This special edition of FP&A Today sees 1,524 finance pros join a LinkedIn Live session. In this episode Paul and a special panel answer all your burning questions on how Finance teams are practically using AI to advance their careers (and bring instant productivity to their businesses).
Joining Paul is Adam Shilton, Founder, Tech for Finance and a world expert on “Helping finance pros turn systems into superpowers with AI”. In addition he is joined by Sloane Kolt, Head of Datarails Labs, which recently launched FP&A Genius, an AI powered solution transforming FP&A.
Some of your big audience questions answered:
- What was the last thing you used Generative AI for?
- Ethical questions in AI and Finance (eg. Do we own the model that AI built?)
- What are the most common use cases for AI in FP&A, and accounting post ChatGPT?
- What are the differences between AI in big vs small companies?
- How should finance leadership use AI and adapt it in their team ethos and processes?
- Privacy, security, AI and Finance – what are the risks?
- What are the secrets of effective Prompting in finance?
- Clarifying tasks with your AI Chat
- Lessons from building a new finance AI toolÂ
- Favorite non ChatGPT AI tools you have seen
- Best advice for getting started with AI in finance
- Rapid fire questions: favorite Excel function and person I would most like to meet.
FP&A Today is brought to you by Datarails,the AI-powered financial planning and analysis platform. Keep your own Excel financial models and spreadsheets and benefit from AI for data consolidation, reporting and planning.
Follow Paul Barnhurst on LinkedIn
Paul Barnhurst:
So I just wanna take a minute and welcome all of you to the special LinkedIn live episode of FP&A Today. As you all know, my name’s Paul Barnhurst. I host FP&A Today, I am lucky enough to get to talk to great people all over the world. Today’s episode’s gonna be focused on AI and how finance can prepare for and think about AI. If you haven’t heard about AI at this point, just go spend a few minutes on LinkedIn and you can bring yourself up to speed. So I’m thrilled to welcome our two guests with us. You could see their name on the screen there, but we have Adam Shilton with us. And remind me, London, right? You’re coming from
Adam Shilton:
Middle of England, so Doncaster.
Paul Barnhurst:
Okay. I knew it was England. So, yes. There you go. And then we have Sloan Colt. And where are you from again? Sloan? New York City. New York City. All right. Yes. I know you guys have a few headquarters. They have a headquarters in Israel and New York. So we have Sloane and Adam here with us. And what I’m gonna do is I’m just gonna start off by giving each of them a minute to introduce themselves. So, Adam, why don’t we give you the chance first to just tell us a little bit about your yourself?
Adam Shilton:
Yeah, certainly. And, and thanks again for, for having me on. Absolute pleasure to, to be on FP&A Today. So, so thanks. So yeah, I mean, I, I first got into to technology, I think back in 2014 now. So my background is finance software and ERP, so coming up to, I think about nine years. During that time, I’ve also done quite a bit in, in the business intelligence and data space as well. And with the, the companies that I’ve worked for, I’ve been quite lucky to work across industry as well. So manufacturing energy, agriculture, professional service, nonprofits before I got sucked into the, the rabbit hole of AI last year, I, I was a little bit late actually, because of course ChatGPT I think was released in November, wasn’t it? So yeah, I was a month late and unfortunately lost my, lost my Christmas to it last year. So bit, bit of a telling off from, from the family, stuck in the a chat interface for however long it was over the, the break. But yeah, that’s, that’s me. So yeah, say happy to be here.
Paul Barnhurst:
Thanks Adam. Appreciate that, Sloane,
Sloane Kolt:
And thank you also, Paul, for having me. Really appreciate it. I’m really excited to think we’ll have a lot of fun today. So my background is mostly in accounting and finance. I’ve worked in large companies, small companies, PE-backed, venture-backed publicly traded, mid-sized large. I think the only thing I haven’t worked in, in a finance capacity is just a very, very small company. But I’ve seen a lot. I’ve also done a lot of work on ERPs and I’ve always been an automation nerd. I have a little bit of a, my background is very, very strange. Anybody who ever looks at this, like, what are you? And I’m like, I’m an amorphous piece. That’s what I, my new feeling, <laugh>. And it’s, it’s true of my history, you know, sort of until landing here at Datarails and at Data Rails, I kind of I have my hands all over the place which is just how I like it as it turns out. So at the moment I’m focusing on building Labs team that we have here at Data Rails. New and wonderful finance features with AI. Very strange change from finance to working on the technical product R&D go to market, like going from like fully back of the room to right up front. It’s been really interesting <laugh>,
Paul Barnhurst:
I, I can imagine. And, and thanks for sharing that. And I know we’ll talk a little bit about this. I know Datarails has done some things and released some tools in the AI, their genius that came out that people can use in data rails. And we’ll get to that a little later. And just really excited to see all these people rolling in. Looks like we have a great, great audience with us. I’m seeing Kenya, I’m seeing Italy, you know, Bangladesh, Germany, all over the place, so it’s really fun to have a global audience DC California. So thank you for sharing that. You know, continue to do that. This next question, please feel free to put in the comments, you know, how you’ve done this. But we’re gonna start by asking our panel here. And on this one, we’ll start with Sloane. So what is the last thing you used Generative ai. And I say generative AI instead of just ChatGPT, you know, so the conversational AI as it relates to finance fp and a or Excel. The last thing it helped you do
Sloane Kolt:
That one’s really easy. The last thing that I used was actually Data Rail Zone fp and a Genius, which is our conversational tool gives you access to your own data where you get your personalized results. Very, very fun to see your data in a secure way. <Laugh> being responded with graphs and charts as well as text. A lot of fun.
Paul Barnhurst:
Great. I love that. And please, as an audience, feel free to share your experience as well. Last thing you may have done, and Adam, what’s the last thing you did? Was it the Llama 2 versus chat G P T comparison?
Adam Shilton:
Yeah. Even though technically that wasn’t strictly and FP&A.
Paul Barnhurst:
That’s okay. We’ll forgive you.
Adam Shilton:
Yeah, that’s it. So, yeah, no, Llama 2 I I did a, a bit of a head-to-head. Very interesting actually. So, so Llama2 is the LLM that came out of Meta. And technically, you know, you can only really directly access it if you’re a developer. But I did it through an interface called Hugging Chat provided by hugging face which allows you to, to chat to it. And it’s free at the moment. So that, that was a plus. So the, the whole point of the head-to-head was, you know, is a free tool equivalent to GPT 4, which of course, as we know, requires a, a paid subscription now. But that was quite interesting. And then the, the finance use case was code interpreter before that. So I was taking data using code, code interpreter to basically pull it apart, reformat it before then doing some analysis and some future sort of forecast predictions there. So, yeah, lots going on.
Paul Barnhurst:
Great. So thank you for sharing that. And we’ll talk more about that. You know, we’ve seen some people’s comments come in. We have Malachi here using it as a personal and creative adminis administrative assistant on steroids. And I love the on steroids, ’cause I used it the other day to write some emails and I’ll admit this writing isn’t my strength and had this big email that needed to go out to a lot of vendors that was important around something. And I asked ChatGPT to write it all, and I made a few edits and I sent it out. And I had two people I know that are really good at writing a marketing respond back great email. And I just kind of laughed, like I didn’t even write it. Like, I let you know, ChatGPT write it. So I can tell you it does a good job on the writing.
