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 FPA Today.
Welcome back to FPA Today, where we talk about the latest trends and technologies shaping the world of financial planning and analysis.
I'm your host, Glenn Hopper.
And today, we have a very special guest who I think his life's mission is to make work easier for literally everyone he encounters.
With over 30 years of experience spanning accounting, finance, and consulting, Don Tomoff is not just an expert in his field.
He's a passionate advocate for using technology to drive efficiency and innovation.
Don is the founder of Invenio Advisors, a consulting firm, and a speaker and trainer on data and tech -related topics.
From his early adoption of chat GPT to his ongoing mission to educate finance professionals on the power of AI, Don has been leading the charge in helping organizations think differently and work smarter.
Today, we're going to tap into Don's wealth of knowledge and hear how generative AI can revolutionize the way we work in finance.
Don, welcome to the show.
Wow. I can just go now.
That's such a great intro.
And Glenn, just to the side, I always hear this, that I'm passionate about this.
And, you know, you are right.
I try to help people.
I always say I'm not really, but I am nuts about making work disappear.
Okay. And I always have been.
You and I share this in common.
It's that quote, and I think it's often misattributed to like Steve Jobs, but it is, if you want to find the most efficient way to do something, give the job to a lazy person.
And that's me. I'm like, if there's something that you're going to expect me to do every day, I'm going to spend 40 hours if I have to automating it and never do it again.
No, that's why you and I get along so well.
Yeah. Yeah. And our shared passion for generative AI, which we've got so much to cover here.
We've got to get into this.
So just a little background, Don and I have known each other a couple of years now.
We've done several speaking engagements before and Don brings the energy and I bring sort of the pedantic lecturing.
So Don, I'm going to count on you to keep the energy up in this conversation and I'll try.
That's hard. That's hard, but I'll try.
So I guess to start, you and I have been talking for years on this and you've been deeply involved with generative AI since it first became available.
Walk me through kind of what initially drew you to this technology and how it's evolved.
And it's funny, I'm saying evolved.
This has only been a couple of years since we were at like GPT 3 .5.
But walk me through your experience so far with.
I can remember in Glenn, it's December 5th, 2022.
And I can remember starting working with it.
And one of the things, and I'll refer to Seth Godin, who's a marketing guy that I follow and one of his blog posts is build a hundred hour asset.
If you invest a hundred hours into learning anything, you will have a skill that very few people will get to, even at that.
And as you know, I have a twin brother, Bill, who's a finance guy and we just started saying, okay, we're going to start tapping the pipe every day, at least 15 minutes.
And literally we'd get up.
I would get up and then first thing in the morning, all right, I'm going to try this.
I'm going to try this.
And as you know, it doesn't always work.
And especially back then, but it would just get things so far along that I was like, I could just see an explosion of opportunity, which has only gotten crazier since then.
And we both know Tom Hood.
And the one that I remember, it was probably mid December.
I sent Tom an email and I said, Tom, AI, let's not miss this one because there's so much opportunity.
And it's just gone crazy ever since then.
Yeah. And Tom at AICPA has been, I mean, he's all in on it too.
So the three of us get together and it's a, we're just nerd fat.
Yeah, exactly. It's a lot of fun.
And the other comment I was going to make is, as you were asking about, how's it evolved since then?
And I go, it really comes down to a mindset.
You're going to get bad answers.
You're going to get things that don't work.
Don't give up on it.
You just got to keep trying things.
And generally I always say, if it's not giving you the answer, you're right.
It's generally because I'm not giving it a prompt that is accurate enough or fully developed enough.
Yeah. And on that note, so I think one of the things that I see talking to finance people is, well, it's great for writing marketing copy.
But in finance, I can't just have black box things being created.
I don't know how I could possibly use this.
And I always want to tell them, you should meet my buddy Don.
I know you've tried so many things with generative AI.
Can you give us some examples of how you've implemented AI to streamline some tasks?
Yeah. Yeah. I mean, I spend, like most accountants, we spend the majority of our time in FPNA, which is your focus is no different.
We spend a lot of time with data and we spend a lot of time in Excel and Power BI and those tools and summarizing stuff.
So when I started using it, I said, okay, if I'm using Excel or I'm drafting a document, I would start with AI.
And there's a phrase that's become a little bit popular in the past month or so.
It's called an AI first mindset.
Okay. And I'm sure you do this.
If I'm thinking about something, I go right to AI.
I don't go to Google.
Probably don't go to Google.
I haven't gone to Google first, probably in a year and a half.
So it's making that a priority and just seeing what it returns and working with it.
So I always say, start where you are and people will start to focus on, okay, well, but it does bad things and it can hallucinate.
Yeah. But if you're doing things that you know the answers to, it's just going to get you there and go, yeah, that's it.
An example would be, I can have it write an Excel, a complicated Excel formula.
Well, if you know Excel, you'll quickly realize that's not the best way to do it.
Can it use the IFS function instead of a nested if function, things like that, you know, and you're not muscling through the syntax yourself.
That's probably the biggest thing that hit me immediately.
Yeah. You know, I love your approach to this and it reminds me a lot of, I'm sure you follow Connor Grennan and his, and he's got a new course out on it too, but it is...
Everybody should follow him, by the way.
