AI projects.
I think we're in lockstep on this in every aspect.
But you said when we were talking about it back then that you've seen projects fail not because of the tools themselves, but because of the rollout.
We can think about this as in just projects in general.
Maybe now, with everything going on in AI, what are your thoughts on what good adoption and change management looks like in practice when you're rolling out, whether it's an AI ML, data project or whatever the case is?
Because a lot of it really is change management or unrealistic expectations, or whatever the case may be.
Yeah, I'm seeing the same sort of issues coming up with AI rollouts.
Both of the places I've worked and with some other friends are similar.
You know, one of the first things you have to do is give people a motivation, like a why.
We did a rollout of something one time and we didn't have a lot of good instruction or a lot of good handling.
We kind of sent a link out to everybody who thought they'd all like it.
And one of the consumers came up and said, you know what?
If you don't make my life easier, you just made it harder.
So you have to give people reasons to adopt this AI.
And everybody knows the word, but there's still a lot of people that just don't know what to do with it.
And this isn't promotion for this, but I actually ran a class a week or two ago doing just that, like showing people how to get to chat GPT, how to make a login.
What does that mean?
Do they get all your information once you log in?
No.
And then some basics on how to do it.
What I've seen, and like I said, the company I'm in with and others is the same thing.
We've got an AI tool we use, and I hate to say it, but it basically was put on everybody's desktop.
And they just say, OK, go now be be more productive.
And I'm telling you, I use it every day.
But there's so many people in the area I work with.
I'd say probably at least 85 percent of them are not even have not even opened the tool.
So companies need to show people how this is going to be better for them.
You know, say I'm working my 40 hour week.
I don't want to do something that's going to make me have to work 45 or 50 hours a week.
So it's incumbent on the change management process to go and say hey, here's a good example of something that you do.
Like, Glenn, have you ever written an email?
A couple, yeah.
Have you ever written an email saying, to somebody two, three, four levels above you, yeah?
Now, let me ask you this.
On that email to that higher up.
How much time did you spend wordsmithing it to make sure that it was perfectly toned and the right words and the right adjectives were used?
And it takes a few hours usually, right?
Yeah, yeah.
And so that's.
You know that's one of the areas that almost everybody can relate to is helping write an email.
And I found I'd like to get your take too.
But I found too, I used to still try to write a good email.
And now I just write exactly what I'm thinking.
Hey, the Q3 numbers, you know, Q3 numbers for last year are horrible.
And I think The Raleigh office is way under target, so they're not performing.
That's what I would write.
And you can go to AI and send it that very terse and raw email and say write this email as if it's going to a senior leader three levels above me, and I want it to be in a constructive fashion.
And it just does it.
So have you had that experience at all?
100%.
And I'll tell you, for me, when I come in, so many people right now just have this expectation.
And that's not unfamiliar where there's a rollout of a new AI tool, but no clear policy around it.
And there's a lot of confusion and fear around what data can I put in?
What can I not?
How can I use it?
When can I use it?
When do I need to log my prompts or whatever?
Whenever I come into a company, I say one of the first things I look at is give me your generative AI usage policy.
Is it clear the employees understand what they can and can't do, and is it too restrictive?
We uh with senior leadership um, we will sit down and say what is everybody comfortable with, and we'll sort of hash through the security side of it and then we say okay, that's great, you're all aligned on this.
Let's update the policy to reflect that.
Then Let's find the evangelists.
Who are your power users?
Let's get them training, make sure they understand.
Then we're going to do broader training for the whole group and then make it very clear.
And then you take it out of the shadows.
Because what's?
I cite this all the time, but Ethan Mollick, who writes on AI, talks about secret cyborgs that are.
Some people have figured out how to really use AI, and maybe they can't use it in their workplace due to draconian policies, or they're just unsure of what they can do.
So they're supremely maximizing their capabilities but they're doing it under the table.
And when they're doing that, there's a danger data could leak.
They could be doing something stupid.
So that's great for them.
I mean, they may be spending four hours a day working and getting four hours of time back that they're not giving back to the company.