And I can see both of you laughing ’cause you probably have used it in that manner before. So that’s, anyone wants a great application. That is one some others coming in for students and teaching. Yes, I’ve used it to help me prepare courses. Shouldn’t admit that, but it’s helped. I still have to go through the content and put in the work and come up with the assignments, creating data. Lots of great ways. So we’ll move to our next question here. And this one here, we’re gonna start with you, Sloan, on this. So over the last nine months, right, we’ve heard a ton about generative ai, whether it be Barred Chat, G P T, now we have Facebook or Meta’s version called LAMA two. Can you tell our audience what generative AI is? How do you think about it? And why over this last year has it become such a big issue? It’s not like it didn’t exist a year ago, so maybe if you could start there.
Sloane Kolt:
Yeah, sure. So generative AI is, is right. First off, AI is a huge, huge, huge thing. It’s, it’s, you know, what we’re talking about today is one actually small part of what AI is. A lot of people on this call, whether you think you’ve used AI or not, you probably have, if you’ve used any technology, if you have a phone, any of those things like AI is in there. We’re just not talking about it quite as much as we are with generated ai. Generated AI is really, think about like the chats that you’re having. We say chat G P T, right? You’re able to have a conversation with something, you put in a prompt, a text prompt, and it’ll it’ll respond with more text, right? You can have a conversation with it. It’s generating that next piece of information. There are other tools that are visual.
For example, you can create, you know, there’s the the Dalles and the mid journeys out there where you can create images, you can create movies, the event you can create music. That’s what generative AI is, where you provide some sort of instruction and the AI will return something that it’s created, that it’s generated from that. Now the reason why we’re talking about it now yes, you’re absolutely right. This has been around for a long time. Anybody who’s been paying attention to AI would’ve known about GPT three, for example. I remember a few years ago, and it was like, wow, have you seen what GTP3 did and how it, you know, how it interacted and all this little whispering, but it, because it was so contained in like this AI community, nobody outside of it really knew. I, I think the defining moment really was in November of 22 when Open AI released, they did a public release of chat GPT, and they happened to have done a, a very spectacular job with their model.
People started playing with it and all of a sudden they’re like, wow, it can do that. Wait, a computer just wrote that <laugh>. It’s pretty cool. And I think, you know, a lot of us suffer from that blank page problem. I think Paul, you know, to your <laugh> your email to vendors, right? It’s such an amazing place to start. Everybody’s imaginations ran wild and we kind of went on a very, very fun journey just, and I think we’re still in the midst of it. Although I keep hearing a lot about, well, is this a, is there like a, is the boom over and there’s lots of thoughts on that? I don’t think it is, but I think it’s different.
Paul Barnhurst:
<Laugh>. Yeah, I agree there’ll be changes and things, but I don’t think it’s going away. It’s not a fad. I think I liken it a lot to the internet. And I say that for a couple reasons. One, the internet has changed our life as we look back 20 years. Two, the inter internet took time there, you know, the bandwidth that’s needed for every single person to be using generative AI is huge. The cost is huge. The privacy concerns, the accuracy concerns, the bias, those are all things that are gonna take time. Some less, some more. We’re gonna continue to use it. But I think before we see it be, you know, at the level some people predict there’s some things we need to work out. And that will just take time. You know, kind of really interesting thing I’ll share, and this wasn’t in the questions, but I just wanna get your thoughts. I don’t know if you guys saw this, but the courts ruled in the US so this is for us audience by see similar precia globally, that you cannot copyright images created by AI. Did you see that? Sloan, any thoughts on that?
Sloane Kolt:
I think it makes sense that you can’t copyright the images created by ai. Well, I dunno. I think that’s a, it’s a complicated topic because somebody would argue that their prompt was so particularly special that it came up with something. My opinion on prompts, we’ll probably talk about it more later, is that eventually it’s not gonna matter because the AI’s just gonna get better. You’re not gonna have to be so specific about it to get what you want. I think the bigger thing is should, should copyrighted images, right? That’s the, that’s kind of the conversation. Should they be allowed to be used as inspiration for ai? And that’s a whole other bucket of yeah,
Paul Barnhurst:
We won’t try it on cable. That one here. We got a finance audience, but I just thought I’d ask your thoughts ’cause it’s really topical and recent. It’ll be interesting to watch. Because I don’t think we realize yet how many different ways this will permeate our lives. You know, if it writes something, can we copyright that? Even if we made edits? How much of edits, what percentage, you know, in finance, if we used it to build a model, do we own the model? At what level do we own the model? It’ll be interesting to see how the courts rule and we’ll have different rulings in different countries. So there’s all that legal stuff that needs to be worked out before mass adoption, not even thinking of privacy in the other things we talked about. So it’s, it’s just fascinating to watch it all unfold. At least it is for me for sure.
[Datarails Ad] You know what it is like 13 different spreadsheets emailed out to 23 different budget holders, multiple iterations, version control errors, back and forth updates you never really feel in control of the consolidation and collection process. Yep, I’ve been there. Stop. Breathe.
Data rails is the financial planning and analysis platform for Excel users. Data rails takes data from all your company’s disparate sources. No organization is too complex, consolidating everything into one place secured in the cloud. Now all your data finally talking to each other, everything is automated. Back into your report in Excel. Cashflow FX conversion, intercompany transactions now automated and UpToDate drill down and variance analysis in seconds. Don’t replace Excel, embrace Excel, turn your Excel into a lean mean fp and a machine. Find out more@www.data rails.com.
We’ll go ahead and start with you on this one. Adam, how do you see AI being used today in finance FP&A, accounting, what are some of the use cases you’re seeing at the moment?
Adam Shilton:
So when I answer this question, I I tend to split it in two. So there’s the use cases that relate to individual applications. So how can it help me as an individual with my work personally, and then how can it help my organization? Because the applications are different, right? So from an individual perspective, we’ve alluded to some already with Paul, you saying, you know, give me some inspirations for a, for an email response, for example mm-hmm. <Affirmative>. So there is a, the creative text-heavy document creation or all of that sort of stuff, you know, so when we look at documenting processes and all of that sort of stuff, you know, it doesn’t necessarily know your business, but you could say, right, well I need process documentation for this workflow. Please go ahead and at least give me a structure behind that, that process flow.
So from a finance perspective, you know, is it onboarding a new financial controller, you know, build me an onboarding plan for, for a new financial controller, or is it decision support? You know, I’ve got a decision to make, I’ve got a lot of options to weigh up. You know, it’s giving me decision fatigue. Can I just use it as a sounding board initially? Obviously making sure that I’m not containing any sensitive data in, in what I’m talking about, right? And I know we’ll come onto the data bit a little bit later. But then when we look at the business application and then could be you personally just wanting to do some analysis on data, for example, you know or you could have a massive database from, you know, lots of areas of the organization and you want to just maybe do some quick working out some analysis on the numbers that you’ve got and you might not wanna wade through a lot of spreadsheets to do.
And I think it’s probably what, what Datarails is, is moving to do at the moment is, you know, can I feed it some data, removing all the sensitive elements, and then just do a bit of Q&A to say, right, well un some stuff that isn’t necessarily obvious. You know, are there correlations that you can see between the data that, you know, might take me an hour, two hours to, to do myself compared to an AI that’s immediately able to, to troll that through that data really quickly. And then of course, we’re into the territory of visualization analysis and then the predictions as well, which we can get onto if we wanna. But that’s, that’s just a couple that I’m, I’m seeing right now.