Yeah. He's got a great approach to this.
And every time someone asks me about prompt engineering, I kind of have the same response.
I think that you would and that he would it's, I'm not going to just give you canned prompts that you just copy and paste in because that's not how you're going to get the most benefit.
That's like, that's handing you a fish and the AI is not always going to respond the same way.
So instead of just thinking, I'm going to copy and paste prompts in here, or I'm going to be this great prompt engineer.
Instead, it's understanding how to use it.
And like you said, I think that's great advice because whatever domain you're in, you're the expert in that, you know if it's on the right track or not, and then you can tweak it.
I'm not going to know the expertise that you have and what you're doing.
So if you're able to actually fine tune and tweak the prompting, then you're going to get better results that way.
And here's one thing I think I talked to a lot of CFO level people and people in the upper ranks.
And what I always tell them is, AI is not like, if you're putting in a new ERP or a new CRM system, it's not going to be a top down directive.
The way you're going to get the most results out of it, it's going to be bottom up.
You give it to the frontline workers, put the policies in place, let them understand data privacy, make sure you're in a protected environment.
But people are going to figure out, if you give them basic training on how to use this, they're going to figure out how to make themselves more efficient.
And it's going to spread that way rather than you telling them, use it this way, you know, coming from an on high.
Absolutely. And as soon as you start, I've had a few people at talks I've given said that what really triggered them, and I think this is a great idea, we get into this a little further, but they say that if they used it for personal reasons, planning vacation, doing whatever, and when they started
to see those dots connecting, they immediately went, okay, wait a minute.
I can do this for work.
And it started the momentum going.
And that's really the best way to, you know, do something you enjoy with it and see where it goes from there.
Yeah. Yeah. And I think at this point, and I know we have a running bet with someone at the AICPA about how many people are going to be exposed to this and using generative AI.
But I think that that's good advice is, okay, I understand day one, you're not on a paid plan, you're not in a protected environment or whatever.
Yeah, you're not dumping proprietary information in there, but you've at least got to go use it and understand the potential here.
And once the more you use it, the more you understand the potential.
And I think I'm now talking to companies more and more who are leaning in and they are giving their employees this these tools, and they're starting to see some pretty amazing results from it.
One example I saw to your formula comment at Topalti, they've leaned pretty far into AI.
And this guy who pretty good Excel guy, finance guy, had a report that it was going to take him an hour every day that he had to put together.
And he'd never done macros before or any kind of programming.
And he just thought, I'm not going to spend five hours a week doing this.
So he spent about because he didn't know what he was doing.
He said it took him 12 to 14 hours back and forth with I think he was using chat, GPT and Claude, but he figured out he'd never done any kind of programming figured out how to automate this report saved an hour a day that he had a report he was going to have to do through the end of the year.
And that is the perfect example of something that would have made mindless work.
He figured it out. And now though that he's done that, he's sharing it with other people in the organization, and they're figuring out ways to automate things.
So this is just, I mean, this is the biggest efficiency booster to come along since the calculator.
I don't know. Well, what you just mentioned, Glenn, and I, there's really three categories that I think of.
And I think you and I've talked about this before.
Everybody thinks of efficiency.
Okay, I'm doing what I do faster.
Okay. Yeah, that's a piece of it.
Another piece is because you can do things so much more efficiently and effectively, you start doing things that you wouldn't have done before.
And the example I use is we're documenting things that we really should have been documenting all along, Excel, Workbooks, processes, whatever it is.
Now it's easy to do.
So we do it. And then the story you just told goes to the top level, which is we're doing things that we could not do.
Okay, this person writing macros, that's my biggest win organizationally is I used to outsource automation of things that I would build for clients.
I haven't had to do that since January of 2023.
And it's enabled building some really sophisticated models.
Okay, which that's what is hidden.
There's an effectiveness there that is like, once you get the hang of tinkering with it, you start to put things together that will literally rock your world.
Yeah, thinking about it.
And as companies start to figure it out, and as the employees get good at their particular jobs, it's going to trickle upstream.
And you're going to start seeing an impact at a department level.
And I love that you mentioned documenting things because one, the former CFO in me is thinking about controls and thinking about socks and having more documentation around it.
That's huge. But then it all builds on itself.
So maybe the SOPs that you had before were lacking, but you have to create these SOPs so that you can feed it into the AI so that it automates.
But you also in turn get better SOPs.
And as the AI gets better, you can automate more.
So you can, if you take functions and you have all these SOPs across them, well, once you have the AI moving across, connecting to various systems through APIs, connecting to data through it, and you know what the process is, that's what you use to train the AI.
And then you get even more efficient.
So it just builds on itself.
That's exactly right.
It's hard to comprehend.
And we know this, but the speed that things happen at, we cannot conceive of until you get in and start doing it.
That just output and the quality is so much better.
And I like to say if you're a football fan, it gets you to the 20 yard line every time.
Yeah. We're talking about individual use cases, but how do you see generative AI transforming FPNA?
What specific areas have you seen, have you played around with where you can say, hey, this area is pretty ripe for disruption by AI?
Well, the big one to me, and this is something you've touched on, is analytics.
I think what you're doing is you no longer need to understand exactly how to clean data.