They're out on the golf course or whatever.
I think I agree with all that, but we've got to show people.
Then one of the things that I do on a daily basis is write SQL code.
In this role, I was supposed to write SQL.
Almost all of my SQL code I write through the AI tool now.
Now I will say too and I don't know how much we'll get in the weeds you know the way that LLMs work and the way they calculate things.
You know there's a lot.
They get a lot right, but every once in a while they get something wrong, and every great while they get something really wrong.
So you always do have to, You have to proofread it, but almost all the SQL commands I've asked it to write have been just perfect.
And the nicest thing about having AI write code, it's formatted perfectly.
You have comments in the right place, the right indentions, all that good stuff.
That's just extra nice to have.
Yeah, and for me I've always said grew up as a citizen developer, but no one would ever pay me to sit in front of a blinking cursor and write code.
But I would spend a couple of hours writing one SQL query trying to make my work easier, but I wasn't familiar with it enough and I didn't do it enough.
I would be missing one comma and I would spend as much time QA-ing it trying to fix it.
Now, I just dump my messy SQL query in to AI and have it clean it up.
And that's a massive time-save right there, without even requiring it to write the code.
But if it knows the database schema and what you're working with and everything.
I still am not fast at writing SQL queries, so having AI do that is a huge time-saver.
I have to share one example with you here on something that I did recently.
That was amazingly helpful.
It's kind of in a new unit.
I'm learning the data source, which is always a learning curve.
No matter how many years of experience you have in SQL learning, a new organization's data model is always new.
And fortunately almost every department in the planet has a very detailed with pictures guide to that data model.
Now I'm being sarcastic here.
Like that never exists.
So I was stuck looking at ETL code to learn this model that was.
You know.
I remember the one query I'm talking about right now.
It was probably about 3000 lines of ETL code.
And it was about nine data sources unioning together and getting all these fields.
It was really overwhelming for me.
And I would tell you what I wrote yet.
I hope you think this is cool, but, um, It used a lot of CTEs.
And for non-SQL people, that's just a construct like a table.
It's just one of the main constructs to build this up.
So I went in the AI and I dropped the 3,000 loads of code in there.
And I said, can you write me a CTE map?
And then also on the second page like write me the field definition of what and what the source fields become like when i'm doing this query, what you know, what the transformation is from field a to field a1 on the final and man, it came back and it just gave me this little chart.
It was text chart, but it gave me a chart on all the base queries that were feeding into the larger one, but it was in one view so i wasn't overwhelmed with everything.
It's fantastic Yeah, it is great at summarizing and things like that, just like worksheets.
I think about all the Post-it notes I have sitting around my desk where I'm all these notes to sell, things like that.
But when you can have AI, generate it quickly.
And I think you know a lot.
Certainly, not all of our listeners are working in SQL, but I do think more and more are.
And it is.
In a lot of ways, we are, depending on the company structure, but we're customers of the data team.
But even being downstream, I guess, bringing it back to FPA, To your mind, as someone who's maybe data first, finance second I don't know if that's an accurate categorization, but how do FPA teams need to think about their role in data governance and strategy?
Well, I mean, I think, like I said, you know you always want to trust the data people, but sometimes you have to verify.
So sometimes it might be going to the data people saying hey, I don't need to see your code, but can you give me, like you know, a higher level mapping of what your things are?
What are your data governance checks?
Like I said.
Most places I've worked have that rudimentary check in there that, did we get data from system A?
But then that's where it stops for most of them.
Do we have any threshold checks to see if we get the right rows?
So I think you have to have a reasonable amount of knowledge sharing in that.
And that helps people trust the data.
Because you know this is the other topic we get into why everybody still uses spreadsheets because somebody sees a number the data team comes up with.
They don't believe it.
But everybody knows how to work in Excel.
So they just go to Excel and they do whatever they do.
And then that becomes their number.
But then it doesn't match the corporate database.
And that's a problem.
So I guess I'd have to say that's a tough question.