Paul Barnhurst:
Yeah. And, and, and I’ll share one and then we’ll go to you Sloane, just the other day. Yeah, I was I asked it to help me prepare a P&L I wanted to do an, an exercise with a bunch of students and I needed some sample data, so I asked it, and then normally, you know, I would’ve created the data myself, spent 45 minutes to make something up and then figure out all the ratios. I fed the data via Excel to the code interpreter and I said, here’s all the ratios I want. Give me the ratios. And it got ’em all right. And I just, I checked a couple of them, dumped ’em into my file and moved on. And the first time, ’cause I had to learn the prompts and stuff, it probably took me as long as it would’ve to actually calculate ’em all.
But next time I’m there, and it had been a bigger assignment, it could save hours. So there’s lots of opportunity and We’ll, you mentioned, you hinted on this, Adam, around privacy. We’ll talk about why not to send your company data to an open system here in a little while. And I see Sloane laughing because just all I’ll say at the moment is think of would you put your company data on the front page of the internet? If the answer is no, which I’m really hoping it is, then don’t put it in open AI. We’ll leave it at that for now. But Sloan, how are you seeing it being used today?
Sloane Kolt:
Oh, gosh, I, so I think there’s like a, a very wide there’s a very wide difference, right? The, the kind of company, how big it is, right? Users at smaller companies, they’re sort of still exploring it. I think you’re seeing it a little bit more on those personal levels. They’re not quite at using it in, in some of the tools. Part of that is at the moment, unless you buy a proprietary tool that will help you, you have to still have a decent amount of technical skill. You kind of have to know what you’re doing to make some of these things work, right? I even saw like a really cool thing, you know, Python in Excel, right? Microsoft, that’s not necessarily ai. There’s so many these cool things that are starting to come out and there’s still a little bit inaccessible to to, to some folks, especially if you don’t have, you know, teams and teams of people and you gotta do everything.
But I think a lot of, so the smaller things, I think it’s a little bit more on the personal side. It’s starting to be adopted with things like and there are other tools out there as well that are helping to automatically, you know, pull data and get provide answers to what people are looking for. I think in finance in a broader way, you’re also seeing a ton of work with predictive analytics. Which Adam, I think you, you know, you covered this really well. It’s a, an incredibly useful tool to understand your data, to see trends that you can’t see. You’re not gonna review every single transaction, but the computer can. So putting the computer to work for you so that you can then focus on like the things that really matter. That’s what I’m seeing a lot more of in the finance world.
Paul Barnhurst:
Agree. And so many applications, we’re seeing people mention, you know, vacation we’ve seen people talk about using it in HR. I will admit my, my daughter loves to use it to create new characters for her favorite shows and to write bloopers and she’ll sit in for hours and ask her questions and have it write things. And so the, the just a number of use cases are amazing to think about. But there’s one question I I saw in here that I’d like to ask you, Sloane. So I’m gonna put it up on the screen here. So can you talk a little bit about, you know, she said AI and pro product management is something I’m very keen to hear the speaker’s view on. So maybe could you give your thoughts there, because I know you’ve been involved in releasing Datarail product and maybe a little bit of the product management view, if you could talk to that.
Sloane Kolt:
Sure. I mean, I think product managers in particular can use AI in so many things. If you think about what a product manager is doing, do you need help writing a user interview in the, you know, voice of, you know, with the inspiration of the mom test or something like that, that’s a really great place to start building questions. And, and in this context, I’m talking a lot about obviously tools. Like you wanna start building a PRD (Product Requirements Document), all right, <laugh> here, what’s, what’s your template? Here’s the basic, you know starting point. And then start having it, you know, mock up your product requirements document giving you ideas on how to research different items. And honestly, you could even like start putting out there, Hey, you know, this is the user problem. Here are our you know, our constraints.
How can we work around that? And just maybe starting to see things that you couldn’t otherwise see. It’s useful, it’s like another sounding board for you to play with ideas and, and walk through, you know, a, a a bunch of different ideas that are coming your way. So honestly, in product management, the use of AI is like incredibly, incredibly, incredibly helpful. You have to use it to your advantage, right? And I would like urge any product managers on this to like really get up to speed on AI in general, because if you work in tech anyway AI is going to come your way. I can pretty much promise that to you,
Paul Barnhurst:
You know, thank you for sharing that. And I’ll, and I’ll share two thoughts here and then we have a question for you, Adam. First we had someone ask Mary Theresa Walker asked, are there basic AI classes out there? There definitely are. If you’re looking for a good one in the finance realm. I know Adam has a guide he’s put out to help you with prompting. You can go to his website, tech for finance. He also has a podcast where he talks to people in AI and different things in tech. So we’ll give him opportunity to share a little bit more about that. Another one that’s very well known in the space, Nicholas Boucher, he’s great in the finance area. He is come out with a digital course, he has a virtual course, he has a guide. Another person I recommend following, and I would encourage you to listen to his podcast on FP&A Today today, it came out I think three, four months ago now, is Glenn Hopper.
Glenn is one of the four most people out there in AI. And I can see both Adam and Sloan shaking their head. I’m presenting with him to the audience. We are teaching in India tomorrow, the senior leaders, he’s gonna come talk about AI and he’s great. So yes, there’s lots of resources out there. I mean, just do a Google search. Because you can’t ask ChatGPT to tell you the courses ’cause they’re all created after it was trained. So if anyone knows what I’m talking about, that’s its data goes through 2021. So topical recent information is not what you wanna be asking. Don’t ask who the queen of England is as an example.
Adam Shilton:
So was Queen of England was,
Paul Barnhurst:
Yes, you could ask who it was, but you know what I meant by that. Adam <laugh>. Well said. All right, so this question is for you, Adam. How should finance and accounting leaders, so not the average worker, but how should leaders think about AI?
Adam Shilton:
I think, what’s the best way to answer this? So, taking AI out the question for a moment. One of the problems in finance is repeat work. You know, a never ending list of tasks that never ever gets done. And, and the problem has always been very intelligent people working on work that isn’t adding much value to the organization, right? So when I think in terms of a spectrum, low value work, at one end, high value work at, at the other end, we need to be pushing people as far into that high value add spectrum as as possible, right? So, so if I’m a finance leader thinking about how I do that, then I need to give my team the tools to soak up a lot of that, you know, low level, low value work as, as quickly as possible, really. So in terms of how to think about it, I know there’s a lot of fear, you know, especially, you know, around data, you know are people cheating?
Do I really want people in my organization with access to, to such powerful tools? And, and I don’t think we’ve got time to answer all of that on, on this session, but my, my view is at least give some sort of autonomy over staff to start playing around with these tools. And what I’ve also seen in organizations is, you know, having sort of group sessions where companies have openly said, right, well, here’s some of the example use cases, spend the next week on it and then next week we’re all gonna present, you know, some of the cool stuff that we’ve been doing. Yeah. And I think that just helps foster a culture of, of openness when it comes to AI. I mean, you, you’ve heard of, you know, schools that are actively encouraging students to use it because it’s not going away, right? So that would be my recommendation is, you know, look at the, the low value stuff that you don’t want people doing address, whether there’s any way of soaking that up with ai and then just try and push people towards that high value spec end of the spectrum as quickly as you can.