It'll explain to you what you can do, or if you can give it the data, it'll do it for you.
So I like to explain that the days of as a leader or a senior person handing off data to a team and saying, hey, I need this analysis.
We're going to quickly move away from that where preliminary analysis will be coming out of the senior level because they're just going to be able to explain what they want and get it done.
And there's people far more knowledgeable than me about integrating Python into advanced data analysis uses Python.
We're all going to be using Python and code via chat GPT and doing things that never before would we have even thought about doing.
So I mean, to me, that's a huge win.
And you look at that and you go, how does that ripple through an organization?
Everybody's going to have data skills and automation skills because you don't have to know the nuances necessarily of every single little piece that is getting done.
Yeah. And that's a great point.
And I always think of it like this.
So we've been talking about democratization of data for what, I don't know, 15 years or more now, but what AI gives us, what generative AI gives us is democratization of data science, because there's a lot of cool stuff you can do with data, but historically to do that, you had to know Python or R,
and you had to be able to write SQL queries or no power BI and all that.
But now, you can in natural...
It's like English or whatever your natural spoken languages becomes a new programming language, because you can say, build me a random forest machine learning algorithm that will determine if these loans are going to default or not.
So there's so many applications and you can do it now without having to wait on data engineers, because the data analytics tools will do it.
I'm really in a couple of my courses that I'm teaching on it, I'm really pushing the limits on it.
And loan approval is one that I just did in one of my courses.
Also, I just built one yesterday, trying to predict the likelihood of a recession based on 30 years of financial data.
So just grabbing stuff from the Federal Reserve and plugging it in and trying to predict recession.
And that's the kind of thing, I did that in a couple of hours, that would have taken me days in Excel.
I'm just thinking of the old days of doing that.
And you wouldn't have done it because we don't have the time.
One of the examples I use when I speak is I use a slide of a growth stock of Amazon and Netflix, because I'm making a point.
That entire chart was built just by feeding it stock activity by week for 20 years and asking chat GPT to build it.
So I didn't have to clean the data.
I didn't have to do anything.
As you know from doing the chart, you got to play with that a little bit to get it right.
But it's right there.
Literally in five minutes, I have a chart that I wouldn't have been able to do that.
Now I will say one of the things I asked it, and this is to the point of validating, I asked it to put a box on the chart showing me the annualized return over those 20 years.
And intuitively I said, yeah, those are right.
But I went to Excel and I proofed it.
Okay. I said, I'm not going to put this out there if I don't know for a fact that those are right, but it nailed it.
Yeah. So trust, but verify.
We're not turning over the keys to the kingdom to our new robot overlords, but we're leaning on them pretty hard, but we're going to check everything they do, especially if I'm a public company CFO.
I'm not signing anything that the robot put together for me yet.
Well, I always say don't trust and verify.
Yeah. Just don't even think it's right.
Which Glenn, you did this at our session back in December last year, it's the future of finance where we were kicking analysis on public filings.
Okay. And one of the things that you can do now, and I think this is an improvement because they've tied into the internet.
I can give it a link to a 10 K and have it summarize the MDNA for me.
Tell me what the risk factors are.
And when I came out of investor relations, it's unbelievable that you can just literally give it a link and say, you could look at 10 companies in a matter of minutes, knowing that you have to check the information it gives you.
Yeah. So I've gone a step further since then.
So I'm using automation tools where...
So instead of building the GPTs, you can build the behind the scenes, the assistance on open AI where it's basically GPT, but it's not...
You're still not coding, but you're just writing in more.
It's not a prompt back and forth, but you can put in documents and for that 10 K analysis, what I've done is created an army of bots.
And so some of them are experts at creating charts.
Some of them are experts on sort of the qualitative overview.
Some of them are experts at financial ratios.
And what I'll do is I've used a couple of them.
The most accessible is make .com.
It's just a drag and drop automation tool, but I've used Flowwise and some others that where you can use retrieval, augmented generation, where you can drop more documents in and have it referred to it.
Now it's just behind the scenes in chat GPT.
So it's actually you interface it through...
So you can put a web hook in.
So I've got a web interface and I can just upload a 10 K into my web hook and we're still testing.
So it's not publicly available yet, but when you put the 10 K in and all you do is click a button, it costs about a buck to run because I've got so many assistants in the background.
Yeah. And you're tying into the API to do this, obviously.
Yep. Yep. So it's passing through, but it takes that 10 K and it gives it to all these different bots.
And then they aggregate everything and assemble it.
So because you know how it is, if you ask too much, you could get a good analysis, but you've got to go back and forth and back and forth.
But this through some chain of thought prompting and through having these bots, you walk away for...
It takes about five minutes to run because it has to go through all of them.
Some of them are sequential.
Some of them can run together.
But then you get this report that, Don, it looks as good as what I would expect a human financial analyst output, and you get it in five minutes and it costs a buck.
And somehow I'm on the shortlist to get that.
I don't know when, but I am.
Yeah. Maybe we talk about that at Future Finance in December.
Yeah. No, that is what we talk about.
But anyway, what you said is I've taken it to another level, as usually you've taken it at about a hundred levels.
That's okay. I can't wait to actually see that work.
Yeah. I figured since I have AI R &D in my title, I better be pushing the limits.