It's kind of a judgment call.
I mean, if that first meeting goes well and they understand, and then what the data team is coming out with is matching what's in my spreadsheet, then I'm good with it.
There's nothing more challenging for a data person than having to justify their data process against somebody's spreadsheet that may or may not have controls on it.
Yeah, and I think you answered this, but I love when we were talking before the call, you know you're like okay, we've got our single source of truth.
It's the ERP.
So we run a report and then it has to reconcile with, you know, Dave's spreadsheet or whatever.
So yeah yeah, the original question that I had set for the show, I think, and you've kind of addressed this but is why that dynamic still persists.
But I guess the real question is, how can we move away from that?
And I think what you've said before and I'm in agreement with is it goes back to that data dictionary and understanding.
Okay, is this a source of truth?
If it's a source of truth and we're all singing from the same sheet of music on our data dictionary, then Problem solved.
We can take Dave's spreadsheet out of the mix, and this is the answer, which is what we should strive to.
But, you know, then the response is, well, we're not going to go back and change the source data.
So you're going to have to transform it somehow.
It's always, it's like peeling an onion.
Yeah, yeah, yeah, yeah.
And it's tough.
I mean, I think...
One of the things that, as I've evolved in my career, data.
People were generally smart and we just always feel like well, this data is right, I know it.
It's Dave's fault if he doesn't believe it.
But I've gotten to the point where it's like no, it's on the data people to prove that what they're doing is accurate.
If Dave is not doing his math correctly, We need to sit down together and understand.
Either we need to show Dave why he's not doing it correctly, or maybe we learn that we're not doing something on our own.
But it can't be this.
You know, I've seen most of my career.
I've seen just this this blind throwing something over the wall and like.
That's it.
I did my job.
I know my codes right.
If they don't want to use it, that's not my problem.
It is your problem.
That's another show is talking about people in data.
But yeah, it's really important that the people believe in the data that's coming to them.
Yeah, and that's in data flow.
I mean, communication is a two way street, there has to be the feedback.
And then if the end user hasn't trust the data, then they're not.
They're going to be frustrated with the data team.
And if the Data team thinks FPA doesn't understand the data that they're getting and they're just mucking it up?
And I know it would be great if everything could be simplified to a playbook or a framework.
But is that Do you have?
Are there some And maybe you've already hit on a couple of these, but are there some key elements of a framework that gets business users to actually trust the governed data?
I mean, is it a collaboration?
How do you get to that point where FP&A knows, okay, I received this data, I can trust it?
I think that comes back to what we talked about earlier, when you were asking me about the pillars of a data governance practice.
Like You know, ideally the first stop for everybody in making new information or a new table or new product would be the data governance team and saying hey, I'm looking for a field that you know, I'm looking for a metric that shows me how many widgets we shipped out of Raleigh every month.
They go to the data governance team first to see if that's already in existence.
But that doesn't happen.
And that's because, like I said, nobody really gets excited about spending money on data governance.
They get excited about machine learning.
Right before AI and I think it's still a valid buzzword.
But the other big buzzword in our community was storytelling.
Everybody wanted to do data storytelling.
So yeah, we need to put money into data storytelling.
Let's like have these seminars and things.
Nobody wants to spend money on data governance because, at the end of the day, you're coming up with something that you thought was right anyway, but you just had to spend more money to make sure it's right.
I could sketch out the ideal process for an organization, but nobody's going to push it to that level.
In my mind I envision this great data hub website on the internal site where you could go if you want a new field site.
We're counting widget A out of these different regions.
Is there a metric in here that's already measuring that?
And then if not, you partner with the data commerce team to build that metric.
And then everybody trusts it.
But when everybody goes off in their silos and does their own thing and there's no governance around it, that's that's tough.
So I hope I answered your question.
Yeah.
I mean, yeah.
I mean, it's, it's a difficult problem and it's, I've gone both ways.
I've even been in companies before where we built a self-serve data mart, thinking this is great.
All the report requests people can just go get it themselves.
But then we ran into the issue of the users.