Paul Barnhurst:
I love that ’cause right? Historically the biggest complaint, or I’d say probably the two biggest complaints for FP&A professionals too much work, you know, and too much low value work. I’m a data junkie or data wrangler instead of a someone who helps with decision making and analytics. So anything you can offload to a machine now, you still need to review it. Don’t just blindly trust it, but reviewing is a lot easier than doing all the work. So whenever you can offload it to the machine, look to do that. And so, you know, next question here, and we’ll, we’ll give this one over to you, Sloane. We are two or things two or three things leaders can do today to help prepare them for AI. So, you know, Adam talked a little bit about how they should think about it, but what’s some advice you’d give them to prepare themselves that they should be doing today?
Sloane Kolt:
Yeah, I mean, the one thing I’d ask for clarification on that, are you preparing your organization? Are you preparing yourself? Because those are different, different questions.
Paul Barnhurst:
Leader, you’re helping prepare your team, your team, so maybe not the whole organization, but really your team, at least in some cases can be the organization.
Sloane Kolt:
Absolutely. The first thing I’d say is be prepared to make the change in the first place, right? I think a lot of folks, especially in finance, we get used to our rhythms, right? We don’t like to change things very much. You have to be open-minded to where, you know, you can bring in these these new tools. You know, you didn’t ask me that question, but I’m gonna answer it anyway, , what finance leaders should think about AI, and I think you need to think of it as your new best friend, okay? Because at the end of the day, just like Adam mentioned, you’re gonna be turning your team into a productivity powerhouse. They’re actually gonna be able to do the things that you weren’t able, that they weren’t able to get to. You know, every single finance person I’ve ever talked to has a to-do list that they never, ever, ever, ever finish.
Using AI tools. There’s like a, like any other change, right? So any other change management requires getting your team comfortable with it. I think a lot of folks worry that their jobs are gonna be taken away by AI. You mentioned Paul, you should always check it. And I can’t stress that enough. Anything that comes from AI, it’s actually kind of useless, especially when it comes to finance. But it’s, it’s kind of useless without the human touch for a few reasons. Once, one, if we talk about predictive analytics, okay, all you’re, all it’s doing is taking the data and maybe some assumptions that you provided with it and then providing trends and oh, it could happen like this, it could happen like that. You don’t have the nuances of what’s going on with the business. You don’t understand, you know, those extremely important business partneringthat information, the context of it all to actually properly understand whether what is being predicted actually makes sense, right?
So getting the team comfortable with it, their jobs aren’t going away. I’m really sorry for whoever’s looking for budget reductions, but like, you’re still gonna need the people. What’s gonna happen is they’re gonna be a lot more productive than they were before. A question that I actually just asked on a webinar yesterday was, you know, do you feel like finance employees are more or less productive with the introduction of Excel? Right? And I think this is kind of that same mindset. So think of it in that same mindset. Use it as a tool, get your team comfortable with the change and introduce it slowly and build trust, change management.
Paul Barnhurst:
Thank you. Appreciate that, Sloane. And someone asked where’s Christian? I should have mentioned that at the beginning on, unfortunately he had a conflict. His work required him to be on a plane at this time, so we couldn’t have him join us. But I’ll hopefully have him on a future episode and he’s another great one to follow. If you want someone to follow around AI on LinkedIn and see what he’s doing. He just released a course, how you could use Python and Excel. Yesterday’s announcement was around Python and Excel. So he, he’s put out a two hour course around that, so you know, he’s a great person to follow. Again, it’s Christian Martinez and I recommend following him on LinkedIn. So I’m gonna share something someone asked about Adam here, and you can kind of address this in your next question as we talk about security and privacy, which is the next question I have. So someone mentioned there are some companies banning the use of AI which is completely true. So Adam, you know, you’ve shared a lot of resources around it. A lot of people have talked about the concerns with that around AI and chat G P T. So you could talk, can you talk about how people th should think about the privacy and security concerns that so often come up and lead to some companies banning the use of it within the company at this point?
Adam Shilton:
Yeah, of course. And, and I’ll, I’ll do my best to embody Christian, because a lot of what I’m about to say is some of the work that he’s done. So as I say, ano another shout out to, to Christian Martinez for this. So but start things off. I I suppose without getting too technical infrastructure is the first thing. So where is the AI phosted? Yeah, so we look at the likes of chat, GPT and some of these free AI tools. They’re hosted on what’s called a, a multi-tenant infrastructure, which means that you are chatting to something on a server that a load of other people also have access to. Obviously they don’t have direct access to your chat conversations, but your data is hosted on a server with a ton of other people that are also chatting to the same AI right?
And this is where the concern comes. And obviously when you log into chat GPT or the equivalent large language model that you’re using, you get a prompt saying, you know, your data is used to train the models, please don’t enter anything sensitive. So the first thing that you can do, obviously to, to keep sort of your, your, I guess, sensitive data ring fence is just to anonymize your data. Yeah. So that could be removing a couple of columns from a spreadsheet that points to, you know, customer or supplier information. But even then, you know, template is still company data in theory. So, you know, that might still be of course for concern. But coming back to a point that Christian describes very well is the concept of using a a, a tool that’s on a public database and then taking the results and putting it into a private database.
So one of the posts Christian did was on using the code interpreter with dummy data just downloaded from, you know, what, whatever the resource was. So fake company data, essentially feeding that into the code interpreter to get the output that you want. And of course, code interpreter, if you expand the response will give you the code that it’s produced. You can then pick up that code. You can put it in a private tenant somewhere. So the example he uses his Google CoLab, for example, which is ringfenced, it’s your company’s instance of that particular tool and run it there. Now Google CoLab again, you know, I guess it’s it’s a coding. I think it works on a Jupyter Notebook or something like that. Again, forgive me, people tell me off if I’m incorrect there, but essentially it’ll enable you to, to run the code.
The other thing that he recommends for companies that have this, and not all companies do, because sometimes it can get quite expensive, is to run chat GPT on your own server. So a lot of people use Microsoft Azure. You can now install chatGPT run chat g p t on your own tenant or your own server, meaning that it’s immediately ringfenced. So you only use that server. Nobody else has got access, nobody can train on, on that data, right? And then the last piece that I’d say, sorry, darting around a little bit here is generally in the settings, and I know you can do this in ChatGPT less familiar with some of the other tools, is you now have the opportunity to switch off data being used to train the model. It used to be an opt-out on a, on a, I think it was a Google form or something like that.
But now if you go into the settings in ChatGPT, there’s a toggle when you go into privacy that essentially says, don’t use this data to train. Which makes me feel a lot more comfortable. But the only issue with that is as soon as you turn that off, you can’t use any of the advanced tools. So you can’t use plugins, you can’t use the code interpreter, you know, you literally just use the, the chat model in its most basic sense if you turn off the data. So read into that, how, however you will. But that’s, that’s just my 2 cents when it comes to, to privacy. But there’s no clear cut solution right now. So just tread carefully.
Paul Barnhurst:
Yeah, there is no clear cut solution, but yes, there are things you can do. If it’s not, I think the final word I’ll say, if it’s not in your environment and it does not have company approval, just don’t share company data. Use randomized, use dummy data or don’t use it. I love Adam’s solution of, hey, give it the dummy data, get the code, take the code off the, off the tool, and run the code in a safe environment. That is definitely one solution I’ve seen quite a bit, but just, just don’t do it. Anything you’d add here, Sloane, on the privacy and kind of data security concerns? Because I’m sure you had to deal with those in building it out for your tool.