Otherwise I don't get to use that title anymore.
What are we paying you for?
No, actually that's fabulous.
I can't wait to see it because when you look at just...
People worry about confidential information as they should, but when you're dealing with public company information, that's just a big sandbox.
You're going to learn a ton, like you're doing, just playing with that.
If you're an accountant, I always tell people, if you're an accountant and you're doing reporting, you should be looking at SEC filings.
Whether you're public, private, or whatever, and you're going to learn a ton and chat GPT has just made it insanely easy to do that.
Yeah. And to that end, if you wanted to build a GPT that was an expert at filings, what you would do is load up a hundred...
You'd have them from different industries or whatever, but here are a hundred different filings from these different companies.
And you don't have to train it and tell it what to do.
It's just got access to them.
And so you want to do your own filing.
It's got these in its knowledge base.
You ask it to start drafting a filing.
Now, again, you're not going to just take the human out of the loop and just turn it over, but you suddenly have a shell for your filing that is solid and you get it immediately.
So let me ask you this.
What you just described there, it sounds like a GPT on steroids, overdrive, whatever.
I mean, because you're really limited in a GPT to just upload like 10 documents.
Yeah, good point there.
But with the assistance and with something called retrieval augmented generation, where you can actually put these external documents in and have it referred to them.
And then you can also use these documents to fine tune.
So now chat GPT just announced that you can fine tune 40.
And it's, again, you don't have to be a programmer, but you take all these documents, you fine tune it, then you use the retrieval augmented generation.
And you've got not just the generalist, you've got a subject matter expert on whatever you're tuning it for.
But this is to your point, to be fair, it's not just the GPTs you create, but anyone who has the paid account can go into the playground at openAI and make it.
But then it's asking a little more, you have to go and be able to have the API that goes into it in a way to call it an output and all that.
And I want to make sure that you jot down a note right now that we're going to set up a call for you to show me what you're talking about.
Okay, we'll do it, we'll do it.
And so you know what, Don, what we've done is we've gone too deep, though.
And you are the master of taking the fear away from people.
So actually, I want to get out of that nerd level.
And just in case we've lost anybody, stick around for one more second because Don's going to make this approachable and accessible here.
And he lost me too.
So that you guys are all good.
So to that end, what advice do you have for finance people who have you know, they've heard about this, maybe they've done some things on the personal side.
But if they want to start, they want to upskill, they want to start thinking about using this, what advice do you have for people who aren't out there building their own assistants?
That's a great question.
I always say, one, and this is not very successfully do I get this, but I always say get the paid plan.
Okay, I that's just a given.
Okay, there's just that big a difference between the free and the and the and everybody always asks what plans I pay for, I pay for Claude and chat GPT.
Those are the two. Okay, and I also use perplexity a lot and some of the other ones but get the paid plan.
Okay, let's assume you don't What else would I recommend?
Go mobile. Put these things on your mobile devices, make it so that they're right there that I can open it up and ask questions.
And I'm sure you've done this Glenn, but being able to talk to the app of chat GPT, when I'm in my car or out on a walk is absolutely insane, what you can get done at any time.
And then I say I talked about this earlier, but start just start at everything starts with just starting, and then don't stop.
Okay, it's just you've got to keep stick with it.
Too thick or thin, you're going to have some wins, you're going to have some failures, just stick with it.
And then look for the easy wins.
And then finally, collaboration, find a group of people that you know, kind of trying to figure this out and do it together.
It's your organization, whether, you know, Glenn and I are a perfect example and Tom hood, we, we trade ideas all the time.
Okay, and it's just making sure you stick with it.
And I know we're going to talk about this, but it's changing.
It has changed so much from when we started to today, you really just got to start running on the treadmill and keeping up with it.
And I love the collaboration and networking part of that too, because there are so many groups out there, there's, I'm involved with the chief executive group, there's a AI connect that they have, where we do monthly calls, there's a, you know, interaction going on between the community, we do a bunch
of tips and demos, there's a Nicholas Boucher has the AI finance club, I mean, there's a million groups out there.
And really, the best learning comes from everybody's out there trying it, they're coming up with new use, it goes back to my idea of it's not top down, it's coming from the community itself.
And this is a really unique technology in that way.
And you described the person who developed an automated program, you know, on a much smaller scale, you can automate things in Excel in like two minutes, never haven't written code before.
You know, you start with that, you'll start to feel what this can do.
And it will explain, and this is one of my favorite features, Glenn, I don't know if you do this or not, but when I'm doing training, some people say, Well, okay, how do you do that?
And I always ask if they're with their laptops, have your laptops open, okay, ask chat GPT, how do we do this?
Okay, ask it, how do I get it to open to a specific page in an Excel workbook?
Oh, that's the auto work, there's the workbook open VBA.
That's exactly right.
You don't have to explain it, because the answer is right, immediately available, which one of my favorite expressions, Glenn, which when I talk to people, I go, not having the answer is no longer acceptable, because you are 30 seconds from having the answer, you've just got to ask.
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 and 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 1000 finance teams use data rails to uncover their company's real story.
Don't replace Excel, embrace Excel.
Learn more at data rails .com.
You know, another thing that I think about is we're talking to finance and accounting people who are inherently risk averse.