We would have a data dictionary that defined this is the KPI.
This is the source of truth.
This is how we measure it.
But they also wanted their own data because sometimes you want to slice and dice it different ways.
And we thought we could really, and this is back in the days of crystal reports.
And, you know, this is I'm dating myself here.
But, you know, we had a huge backlog of report requests and people frustrated with it.
And so the solution was come up with a self-serve data mart.
But then you run into the problem of people not following the playbook.
You're coming up with your own metrics.
Usually whichever one looks better that they're going to present to management.
Then you have management confused with different information.
We ended up severely limiting access to the data mart and to the fields that they could pulled, because it's like we can't have this division head and this department head and this senior manager all pulling different numbers saying they're reporting the own thing.
That was a recipe for disaster.
Oh, yeah.
Yeah.
It's amazing.
You got to look at human behavior and designing these systems and processes too, because when you were talking about that data dictionary, I had a visual in my head, because I've been in projects like that too.
We have data dictionary.
It's incredible how many people will not go to the data dictionary if it's on another tab.
Yeah.
You can have all the fields available on the next tab.
Nobody's clicking there.
They're just going like, I think this field is right.
I'm putting it in there.
They don't.
It's got to be like a pop up right on the field on that screen or it's not getting looked at.
That's the thing where okay, maybe not generative ai, but even with some uh you know rule-based sort of automation, that would be an interesting way.
Okay, you have a self-serve data mart, here's the metrics we're going to give you.
But you know, if you, if it has a natural language interface or whatever, and you say what you're looking, that it's going to default to, this is what's defined, this is what you get, and if you say No, I want to see it a different way.
You know, the response can be fine, but that's not metric X, Y, Z or whatever.
Well, let's let's onboard your new metric.
We'll make sure we'll we'll make people really look at it, make sure it's absolutely not anywhere that, you know, just because you didn't find it doesn't mean it's not out there.
And then, if so, let's onboard that metric and put some rules behind it, get the lineage of where that's coming from, and it'll be part of the process.
Yeah.
So we had Dave's spreadsheets, and now we'll have Dave's KPI.
Well, this is not the official one.
This is Dave's.
So I think you and I could get I mean, we could go on data all day.
But I know another big area of interest to both of us, because data is the foundation of it.
But I'd love to talk about AI and productivity, obviously specifically around finance.
But I think we hear about all these failed AI projects and then conversely, we're seeing all the amazing things it can do.
But when we talked before the show, You grouped AI into two categories.
You were talking about productivity tools and innovation tools.
Give me your definition.
I mean, it's pretty self-explanatory, but give me your definition of each and tell me, based on and this is generative AI specific, I guess, where should companies realistically be focusing today with their generative AI usage?
So I, you know, I'll start with defining those a little bit more.
And, like you said, it's pretty self-explanatory.
But innovation is where AI is allowing some you know, some people to really hyper model things and get a lot more variables into the models.
So you can have even better projections, like even further out, five, 10 years out, if you want.
15 years, but I think that's a small subset.
It's the coolest part of AI, but I think it's the productivity gains that are really where most people are going to see the benefit.
Like I said, just simple as writing that e-mail to two levels above.
That'd take me a couple hours probably to get it right and get to a place where I felt good and look at things.
With AI, I could do that in five minutes and be fine with it and send it on its way.
One of the SQL queries.
When I first got in this group, one of the SQL queries I had to write was I had the base query, but I needed it.
I needed to have aggregations by week, by month, by quarter and then by half year.
So I could have done that, but I fed that into the AI and told it exactly what I just told you.
Hey, this is my base query.
Now I need you to summarize it by week, month, quarter, and year.
I want it all in the same table, so give me an identifier on what level of aggregation it is.
It spit out 200 lines of code or whatever.
When I looked through it, it was right in about a minute.
Some of these productivity saves will be an hour or two in a one-time deal.
Some of them will be repetitive, like something you do every week that takes you three hours.
You might be able to get it down to 30 minutes.