Sloane Kolt:
For sure. So I would say there’s a few things. One, if you’re using something like an API, right? Even even with open ai as of March of this year, they changed their policy. They actually did use to train on that just like they do with regular chat GPT. But now that’s, that’s shifted. Azure, it’s gonna be, you know, the same thing where, where your data’s as mentioned, is gonna be secure. You should trust it the same way you trust Azure, which is always be cautious because way nothing’s ever that great. Or nothing’s full proof I guess I would say. So I think a lot of it is I I think that’s a great idea. You know, grab the code and pull it somewhere else. That does require a, you know, some technical skill for a lot of folks. Mm-Hmm. <affirmative>, you can use proprietary tools shameless plug, FP&A genius we’ll have other things that are coming out as well. But you wanna make sure that anything that you use is SOC 2 compliant, right? You don’t want anything that’s gonna potentially have any, any possibility of leaving your trusted network. That’s, that’s the most important thing. If you’re putting something on chat GPT.
The next prompt I put in, maybe I’m gonna see exactly, you know, maybe I know all of your private company revenue metrics I don’t know, or how much money you’re losing or making, you know, like you don’t want that to happen. So I think it’s be very cautious about where you’re putting your data, make sure that, again, SOC two compliance is what you wanna see. Make sure that the tools that are being utilized actually do have data protections. And if you don’t see that, you know, assume that they’re not and be very cautious about sharing.
Paul Barnhurst:
Yes. Thank you. Appreciate that, Sloan. And as I’d like to say, we don’t wanna hear of any of you getting fired ’cause you didn’t think about privacy. So that’s the last thing I’ll say on the privacy front. Just be smart talk. If you have questions, talk to your legal or your technology or whoever the appropriate person is in your company. Just don’t do it till you’ve talked to ’em. If you have any question. That’s the last word I’ll say on that. So I,
Adam Shilton:
Adam, go ahead. Can I add one last point? Sorry Paul,
Paul Barnhurst:
You can have last word then. Yeah,
Adam Shilton:
Yeah, <laugh>. So, so talking about company data that’s fine. Obviously, you know, be, be wary, but there is also an element now, especially with a lot of the tools that are coming to market. So, so one of the ones I investigated was Perplexity. So we, we mentioned previously that chat only goes up to 2021. Some tools now enable you to search the internet usingAI. So perplexity is one the Llama2 using the hugging chat interface, you can connect that to the internet as well. So you know that there are ways to get around it, but obviously your company’s not gonna be comfortable giving away company data. But if you are using it in a personal sense, you know, planning holidays or you know, searching the internet your data is still important. So you’ve got to decide what is important to you personally.
Mm-Hmm. <affirmative>, one of the more recent releases with chat TP two and Perplexity because you can do an AI profile on it, is you can essentially input your characteristics, your preferences, you know, so that you don’t have to give a scenario every time that you enter a prompt, right? So instead of me having to say, you know, actors, individual, the skillset, you can pre-train, chat GPT to say, you always need to work on this context, right? But just be careful what you put into that context. Yeah. Because in theory, OpenAI or whoever the equivalent AI provider is, has now got your profile. You know, and there’s, there’s an argument to say that, you know, everybody’s got my, just using my phone like a Google Pixel. There’s an argument to say that Google knows everything about me. Right. You know, so it’s, it’s one of those where you’ve just gotta be comfortable personally, not just from a, from a company angle. And that’s, I just wanted to get that in there.
Paul Barnhurst:
Thank you. Appreciate that, Adam. So real quick, I have two questions here for you, Adam. First one, can you talk about some of the resources you’ve put out there? I know you’ve put a number of resources out on LinkedIn around AI and chat GP . And so could you talk about what’s available that you’ve you’ve put out there?
Adam Shilton:
Hmm. Yeah. So this comes back to what I mentioned earlier about getting into a a, an AI rabbit hole. Sometimes it’s difficult to, to escape. And, and I remember kids keeping me up one night. I spent probably about four hours in mid journey. And, and I guess like side, side topic I guess one of my most popular posts on LinkedIn and the newsletter was how to automate 90% of your presentations using ai. And I did is I, I basically got chat GPT to write me the outline of a presentation. I then used Gamma Gamma app, I think it is to plug in the outline and then it automatically generated the slides for me. And then I just had to, had to tweak it afterwards, right? I actually got told off, ’cause I gave a presentation and then the last slide on the presentation said, just so you know, 90% this was generated by AI and it was like a mic drop moment because everybody in the room thought, my goodness, like literally 90% of it was generated by ai.
So, so that’s, that’s one. But in terms of the resources that I’ve put out it’s partly to scratch my own itch and, and partly to educate. So obviously finance get labeled with this, you know, always behind the times and, and that sort of stuff. So in the weekly newsletter, I try and investigate something new, you know, and it’s partly to, as I say, scratch, scratch my own itch. And then obviously make that publicly available so people can experiment themselves. I suppose the advantage I’ve got at the moment with the podcast and newsletter is, you know, I’m kind of free to do do what I want. So, you know, I, I investigate these tools and then, and then put it out there. And the aim is obviously, you know, to build a network, bring people up to speed and basically make everybody their own experts. Because as we said previously, it’s, it’s not going away. And I, I’d hate to think of anybody that’s missed out or, you know, not kept up in their career or, you know, missed a trick ’cause they’ve not seen something that could literally be the click of a couple of buttons. So that’s the whole premise behind the content I share, I guess.
Paul Barnhurst:
Great. And thanks. You know, I’ll just add is he’s mentioned that I’ve seen some of the content. It’s really good, you know, if you don’t follow ’em, I recommend you do. And then also to the audience, you know, feel free to keep the questions coming. Those that we can talk about, we will, we’ve definitely had way too many comments to go through all of ’em today, which is great. We love the interactive nature here and people responding to other people’s comments, you know, a lot of great things coming in. You know, someone made sure to let us know, co-pilot will launch in November, which is the AI add-on for Microsoft Office. So, you know, a lot of exciting things there. Another one shared here, you know, finance, G P T is a brand new tool. I know another guy who’s working on one for finance as well, that’s a G P T, I think it’s finchat is what he calls it.
So, I mean, they’re cropping up constantly, right? It’s just a, it’s a constant thing. And I’ve also heard a few people that have had a drop, just so everybody knows this will be released as an FP&A Today episode. It may take a week or two to get it out, but once we’re done here today, it will go over to our audio team to put it together. So anyone who had to leave early, or if you wanna share it with a friend, they can all go listen to it. It will be available also pro, it’ll also be out on YouTube, so you can watch it as well if you prefer that. So next question here, I’m gonna ask you, Adam, and then Sloane will get your thoughts on this prompting. I know you’ve written a whole guide about how to prompt, prompt. So any thoughts you wanna share, you know, any advice on writing prompts? Why is it so important that you understand how to write ’em?
Adam Shilton:
So I’m gonna, my cheesy fishing analogy again, and people who’ve listened to my podcast probably heard this before, all right? But let’s say I wanna fish, you’ll, you’ll hear the point of this in a second, but I, I want to fish. So, so I say to somebody, go find me a fish. And the immediately response is, right, well, there’s a whole ocean of fish, right? And, and that’s, that’s part of where we are starting with, with a large language model that is trained on so much data. So if you give it a vague question, you’re gonna get no end of responses. That could be the answer to the question, right? So what’s the next iteration of that, right? Well, I want a white fish. Okay, there’s still quite a lot white fish. How do we narrow down that further where it’s native, the to Pacific, you know I want one that tastes great when it’s battered and put on a plate next to chips.