And you know, they think this all sounds great, but they're just, this is unknown to me, it's a it's a black box.
I don't know what sort of voodoo magic is going on.
I'm not touching it.
So, you know, you can you can give the demos and you can talk about it.
And I think you and I probably see a lot of the same challenges, but also opportunities.
So maybe talk to me about when you go out and you talk to these groups of finance and accounting professionals, what are some challenges you've seen when talking to them?
And I don't want to just keep it negative because I think we're also very smart people.
And there's also kind of a positive side.
There's that moment, I love it when you see the light kind of go on with people in these demos sometimes.
But walk me through maybe some of the challenges and opportunities you see just from talking to the groups you do.
I recently did a session.
This kind of ties into what you talked about.
I love it when the light bulb goes off.
I did a session and I do, I repeatedly say, if you're not on the plus plan, get on it, okay.
And one of the young guys came up after the session.
He says, do you collect an affiliate fee for recommending Chachi Piti?
I go, no, I don't. I go, but I get a fee.
It's called a feel good fee.
I feel good whenever somebody adopts it and that light bulb goes off.
So mindset's the biggest thing.
You've got to realize that we don't like change, but if I'm not doing this and Glenn, I'd love your opinion on this.
I mean, right now if you're doing AI, you have a huge advantage against those that aren't, okay.
If you're integrating it into your workflow.
But I think in the next year or so, you're going to quickly shift to it's a liability for you.
If you aren't doing it.
Oh, totally. It's going to be expected that you're going to be able to augment your work and be able to go to Chachi Piti to figure things out.
In fact, Glenn, I got one quick story I want to share with you.
I was doing a panel at a conference and one of the people on the panel, this is an accounting firm.
He was talking about an intern.
He was talking to an intern and he said, well, here's what we need to do.
We need to take the information.
He said, I don't know how we're going to do this.
We got to get this done.
So this gal disappeared, comes back in 15 minutes.
And Daniel was this guy's name.
She goes, Daniel, I was able to get Chachi Piti to write a macro to automate almost all of the data retrieval.
This is an intern. And I go, okay, did you hire her on the spot?
Because everything she demonstrated right there, that's what you're looking for.
Initiative, thought process.
How can I do this better?
And that's the level that thanks to you talking about bottoms up.
That's where it's happening.
Okay, you give them a challenge and they're like you and I going, well, I'm not going to go sit for a day and try and finger this stuff into a spreadsheet.
I'm going to see if I can do it easier.
Okay. So mindset, there is no easy answer, okay?
And accountants and I'm even closer to the accounting world than you are.
I'm a CPA and I've been in the shoes that they're in.
It's hard. It's really hard because you don't have time.
I just talked to a CFO and he said, man, I wish I had time to figure this stuff out.
I go, nobody has time.
You got to find the time.
I mean, I don't think you woke up Glenn back before you started to write your first book when all this stuff wasn't even like on anybody's radar.
I don't think you woke up and said, I got a year to fill.
So I'm going to write this book.
But on the other hand, and I think the thing of having to figure something out is daunting to people.
It's just, I'm not sure.
I don't know where to start, etc.
So on and so forth.
And you mentioned Connor Grenin.
It's one of the reasons I love following him.
He's got a very practical thought process to explain how to approach these things.
But you talked about the opportunities.
If you're a knowledge worker, and we're focusing on FP &A, but just in general, if you're a knowledge worker, this is a career changer.
It's going to just open up so many opportunities and doors for you.
Yeah, agreed. And it goes back to that, how are we going to get more efficient?
And I don't think anybody...
You don't go get a master's in accounting or finance because you want to be a data entry person.
And I think young in your career, it is fun putting together really cool models in Excel.
But as you progress through your career, and you want to broaden what you're doing and the value that you're adding beyond just building a model, then it becomes about interpreting and driving and understanding what the assumptions are in the model and all that.
And if your whole team is spending their time focused on that, rather than on building out the massive nested ifs.
And honestly, it's still fun Don.
It's still fun to do stuff in Excel.
Well, it's even more fun if you do it in 10 % of the time that you used to do it.
Exactly. Exactly. Yeah.
But you just made a point.
It just slipped away from me, but you made a point that...
And this is where I come from, and both of us, Glenn, are senior level finance people.
So we're used to working with the staff people and the managers.
And I always said, if you're a CFO, if you're a senior FP &A person, you need to have an awareness of what's capable, what the possibilities are, so that you know what your team should be understanding and integrating.
Okay? Like, away from AI for a second, if you're using Excel, your team's working with data.
And if they're working with data, they have got to know power query and power pivot.
Okay? They got to understand data transformation, and they got to understand data modeling.
But if there's a leader, we can't point that out.
It's not a priority to them to learn it.
Yeah, very well said.
I don't know how often you do these a lot.
It's almost hard to keep up.
But I'm doing very well.
Yeah. But the tips you give on LinkedIn, they're always so practical.
And also, you have a better skill of, whenever I try to write something, I just ramble and blow it up.
You give these quick hit, fast tips that you can get in 30 seconds.
I think they're all Twins Talk tips.
Talk me through that.
And you're at 1 ,300, 1 ,400 of these tips that you've done now?