But that's the productivity gains, and that's where I think Initially, most companies are going to get that.
I'm in the financial sector.
I work in banking and stuff.
If you had a CEO on, they'd probably argue with me, but they're not in the tech business so much.
Maybe some of these fintechs are, but banks They're not.
They're about banking.
So if they can speed up their things and do more stuff or free up to do more interesting work, then that's really where I see the gains are.
And I have another good example, but I'll pause first to see if you have a follow-up.
I guess for me those productivity gains.
And I'm going to throw a question out, and I don't want you to think long on it, i just want to give me your gut feel on this, but i have this suspicion, and it's that people's ai usage at work is vastly underreported because of that secret cyborg thing that i referenced earlier.
I wonder um How many people do you think are out there?
I keep explaining.
With generative AI and especially as the tools get better, I've had some amazing results recently with Claude Cowork and some of the other tools as well automation tools like N8N.
But I feel like where we are with AI and automation right now, basically every employee who does digital work, you have your own RPA system and you can automate any digital process you do.
You can automate.
So the question, now that I've given you all that exposition, is what's your gut feel on what percentage of the workforce is using AI, whether above board or just on their own, using their own tools and everything to be more productive at work?
I think, to be more productive at work in general, even all the people out there.
I'd say personally 20 or 25 max.
So you think there's still a lot that are behind on the adoption card?
I think the AI is there to give all the savings.
I think the workforce isn't there.
I think the infrastructure, like you said, how many companies have you gone to?
They don't have a plan on how to use AI.
It sounded like a lot of them.
So you've got to get them to have that.
You can't just dump it on somebody's desk and expect them to do twice as much work next week.
Right.
They've got to understand it more.
And it's intimidating.
I mean, one of the things that tech people are bad about is thinking that everybody thinks like we do and they're like we do.
I went to school for computer science.
I'm not any smarter than anybody else, but I had a digital career.
There's so many people that went to college, didn't have anywhere close to math in their degree, you know, but they're still have a lot of value.
But when you talk about AI and things, it sounds robotic.
It sounds intimidating.
And so people just don't go there because they don't know what to do with it.
And then you have a lot of people didn't even go to college.
That Again, I'm not trying to say.
The college degree is there.
Some of the best people I've worked with have not had degrees, and I was shocked.
But it proves you don't need it, really.
So, to get back to your question, I just don't think the workforce is going to.
Unless company leadership can do a better job of guiding people down the path.
I don't think gains are going to be made just because the workforce isn't ready for it.
Is that fair?
Yeah, it's funny.
Thinking about just knowledge work and finance in general, and I get it.
I understand you have to have traceability and audibility and everything.
And just as you're working in data or working in finance, you have to be able to explain where the numbers came from.
You can't just say, I dumped this into the magic black box and it came out.
But there are so many ways that you can improve things.
And I think that so many people in FP&A... I mean, to me, FP&A, we're the original...
BI team.
We are the original business analysts.
And I think marketing got a little bit ahead of us for a while in the machine learning side because they just had so much more data.
If you think about it, traditional finance and accounting we're just talking about the general ledger, but modern finance and accounting goes well beyond that, into other systems and the data and whatever data we can.
You know I've always pushed for the CFO should equivalently be the CDO, the chief data officer, as well.
Now it depends on the skill set and the interest of the CFO or, you know, maybe data comes under them.
But I always think Well, if we're trusted to be the stewards of the financial data that is the lifeblood of the business well, is it the financial data or the overall data?
It depends on the business, I guess.
We're the impartial observer, because we're just looking for the truth as finance.
So then, why shouldn't we also, not singularly, not just with a heavy hand, but define the KPIs with the other groups, have access to the systems and be the reporters of it?
I mean, why is...
To me the financial data and the rest of the data in the business.
There's so many correlations and linkages between them.
I think I'm arguing to the right audience on FPA today about who should own the data, but at the same time I'm sure other groups would have their reasons why it doesn't belong in finance.
Maybe chief among those, well unlike you, who went to school for computer science.
We went to school for finance or accounting.