You know, in the UK we love fish and chips, right? Hence, hence why I use the analogy, right? But that’s exactly what we’re doing when we’re building more advanced prompts, is we’re just guiding the large language model to the best possible output. And that’s where the whole premise of, I guess, advanced prompting comes from. So you need to give it the context, you know, where’s it starting from? You need to tell it what skills you want it to employ. You need to give it examples where possible. Yet you need to give clear instructions, and then you need to tell it what the output needs to be. You know? So I want this as a tidy document. I can copy and paste into Word, or I want this in table form so that I can drop it into Excel. Or if you’re using the code interpreter, just give the spreadsheet to download. You know? And that’s, that’s what we mean by prompting. So I guess in summary, you know, ask a poor question, get a poor answer, ask a better question, get a better answer.
Paul Barnhurst:
I’ve learned out the hard way being an interviewer, right? Hosting these shows. I’ve asked some poor questions and I, I did a LinkedIn Live one time where I asked everybody and they all kinda looked at me like, what are you talking about? And nobody really answered the question. So that’s similar to ChatGPT, except for in the case of Chachi pt, you’ll probably get a lot of information, you’ll get an answer, it just may not have any value to it for what you wanted.
Adam Shilton:
Yeah, absolutely. And and the last thing to end on though is I miss the big chunk of that fishing analogy, which is the experience of the fishermen. Yeah. And, and I think that’s what we often forget. We hold AI to incredibly high standards, you know, because it’s the next big thing, right? But, you know, you can’t train an AI on, you know, 20 years of fishing experience, at least not yet. Right. You know, so if you expect to get the quality of that experience in gut feel, you know, what’s the weather gonna do? You know, oh, you pointless fishing there after these guys have been in yesterday because all the fish will be gone. You, you see what I mean? So there is still a gap for that experience to be built into these large language models and people are training them, don’t get me wrong, but I still think it’s gonna take a while to train the equivalent experience of somebody who’s been doing it for years and years. We’ll get there, you know, hundred percent of it. But for now, we still need that context and we still need the, the individuals that are using it.
Paul Barnhurst:
Thanks, Adam. Any, anything you’d like to add to that Sloane around prompts and prompting the tool?
Sloane Kolt:
Yeah, I think that was a, that was a great analogy. I think it’s really just about being specific. At the end of the day, like, Adam, you’re alluding to this, it’s like if you give something really broad, you’re gonna get something really broad. Because at the end of the, you have to remember you’re talking to a computer, you’re not talking to a person who has all the context of the things around you. They don’t know what room you’re sitting in or what conversation you had that made you ask that question. So if you treat it as if like they don’t know anything, right? Or when you’re talking to the, to the chat bot, you’ll get better answers. There are some tools that are more attuned to answer specific questions, so you don’t have to be as specific. And that would be like a, a case in FP&A Genius, you know, what your intention is, right?
Paul Barnhurst:
Yeah. You’re, it’s trained on a certain area, it’s focused more so it’s, it has a little bit more context, so to speak. If you think of a conversation with somebody,
Sloane Kolt:
Exactly. Be specific. Give examples. I think you also mentioned it’s called like Few Shot, if anybody ever wants to look up any of these things. In terms of prompting, the one thing that I’ll say, and I, I alluded to it earlier, I think that eventually we’re gonna get to a place, and I dunno that it’s even that far away, that prompting and being so specific about how you prompt is gonna be as important. There are certain things that you can do. We can get really advanced where you’re, you know, putting specific kinds of like Mark Demarcations, you know, pound signs to say this is the instructions. And some of those things that people do to get very, very, very specific responses. Those types of things are gonna be much, much less necessary in the future.
Paul Barnhurst:
It’s a lot like Google, when search engines first came out, you had to do Bolean Logic and be very specific. Today you ask a general question and it does a pretty good job. I think that’s probably a pretty good comparison to what we’ll see with conversational AI. It’s very, very similar and probably, I think it’ll be a quicker adoption rate than we saw with the internet. ’cause We’ve gone through it before, but it’ll be similar.
Adam Shilton:
Mm-Hmm. And we’re, we’re already seeing that already. So, you know, coming back to the whole premise of training in AI, right? So and again, this, this comes back to, you know, how people can start adopting it is, you know, don’t just adopt AI willy-nilly without any sort of thought of what problem it’s solving for, you know, so we are now in a, in a space where there’s hundreds if not thousands of AI tools now available on the market, right? So you’ve got a narrow use case that you might wanna fulfill. So if you spend a lot of time trauling through documents, you know, maybe use Claude 2 or something like a petal to, you know, distill that information and summarize it, that’s a narrow AI use case. You know, because it’s already pre-trained on that specific scenario. I mentioned the whole premise of ChatPT now giving you the ability to enter your preferences in that context.
So when you look at my framework, that scenario piece that will be null and void pretty soon because it’ll already under understand what the scenario is, right? But the one trick just before we move off prompting that people can take away to today it’s something that I found out recently and I, and I dunno why I didn’t do it before, but when you imagine the way that you speak to a human being, you don’t just throw a question at somebody and expect them to immediately come back with the best possible answer. There is a conversation. And I think with the advent of prompting, people start getting into the realm of, oh, you know, if I don’t get a good response, it means that my prompt’s bad. It’s not, it just means that you didn’t follow the conversation through, you know, chat is the operative word here in chatGPT you continue the conversation going.
And the trick that I found was on the clarification at the end of a prompt. Yeah. And, and, and it’s funny how the output the quality just goes up. If you end a prompt by saying, before you carry out the task, please, please clarify that you understand what I’m asking of you, and let me know whether there’s any gaps or anything that you change to improve the quality of the output. So that way, instead of it just doing the task and presenting you with information that might not be what you wanted, it’ll come back and say, I understand you’ve asked me this. If you want me to do a really good job, then here’s three points that you could do with answering before I carry out the task. And then you could save yourself hours of, you know, going around in circles with a prompt that didn’t quite work, because instead you’ve just asked for clarification. In the same way as you would with a human being. So that’s something that people could take away today.
Paul Barnhurst:
Thank you. That’s great. I think continuing that conversation and continuing to ask it to refine like you would in a conversation is a really important point. I’ve definitely done that where, no, this isn’t what I wanted. No, your answer’s wrong. Or I, let me rephrase what I gave you and it, it takes that into consideration. So I wanna share one comment we got here. Go ahead.
Sloane Kolt:
Yeah, I just Adam, what you said, I think that’s a really, really, really great point. It just reminded me of one other thing. If you tell ChatGPT or or other chatbots, you’re like, do not do this. That usually doesn’t work very well. I don’t know why <laugh>, but it’s one of, it’s just one of these common hangups.
Paul Barnhurst:
Then you must not have kids. <Laugh>. Sorry, I couldn’t resist. We
Sloane Kolt:
Get, you’re gonna get the opposite response, right? So <laugh> I think, I think that’s one of the things just to keep in mind, it’s much better to say, do this, like this all, you know, be specific about the things you want it to do. When you say, don’t do this, I don’t why I just doesn’t get it.