1 ,327. Okay, Glenn, I'm glad you brought that up.
Twins Talk with a Z on LinkedIn, on Twitter, different places.
It's me and my twin brother.
I tend to focus on the, what I call practical tips.
But one of the things I've really started doing, because it struck me the other day, I don't know if you've ever heard of the two minute rule.
Okay, the two minute rule is, this is a David Allen, whose productivity book years ago, and he says, if it takes less than two minutes to do something, do it right then.
Okay. Don't put it off.
You'll waste more time thinking about doing it later.
And I said, well, with AI, so many things have moved into the two minute task.
Okay, and that's always been a mantra of mine.
If I can get this done in two minutes, so how do I do that?
Well, I get digital information instead of paper, I convert everything to electronic.
I have search capabilities instead of having to click on folders, etc.
But when you start using AI, and you go, which you quickly realize that things now become two minute tasks.
And that's really, I started adding that hashtag because when I do something, and I did this one the other day, you may have seen it, but I was driving home from the east coast, and I'm going to be doing a power query session.
So I just go into Whisper, which is the transcription voice part of GPT app.
I've been driving, I said, I'd like 20 or 10 power query transformations.
Explain what, why and how for each one, I'd like 10 more.
Done it, okay. Explain, give me, well, no, at that point, I said, I would like you, and this is talking to it, I would like you to export an Excel file with a grid and columns for what, why and how on my phone while I'm driving.
And that was a two to two to three minute task.
And I go, you don't really think about how fast these, well, you don't know how fast these things can get done.
But the voice and whisper is like, crazy.
Okay, we talked about personal use, you know, you want to, you want to get into that.
So I focus, at least I hope we do, we tend to focus on things that people are going to be able to immediately use.
And you can tell if you, but by the way, Glenn, if you talk AI, you talk Excel, those posts tend to resonate really well.
If I talk about knowledge management, which we hit a little bit before that.
And why do we do that?
Because if I learn it, or you learn it, Glenn, you're the same way, hey, I can help somebody, it's not hard to just share that.
Okay, that's a, that's passion.
You talk about a passion of helping people do things better.
That's probably where it shines through the most.
And the only other ones that we talked about this, Glenn, but you talked about the prompt engineering.
I really like to recommend that people lean into the custom GPTs.
Yeah, and we haven't, we haven't talked, I know you and I talk a lot about it.
And we and I took us off down the road with the assistance, but custom GPTs, yeah, actually fills in a little bit on that.
It's, you know, I've written, not as many as you, but I only have probably three or four that I share, but I've written 20 of them.
And yeah, it does take a little bit of figuring out.
Sure. Okay. But as an organization, I could easily see these things being everywhere in an organization.
And I don't know if you're familiar with the Moderna white paper, Moderna has 750 GPTs that people have developed.
And I go, this is real.
Okay. And they do anything.
But the beauty of them for folks that don't know is it eliminates the need to do prompting.
So I have one that's an Excel formula.
I call it the Excel formula dominator.
You're no longer thinking about how to write formulas.
You're asking it what to do, and it's doing it.
I was talking to a client the other day, and I said, we're doing formulas and functions.
And I said, pick a form, pick a function that you don't understand.
And we had a list of all these functions.
They said indirect.
I said, okay, let's ask chat GPT using this GPT to explain that function.
It gives me a full, nice, laid out function.
Then I go, now create an Excel, create an example and export, give me the example in an Excel workbook.
And within two minutes, I had a simple example of the indirect function in an Excel workbook that I could share with somebody.
That actually begs another question, because I have used GPT to write formulas.
And it's been a while since I've done it, because I'm not in Excel as much.
I'm more in programming stuff.
But when I used to do it, and maybe there's a better way now, if I wanted to be able to just copy and paste the formula, I would say, okay, I have a table.
It's in C5 to F316 or whatever, because I want to do, I don't know, a VLOOKUP or whatever I'm doing.
I want it to have the exact right cells.
All I have to do is copy and paste.
But it takes a while to say, so how do you have it set up so that it's giving you actually the right cell references when you're doing the formula?
I think that's a complete setup question, but it's good.
I would use vision.
Take a screenshot so it can see the columns and the rows, and it will figure it out.
That's awesome. There's so many things to talk about.
And again, that's the paid plan, which I'm sure you're aware, Glenn, but everybody has access to the full capabilities now.
But if you give somebody a GPT, they'll get like three uses and it'll go, okay, you're done.
Okay, that's nice. You tasted it.
Now you have to pay if you want more.
But you're right, the vision.
And I'll tend to ask it generically how to do something before I get into a real formula.
How do I have a lookup of value?
I have dates across the month across stop and products down the side.
And it'll tell you that an index match will do it.
And you go, okay. And this is where people are concerned that these are going to make a sleazy, etc.
I go, always take that extra step to say, explain to me what's happening here.
And it will do it perfectly.
Okay, so I mean, your observation there is a great example because if you're trying to get a real precise formula, you do have to set that prompt up pretty good.
I love the vision idea.
Okay, okay. So that's good advice.
I know it likes to deal with CSVs just because there are smaller files than Excel.
Have you tried uploading the Excel file and then telling it what you're trying to do and see if it can reference that?
And so does that work as well?