It came up that way.
So I get it from that perspective.
But I think the role of the CFO and the role of finance is shifting.
And I don't know, your thoughts on where data should reside?
Functionally, I mean, in the company.
Well, no, everything you said, I agree with.
You can make arguments either way.
I don't know if I have a thought on where it should reside.
I mean ideally in the assembly line model you have the data people over here that they just manage the data.
Is it coming in right?
Is it getting this?
Then the business people define it, but then they work really closely together.
But I've just always seen companies.
I've worked mostly in big, big banking organizations and I've seen there's a constant ebb and flow of centralized and then moving out to decentralize.
Because, following the argument you said that the CFO and the data person, they should have the same skill set and same knowledge.
Well hey, let's just make one team that has all the finance people and all the data people.
We'll put them together and then they'll they'll run everything.
Well, then you got this big organization.
It's like, it's great, it's all together, like there's so many people can help all the groups.
But then inevitably, some of those groups, they get somebody who doesn't know their data to work with them and it's frustrating.
They're like well, we need to hire a couple of data people in our group now and another group will do that.
It pulls apart that centralized data management scheme and goes out into silos again.
And then they get into silos, and the process restarts.
So I'm sorry I officially didn't answer your question on where I think it should reside.
I get it.
And you know, I think that I was – so when I went back to school a decade or however long ago, for analytics just to – because I felt like that's where finance was going.
One of the case studies I read, and this really is the model that I like, was the embedded model.
And in finance, we talk a lot about business partnering.
And to me, whether it's data or the finance person, if you have an embedded person in each business department division geography, whatever the case is, but that works directly with that team and a true business partnering.
But maybe they all roll up to a more technical understanding, to a CDO or CIO whatever the role is that can give them that technical support.
But they're embedded in the teams and really they need to be a full part of that team.
But a lot of times the team isn't going to have the skill set, so they go back up to the technical lead for that part of it.
But then IT is the plumbing, data science is the algorithms and the understanding of how to get the machine learning and everything to work.
But then when you are actually embedded with the teams, you understand better what their needs are.
I think maybe that's a good data structure.
There's a position out there.
This is funny because this time, and everybody watching, Glenn and I didn't plan this.
It's just happening now in real time.
What you just described is a position of a fully functional, mature data governance team.
It's called the data steward.
If you've heard the role, but basically the data stewards, I think in most cases they roll up to the central data organization but they are kind of.
A data steward is typically embedded in anywhere from one to like three business units.
And they're responsible for all the interaction with the data team.
And when they're trying to come up with a new thing, they're that person that goes in there and checks to see if that metric's already defined.
If not, it kind of helps them work through that.
Again, it's a really good role and it's really needed, but I've seen companies scrimp on that role because it's basically just it's like that guy in Office Space.
I talk to the data people so that the customers don't have to talk to the data people.
Yep, yep.
This is FPNA today.
So FPNA is our primary focus here.
However, as the listeners know, I'm very passionate about the data and understanding the data side of it and the skills that we need to have.
And I love someone on my FPNA team to be able to write Python and understand statistical modeling and write SQL queries and be really ultimately, whatever tool you use, Great.
But ultimately, I'm looking to answer, what is the way to get the most complete answer?
If it's all in Excel, great.
Excel is a wonderful tool.
If it's using RStudio or whatever.
Whatever it takes, be able to explain it, be able to explain why it matters, understanding what goes on.
But now for FBA people and I think this is just my own personal mission on this around data.
But if you were talking to someone since the background you have in working with finance, someone starting out in their finance career, how do you see the balance shifting, especially with AI and vibe coding and being able to clean and write code very quickly?
How do you see the balance shifting between the technical skills and whether that is Excel SQL Python whatever Tableau, Power BI, whatever it is?
And then Just this past week, we've seen, with all the new cloud co-work and connectors, the software crash in the market.
What Satya Nadella said a couple of years ago that SaaS has a real problem because It's just a wrapper on top of databases.
And if generative AI can go in there, what is the future of SaaS?