Paul Barnhurst:
Thank you for sharing that. And I, real quick, I’m gonna add a comment. We had come in here from Serena. I thought this was interesting. She’s like, I saw a post from a C E O and her requirement for hiring is you have to have a great writing skills. Otherwise, it is a big no for hiring. With AI tech these days. What do you guys think? Writing’s still important. Communication is still important, but you can definitely use AI and you should be to improve your writing, to improve the quality. Writing is not my strength. I’ll be the first to admit it. Anyone who’s read some of my stuff will know that. But I’ve run things through Chat GPT and it’s helped give me a better result. So I think it’s still important. Definitely something we need to learn. It’s not like all of a sudden we don’t need to write, but we definitely use GPT to help improve the quality. Without a doubt, that’s one of the great use cases for it. So that’s all I’ll say on that one. And then we’re gonna kind of move on here to a next question. We talked a little bit about the Sloane, but maybe could you add a little bit more about your experience in bringing, you know, FP&A Genius forward? What was that like? And just kind of walk through and talk a little bit about that, what you guys done and how to think about that tool.
Sloane Kolt:
Yeah, I mean, it, it’s been a lot of fun. First off I work with an incredible, incredible team. We have some absolutely amazing developers and, and product managers working on this. And, and, and much more. I think a lot of, you know, this experience is obviously the data privacy concerns, but the, the biggest thing that was always in the back of my head was like, listen, I’m a former finance person, or maybe you can never take the finance out of the person finding that in my day-to-day life, I’m like a finance person. I’m not sure I agree with that. Like, I can’t, I can’t get out of the math of it all. But I think there are things, you know, you have to care about the quality of your data when you’re building something like this, like a, a finance tool, you don’t wanna be spitting out numbers that don’t make sense.
Something everybody should know about the LLMs, they are created, they are built on math and statistics, but that their capabilities are not very good on the mathematical side. So just, just in terms of the chat. So don’t get those things confused because if you do, you’re, you could be very upset with the results you get, you know? So from our end, it was making sure that the data was very reliable and coming from our existing database, things that we, you know, those numbers we knew were gonna be really solid so that we could trust them. You know, I don’t wanna release something where I’d be nervous that, wow, that the wrong things popping up that, you know, that was, I mean,
Paul Barnhurst:
You don’t wanna give a CEO the wrong numbers is getting ready to go talk to the board. That’s a bad idea.
Sloane Kolt:
I getting fired, but getting fired wasn’t on my to-do list most of the time when I was in finance. So I, so I think that’s one of the biggest things. That was actually one of our bigger challenges because we wanted to make sure that it was integrated in a way that kept it a really strong focus on data integrity. And then it’s interesting, you know, you’re, you’re thinking about it. There’s a lot of different things to think about. You know, you can build it in a more general way. You can build it in a more specific way. For those of you who are, you know, working in product and AI, you have to determine how much, and I’m gonna use this in quotes, decision making ability that you provide to any AI tool. So you have to find the place where to draw the line.
And it’s gonna depend on what you’re, what you’re building, right? For us, you know, we’ve, I think we’ve found a, a middle ground. Again, we wanna make sure that the data’s right, that the calculations are right, that nobody’s walking away going like, yeah. So that number makes no sense. And I’m gonna have to go now, I’m gonna to go now and go dig into my numbers because that was just so wrong, and I’m really sorry. It’s like a, just talking about it’s giving you like this massive amount of anxiety. And so I thought that <laugh>, like when you’re presenting numbers, when you’re dealing with numbers and one of the use cases for our product I feel like a broken record every time we talk about it, but it really resonates with finance folks. You’re out there presenting the numbers all the time, right?
Whether it’s in a board meeting to your CEO or department head, whatever it may be. You’re, you prepare for those things time and time and time and time again. And you know, you, you think you know every question that could possibly pop up. And, and sure enough, something gets asked and you’re there, you know, furiously typing away, they’re sending someone on your team to go find the answer or you’re trying to dig through the data yourself. Andfor us it was about relieving that anxiety so that you could get a fast answer that you knew you could rely on and, and move on with your day and, and stop wasting time.
Paul Barnhurst:
Thank you. I really appreciate that. That’s great. And you know, we’re coming up here near the end of the hour. We may run a minute or two over, but we’re gonna kind of move forward here. A few more questions as we wrap up, but please keep the comments coming. We love them and you’re always welcome to, you know, reach out to any of us on LinkedIn. You know, we really enjoyed this conversation. Next question I wanna ask, and we’ll start with you Sloane favorite, and it can’t be data rails tool, but favorite tool beyond ChatGPT, like cool application you seen or something you really like?
Sloane Kolt:
I’m gonna say, so this is something I’ve read, I read about recently. So I don’t have direct experience with it, so I’ll put that caveat on it. Okay. but there’s this tool out there, or there was research done that can predict whether a song is gonna be a hit. And the way that they do it is by measuring a heartbeat response. And apparently it’s like crazy accurate.
Paul Barnhurst:
Doesn’t surprise me, actually. They’ve, they’ve got a formula down for hits for music. It’s crazy <laugh>.
Sloane Kolt:
Well, it turns out sometimes they put big bets and they don’t, they don’t make it
Paul Barnhurst:
True. They’re wrong sometimes <laugh>, no question. But
Sloane Kolt:
It was, it was kind of really cool to just see that, like from, from the heartbeat. And then I think in general in healthcare, I am so, so eager to see what’s gonna start coming out there. I, I feel like for a lot of people it might seem creepy, but the truth is, this is a place where your AI is your friend. AI is gonna help diagnose things way faster than a human can. When you can train, if you think about it in a, in a very, very simple way some doctor, right? And this isn’t nothing about doctors. Doctors are amazing. Please go to doctors. But they’re, they’re not looking every day. They don’t look at thousands and thousands and thousands and thousands of images of, this is cancer, this is not cancer, this is cancer, this is not cancer, right? And these tools are able to see it and they can identify in the progress, oh, this was someone who was diagnosed with cancer, but this was an earlier scan, right? And those tools are getting more and more accurate. And I think it’s just, it, it’s going to be transformational in the healthcare industry. And just for, for, you know, diagnosis maybe great things for prevention are on the way. Really exciting stuff.
Paul Barnhurst:
Thanks. Appreciate that, Sloan. So Adam, what’s your favorite one beyond Chat, GPT And for you?
Adam Shilton:
I only get one
Paul Barnhurst:
<Laugh>, you only get one. We’re limited on time, otherwise I’d let you go for like 20. ’cause I know you have a long list.
Adam Shilton:
Yeah, so, so the, the one that I had most fun with recently and I need to go back to it actually because I, I think they’ve released a load of updates, is it’s called Mind OS, and you can train your own bot. Yeah. so well, and they’ve got some templates as well. So you don’t even have to have an account, you just go to Mind OS and you can experiment with some of the bots they’ve already pre-created for you. But I like it. One of the examples is like an industry analyst, an industry analyst bot, right? So you go in similar to Chat P T, it’s a chat interface, but it’s pre-trained to be able to help you find industry trends, industry data and that sort of stuff. So you can say, you know, show me manufacturing industry trends.
And it will not just do the chat thing, but it’ll also produce visualizations for you to the point where that particular use case actually generates like a two page PDF with a full written breakdown and visual analysis of, of that industry for you. But of course, without using the templates, you can go in and you can basically set it up to say, right, well this is your context, this is what you need to access and this is what your outputs need to be.So yeah, if you’re curious to have a look or, you know, you don’t need developer experience you can’t just follow the wizards, go in, have a gut, create in your own bot. ’cause In my view, you know, we’re, we’re gonna get to a future where we hire ais and we use ais in the same way as we do people.