Yes, it can. In fact, what I primarily do, not to write formulas, I don't do that, but I'll take an Excel file and say, explain the contents in here.
And what's the best way to organize this data for data analysis purposes?
You know, if you're a senior person, and you didn't want to do it, but you wanted to analyze the data, it'll actually do it for you and export it to an Excel file so you could then do the analysis.
But more than anything, and then Glenn, you know, when you have to do analysis, you got to structure your data a certain way.
And it's not the way accounts think.
It just isn't. We think reports, dates across the top, etc.
But it will explain to you how you need to structure it, why you need to structure it that way, and it'll help you do it.
I love that you said that, because I just went through this yesterday, I was explaining it to someone.
So we, you know, dates are the columns and then the accounts are the rows.
But if you're working in machine learning, it is the accounts are features in the model.
And then the rows are observations.
So every account is a feature.
So that's a column.
And then the rows, every time I put finance and I have to tell it to transpose, if I'm trying to get something back out, I have to transpose it back because you know, to the machine learning model, the observations are the rows and the observations are the months.
So it always transposes those financial statements.
And then the other thing it defaults to is instead of doing what we like, where the oldest date is on the left, so if you're plotting like revenue growth or whatever, it always does it backwards.
So it looks like the company's tanking when actually they're going in the other direction.
So there's a lot of that back and forth.
And what you're describing there, these are kind of the nuances of data analysis that nobody thinks about.
And it takes time. Once you kind of figure it out, you go, hey, wait a minute, I can ask it.
And I don't think you're not an M code or power query guy much, are you?
No, no. Okay. Because you can actually have chat GPT write you that M code.
Okay. And for a guy like me, you're getting away from the interface and saying, I would like you to create the M code that will do this.
And then you just tweak it to get it to work.
It's mind numbing. So one more on, because I saw this on one of your twins talks posts recently.
You know, you and I try all these, we've banged around with, you know, Gemini and Llama, chat GPT, Claude, and I always keep going back.
I don't want to be a homer for open AI, but man, chat GPT does so much.
But we were talking before we started recording about Claude, want me through what you like about Claude and maybe some use cases because I do, I also have the, I have the paid, you know, the team account in for open AI.
And then I've got the pro account and Claude, and I do love the artifacts tool.
I just, I don't use it as much.
So what, what me through some of the use cases I think the distinction between for me is if I'm summarizing documents and if it's a large document, like let's say a couple hundred pages or, you know, let's say it could be a hundred pages and I'm thinking like, you know, McKinsey reports, stuff like
that. That's, that's dense.
I will defer to Claude to do that analysis.
Now, Claude doesn't have access to the internet, so it doesn't get any of that right.
But it does just a phenomenal job at going through large documents that are kinda, I won't say beyond the scope.
It just does better than chat GPT.
On the other hand, if you're trying to get links or you're trying to pull things out of a document that are links, it doesn't do very well.
So I go to chat GPT and it's finding what works best.
Okay. And, and tinkering with all these different models and going, okay.
Like if you're writing a document Glenn, and I think you described this before we got going, if you're writing a document, Claude is much better at if you feed it documents that you've written and saying, okay, write this in the tone of Glenn Hopper, it's better than chat GPT.
Okay. So that's another big advantage.
And we talked about artifacts.
The fact that I can create a document right in Claude and then share just that document via the artifact is super handy for me.
It's like, Oh boy. Yeah, that's great.
The last thing I would say on that is, and I don't know if you've seen this, but in Claude Sonnet the 3 .5 is their best model right now, but the Opus 3 is better at writing and it's better at long tasks.
So I don't know if you've, if you've seen that difference.
Whenever I'm writing something, the phrase that I use most often is rewrite for clarity and I'll plug in like my mess of something.
And Opus does a really good job with it.
I'll just take, you know, a couple of sentences or be like, I need a transition between these and it'll, it'll come up with something.
But I also noticed when you do that, it picks up the AI detector will be like, ah, this is written by AI.
And it's like, it was edited by AI.
And, and it was a lot faster than me banging around with it.
Bingo. That's what I say.
So, all right. Well, we really appreciate the tips.
And if anybody's not following Don on LinkedIn, they really are.
They're, they're great.
They're practical and they're quick reads, quick hits and stuff you can use every day.
That's the key. They're quick.
You know, you and I are obviously power users where we, we notice every time a model is updated slightly and we use it enough to see that.
But I think, you know, even us, I'm looking around my desk here.
I don't have my crystal ball, but maybe you have yours handy.
How fast all this technology has gone?
Like, where are we in five years?
What, when you and I are revisiting on the podcast, what are we talking about then?
Particularly using in finance or maybe just globally, what do you think?
Well, just, just in general.
And I think this is not our opinion.
It's what we see and it's out there as practically every job is going to be augmented by AI.
Okay. It's just, it's just going to be.
So understanding how you can take advantage of that today is absolutely critical.
Getting, getting comfortable saying, okay, how do I use this prompted, et cetera, to get it to do things that are helpful to me is key and it takes practice.
Again, back to that hundred hour rule and back to quoting Tom hood.
One of his favorite expressions is we may not be able to learn everything, but we can learn to ride the wave.
Okay. And that's really the key.