So this is the part where you and I have been doing this a long time.
Someone who's just coming into the field, where do they need to be focused right now?
That's a tough question.
Because part of my feeling is I do the vibe coding and stuff and and I use AI for a lot, but I still contend that I came up with the idea.
I'm coming up, I'm shaping what I want this to be.
There's different levels of detail you can give.
Some AIs are getting really good.
You can have a very small amount of detail.
Some you still have to fill in a little bit more, but you still need to understand what you're trying to do and why.
In finance, you're trying to optimize the flow of cash going around.
You're looking at your cash flow statement, your income statements, your balance sheets.
They have to be right.
That's kind of the higher goal.
So if you don't understand the purpose of a balance sheet, Then when you go ask the AI, you know hey, here's a bunch of data I got from finance.
Build me a balance sheet off of it.
That's not going to turn out good.
You know it may look right when it comes out, but you have to know some of the guidelines and some of the.
You know some of the legal things you have to pay attention to, as well as some of the corporate policies in whatever organization you're with, to be able to guide that AI to help you.
So I still think right now me personally that you still do need some of that core subject matter knowledge to be able to effectively use AI.
Yeah, that domain expertise, and I think that too.
Even in domain expertise in your field.
But if you think about AI and how the models are trained yes, I always say they've read the entire public Internet, But what they don't know is the proprietary information about your company, your processes, your procedures and the technology, the full setup there.
So the way that you're going to move ahead is...
If you're a deep level expert on your domain, and oh, by the way, you can use these other tools.
Like if you're vibe coding and you don't know the first thing about you know if then, or a conditional loop or whatever, whatever sort of thing you're coding.
And you don't, even even if you don't write perfect Python, if you have at least a familiarity with it.
When you see the code, and especially AI will put comments in the code and you can understand what each chunk is doing.
But then it's like you said, it's enough to be dangerous.
It's like Someone in that's not in finance.
If you ask them the difference between you know EBITDA and operating income or whatever you know, it's like you have to have that domain expertise.
And you know, if you're not in finance, you may not realize.
Oh, the balance sheet actually needs to balance between your assets and liabilities.
So, yes, I'm right there with you on that.
So I know we've talked a lot about data, but before we go, I know for years now you've been doing Excel videos and Excel courses and tips and tricks.
And I think that for our listeners, that would be a great thing for some of them to check out.
But will you tell me and we'll put a link to it if you've got it online or I don't know how people could access it.
But you had a pretty interesting Excel project that you did just for fun a few years ago.
Do you want to tell me about that?
The Minesweeper game?
Yeah, yeah.
And I know it's here in a couple of years now, but I still love that.
Yeah.
So, for those of you who may or may not know, this Minesweeper game came with Windows and it was kind of a little grid based game.
And, you know, I hadn't coded for a while and I still enjoy coding.
So I was going to try to go write this game in Microsoft Excel.
And this was I think this was a little over two years ago, because it was right when I started getting into AI.
And right before I started, I kind of had the flow in my mind of, you know, of sectioning it out.
Because when you write a computer program you kind of have to have an approach or like a flow chart on how the thing's going to go.
And I had that.
Now what if I see if I can do this in AI instead of me just writing, because it was all visual basic code.
And so I mean honestly, it would take me a lot shorter now, but it took me about three hours to make the complete Minesweeper game.
Like it looks like the real deal, you know, gray boxes in there and you click on it.
There's a bomb.
It explodes out.
I did it in six prompts.
Now, when I say six prompts, that sounds great.
Probably had about 20 or 25 through the course of the evening.
Cause like my first one, I wrote the first one was just to set up the board.
So basically say, Hey, you know, I want this board instead of resizing columns myself.
I just told AI to do everything.
I said, I want this.
It's going to be like a game.
So I want all the cells to be sized like squares.
So it resized the columns.
They're in proportion to the to the rows.
But then, when it had all the setup instructions done, I click go.
It spit back Python code because I hadn't specified do this in Visual Basic.
So again, that's that example.