Right? You know, so I think we’re gonna get into the territory of AI sourcing as well as outsourcing, you know, so we outsourcing some tasks, we insource some tasks with intelligent people that we hire, right? But I think we will get to the territory of AI sourcing, which comes down to those narrow use cases that we suggested before. So I think the sooner people can start going and having a play around with creating their own box, you know, with certain purposes, I think that’s gonna give people a really good window into the way that these technologies are gonna move forward. So mind oss, I’m not affiliated with them, but I had a lot of fun with it.
Paul Barnhurst:
Yeah, I’m sure you have an affiliate commission. No, I’m kidding. <Laugh>. so we, we have about a minute left here, so I’m gonna move, move on. We’re gonna skip one or two of the questions we had and move to what’s one of my favorite sections. These are the rapid fire questions. So you, I’m gonna give you guys 20 seconds on these as we’re kind of short on time. So we’ll start here with Sloan on this one. What’s your favorite thing about Excel? Favorite function or feature?
Sloane Kolt:
I told you it was me boring SUMIFs.
Paul Barnhurst:
Sumifs that that works. You’re not the first, nor will you be the last to mention that one because it opens up a world of things, even though it’s, it’s not the sexy, sexy one people mention, but it get, it’s the workhorse. Adam,
Adam Shilton:
Pivot charts.
Paul Barnhurst:
Pivot charts, you know, I’ve loved pivot tables. I’m not a big pivot chart guy. I don’t know why, but I love pivot tables, but pivot charts are great.
Adam Shilton:
Yeah, for me, I, I struggle with, I struggle with not being able to see stuff, you know, so, so just having a load of information in, in lines and columns doesn’t really help me. So as soon as I get it into a chart, it just helps me out a load.
Paul Barnhurst:
Yeah, I love it. You know, someone mentioned here, Zapier here is a great tool. When we talked about tools for ai, you know, others ask, we won’t have time to cover this one, but just a thought out there. It could be a LinkedIn post for one of you if you want. How do you see AI playing a role in analyzing and identifying underperforming areas or projects? So that’s another one that someone asked. So next question here, this is another fun one we ask, and we’ll start here with you, Adam, if you could meet any person dead or alive, who would you meet?
Adam Shilton:
Is it just one person?
Paul Barnhurst:
It’s one person. You can’t give me a group.
Adam Shilton:
So I can’t have a, I can’t have a dead person and, and a live
Paul Barnhurst:
Person. Nope. You get one or the other.
Adam Shilton:
Okay, fine. So it’s a dead person and it’s not, it’s not related to finance, but it’s a guy called Anthony Bourdain. . So type of coach was Tony Bourdain, who was one of the, I guess, original TV chefs.
Paul Barnhurst:
Yep. I know who he is.
Adam Shilton:
Yeah, so he, he traveled the world, you know there was a political angle to a lot of the stuff that he did, but I love food. So I loved his speaking style. You know, I, I basically lost weekends to watching the, the layover parts unknown, that sort of stuff. And you know, unfortunately he had a pretty horrible end. But if I could meet somebody just to, you know, have a meal with them he’d, he’d be the guy that I’d love to meet.
Paul Barnhurst:
Alright, thanks Sloane, how about you?
Sloane Kolt:
I feel like I don’t have an interesting answer for this. Also I have way too many people. So choosing one very difficult, fun fact about me is that I’m terrible choosing favorites. <Laugh>, I, I’m gonna go with, and there’s so many other people that I really could say here, but I’m gonna go with Nikola Tesla just because I’m super fascinated by him. And I think he’s just like a wacky genius that like, I just really, really curious about. So I’d be interested to see how he speaks and interacts with people and yeah.
Paul Barnhurst:
Great. I like that. So we’re gonna give you kind of last word here and then we’ll give an opportunity for you to just tell the audience how to get ahold of you. So if you could offer one piece of advice to our audience, you know, kind of as we’ve talked about today around AI in the workplace, we’ll start with you Sloan. What would that one piece of advice be?
Sloane Kolt:
Be open-minded and play. If you just play, hop in there, and I think this is probably true for a lot of things, especially when it comes to new technology. Assuming that you set up a, a nice safe environment for yourself where you’re not gonna yell down an airplane or something like that by accident make sure that you know, you’re not going too crazy with your risk taking, but be open play. Don’t be afraid to make mistakes. You’re going to, that’s how you’re gonna learn to be open.
Paul Barnhurst:
Love that, great, that great advice. And that goes far beyond AI. That’s really good advice. Adam.
Adam Shilton:
Mine’s great. Pretty much the same as Sloane’s, to be fair. Always have a tab open with some sort of ai, whether it’s Chat GPT, Claude 2, Bard, even though I’m not a huge fan of Bard, that’s a conversation for, for another day, right? But always have them open and always think to yourself, right, especially when you’re starting an activity. Is there a way that I could start this quicker by dropping the thought process into AI, obviously providing you not giving away any sensitive information, right? So have a tab open and then, you know, maybe find an easy freemium tool to just have an experiment with once a week. You know? Yeah,
Paul Barnhurst:
I mean even ChatGPT has its freemium version. A lot of people think well cost and well, for some things it does, but not everything.
Adam Shilton:
But don’t get too stuck because, you know, we mentioned Microsoft copilot, it will end up morphing its way into our day-to-day without even realizing it. So yeah, you, you’ll give yourself an edge if you can be savvy and use these tools to your advantage early. But that experience that you build now is obviously gonna make sure that you’re ahead when it does start being built into these tools that use every day.
Paul Barnhurst:
Thanks for that. I know we’re a couple minutes over, but we’ll give a minute here. Sloan, if somebody wants to get ahold of you, maybe ask you a question or learn more about you, what’s the best way for them to contact you?
Sloane Kolt:
Linkedin’s always a great place. Sloane Kolt should be pretty easy to find. It’ll be really interesting if there’s another Sloane Kolt on the planet. By the way, if there is, come talk to me. I wanna know <laugh>, but LinkedIn.
Paul Barnhurst:
Great, thanks Adam.
Adam Shilton:
Yeah, Adam Shilton on LinkedIn, SHILTON. I’m not the only Adam Shelton though, but there is a another dude that I actually reached out to, to connect with just ’cause he had my name. I think he’s like a, a media and design expert or something like that, but no, Adam Shilton, the, the one with the Adam Shilton Tech after it, not the other Adam Shelton. And then just tech for finance.com. And that’s four spelled out, not the number four.
Paul Barnhurst:
Thanks Adam. I really appreciate it and I’m just gonna share this comment as we close here. You know, guys, you’re doing a great job sharing all this type of information. Kudos to you and we’ll definitely be waiting for your next live podcast. So thank you. Really appreciate that Marta. And I will add, I’m just gonna throw a plug in here at the end. Adam has a great podcast himself, tech for finance. You’ll go to his LinkedIn, you can find that. And he’s brought on a lot of great guests in this area. He’s had AI conversations, so if you’re looking for more conversations, he also has a great podcast and this will be out soon. So thank you so much everybody for joining us. We really appreciate you staying with us. We know we ran a few minutes over, but thanks. Really enjoyed the time and we look forward to the next podcast. So thank you everyone.