You've got to just do what you can to try to keep up.
And Glenn, I listened to some of what we talk about and I go, I just can't even imagine getting there because it's hard enough to keep up with just the things that I'm trying to focus on.
Okay. It's like, yeah, I'll see a, here are a hundred top apps.
Okay. Whatever. I got like two or three.
Okay. Unless Glenn Hopper says you need to look at this.
I'm not going to be looking at something else.
Yeah. And I'm focused.
One of my favorite expressions is narrow the path.
Make it easy for you to stick with it.
Okay. So if it's chat, GPT and Claude, okay, I'm going to start using them or if it's copilot Gemini, whatever it is in your, in your workflow.
Start with that. Great points on that.
And, you know, there's a couple other things that you and I talked about before.
I wanted to talk about memory, but I think in the interest of time, we're going to follow Don on LinkedIn and you can learn about how to make use of a memory capability on chat GPT.
Well, we'll give them a little something to look for there.
It's really nice. Cool.
Donna, you know, we've, we've been talking for years, so I know a little bit about your background, but, you know, we went straight into nerd stuff.
We forget there's a human element to everybody here.
So maybe for our, for your followers on LinkedIn and the listeners on the show, what's something that maybe, people don't know about you?
Something that they couldn't just learn by looking on Google or asking chat GPT?
Yeah. Well, that that's a good question.
Actually, a fun fact is I got two for you.
I was a CFO with the Cleveland at the Rock and Roll Hall of Fame in Cleveland for a year.
That's cool. Okay. For a year.
That was a, a little break in between two of my public entity jobs, which was really fun.
Some of the things we got to experience.
But fun fact, my, both my brother and I were competitive college distance runners, competitive marathoners for a number of years and both two time Boston finishers.
So, you know, we, we were serious back in our younger day on the, on the running front.
That's awesome. So you two, uh, raising each other a lot too, huh?
Oh yeah. In our best effort, he beat me by 30 seconds over a marathon over 26 miles.
It's like, really? You think he could wait for me, but so do you, uh, do you still do marathons?
No, no, I barely run.
As I like to say, I have no place to put the pain anymore.
Yeah. But I appreciate those, those efforts.
Let me tell you, I was watching the Olympics and I go, these people are, that's crazy.
I rode my bike the other day and it was about 20, it was like 26 .1 miles.
So I was just shy of the 26 .2 and I looked down at my time and it was like an hour and 32 minutes.
And I thought I was out there moving.
I felt, I mean, I was, you know, it's 16 miles an hour or so.
And I was thinking the world's fastest marathoners are, you know, they're not right there with me, but they're done in 30 minutes, like running this whole distance that I just rode my bike.
Two hours and eight seconds is the, and here's a fun fact for, for anybody that's somewhat familiar with distance running, when he, when he set that world record, I believe in mile 24, he dropped a 418 on the field, a 418 mile.
And you just think about running a hundred yards at that pace.
It's like, forget it.
Yeah. So pretty, pretty amazing stuff.
I haven't done a marathon in a couple of years, but I'm in recent years, I've been more of the, if I get under four hours, that is a win.
You know what? That's impressive.
Good for you. It's funny because I've had a lot of I've had some AI guys and some data guys, and I had an author on the other week.
And for some people, they really don't like this question.
Like one of them said, I know you ask everybody about Excel.
I don't want to answer that, but I know you're going to love this question.
And I bet you have a really good answer or two, but I got an answer.
That's for sure. What is your favorite Excel function and why?
Okay. I would tell you two things, not function, but capability would be power query.
So that's the data transformation piece.
Absolutely. The, the biggest change in Excel in the last 10 years.
Okay. Just crazy what it allows you.
You know, we're no longer working in access like we used to back in the 2000s function.
I would say that the unique dynamic array function is my favorite.
And when I use it primarily for, and I get the most benefit out of it is I'm big on creating data validation dropdown boxes and using dynamic arrays or the unique function makes it really easy.
If I had to pick one, there's more than one, but if I had to pick one, that would be it.
Love it. And honestly, so I'm, you know, I was a, an FPNA today listener long before I was the FPNA today host.
And I think that might be the first time I've heard unique, but I get it.
I get it. Really? Wow.
All right. Yeah. So now though, now though, I feel like I need to, um, take one of your Excel courses so I can expand my knowledge, but it's cause I'm not spending near as much time as I used to in Excel these days.
I know. But you give me the course on, uh, rag and then we'll talk this.
We'll talk Excel. Perfect.
We'll do it. We'll do a trade.
Yep. Yeah. Hey, Glenn, this was, this was fantastic.
Thank you for the opportunity.
I love talking about this topic.
And every time we talk, you school me on something, it's been fantastic.
Likewise, Don and I, you know, and on, and to that end for our listeners who, who have gotten, uh, their first drink from the Don Tomoff firehose of knowledge, um, what's the best place for, uh, for people to connect with you and learn more?
LinkedIn, that's, that's where I'm most active.
You mentioned medium and I do, am I, I am on medium, less active than I used to be, but like this best place to find me and I share, we do share the twins talk tips.
So there's a steady diet of hopefully helpful insight for people.
Excellent. Excellent.
Well, Don, thank you again for coming on.
Thank you, Glenn.