Like, oh, but it's really neat.
Each prompt kind of did a different foster of the game.
And I did a webinar with a bunch of people who really thought that was cool too.
And at the end of it, this is just how you start getting into AI and how you realize how powerful it is.
Because I was really psyched that I wrote it in six prompts.
Like, that's amazing.
And one of the guys on my webinar when we were done said Why didn't you just ask it to write a Minesweeper game and see what happened?
We pulled up Claude and asked it to do that.
And I wrote the whole game.
That was like basically the prompt.
But, you know, as finance and data nerds, Excel is our preferred medium of communication.
So I admire that you did do it in Excel.
Yeah, anybody could do it.
Just, you know, writing a Python game.
No, I love that story.
And I do.
If there's a way that we could link it for our listeners to check it out, I'd love to do that.
Okay.
Um, wow.
Yeah.
And we are getting.
I've still got a few more questions, but we're we're getting close to the end of time.
So I'm going to start trying to bring this home.
So speed round.
We have a couple of questions that we ask all of our guests, and I'm going to spring them on you now.
So the first one is what is something that not many people know about you, something we couldn't learn just from your LinkedIn or your YouTube videos?
Okay, I'm really going out on a limb here sharing this, because I will get challenged on this a lot probably.
But in my younger days, when I was in my 20s, I was really into martial arts and I got to be a secondary black belt in Taekwondo.
Very cool.
Very cool.
You know, so I enjoyed it, but it's it's those days are past for me now.
Yeah, I did just have a honestly, I'm trying to remember which guest it was.
I just had a guest the other day that had a same martial arts, martial arts background.
And I thought that's I'd had a previous guest. who did improv comedy.
And I thought, wow, if you had both, you know, improv comedy, you learn to collaborate and be nice.
And, and, you know, and that's all fine until it doesn't work.
And then if you've got the martial arts, you can bring, bring the other side of the house.
Yeah.
So okay well, the last one, and this will be interesting because I know you spend so much time in Excel.
But what is your favorite Excel function and why?
So I saw the question primer earlier and I think I want to get really deep in this.
But if I'm just being honest, like it's the payment function PMT.
For those of you who don't know it, it's a very simple function, PMT, and it lets you calculate basically, you know, definitely it lets you calculate a fixed amount of payments over a period of time.
It lets you calculate a car or house payment.
And so you go to the dealership and they'll tell you hey, you can buy this car for seven years and it's going to only be 350 a month.
Payment will help you realize right there how much interest you're going to pay on that.
And then your house too.
I did a video on sorry, it's supposed to be speed round, but I did a video on how much more interest you pay with a 1 change in the mortgage rate on a 500000 house over 30 years.
People don't think of how much one percent of you, because you and I look at it like oh, it's extra percent, so I'm paying you six bucks instead of five.
Mortgages don't work that way.
I'm paying you like 40 more for another one percentage point.
So Yeah,
And I love going back to the old standbys as a favorite, because you know someone like you that spends so much time in Excel.
There are just those fundamentals that blew our mind back in the day the way AI is right now.
And payment was a great, you know, very early function in there.
And it is a way to think about it.
So, yeah, great, great answer.
Okay.
Well, in the interest of time and wrapping up the speed round, I just want one last thing.
How can our listeners connect with you, learn more about what you do, see the YouTube videos and all that?
I've got a YouTube channel.
It has a lot of Excel videos.
It has some SQL videos, has a couple of shorts.
And as I said, I'm going to be focusing a lot more on AI this year, too.
It's called Tom's Data World.
And I think we're going to provide a link.
And then I have a couple Udemy courses a couple are free and one is paid that we'll provide a link to too.
And the free ones are, you know, one's on pivot tables, one's on formulas and functions and then one's kind of comprehensive Excel.
And then just follow, you know, look up on LinkedIn and follow there.
Great.
Well Tom, I'm glad we finally got you on the show and I've really enjoyed having you on, so thanks again for coming.
Thanks, Glenn.
It's always a pleasure on the side.
