Triumph with Tech: How AI Empowers Today's CFOs

Speaker A: That's why specialized tools have been built on top of these LLMs. Using the strength of the LLMs, which is around content and pattern recognition and reasoning, and then using deterministic formulas, which are, uh, the holy grail of finance and math calculations. And bringing those two together, there are tools which are just focused on finance professionals.

Speaker B: If you're listening today, it must mean you're a strategic cfo. Tune in regularly for tips, best practices, and eye opening stories that will help you optimize your cfo, uh, experience. And don't forget to leave a review and subscribe.

Speaker C: Hello and welcome to FEI Silicon Valley's Strategic CFO Podcast, a series of conversations with business and government leaders that impact our lives here in Silicon Valley. My name is Jen Robertson, your host. My guest today is Gaurav Mandelata, known as Rav, who is CEO of Socio Squares, an AI software development and online marketing firm, and the Chief Product Officer at Propel, a SaaS company that enhances small business visibility through SEO and Google reviews. So welcome, Rav.

Speaker A: Thank you so much. Um, thank you for having me.

Speaker C: It's a great pleasure. So, for our listeners, one of the reasons I invited Rav to be on the podcast today is because Rav gave an excellent presentation at our recent dinner meeting in September. And I know not all of our members can attend our dinner meetings. So I thought it was useful to have Rob as a guest so we could kind of recap the conversation and dig in a little bit to these new AI tools that are, uh, really changing our lives here in Silicon Valley. So before we get into that, why don't we first of all go through your background, Roth, and I'd love to know how you got into this. It was really impressive what you had to say to us. So tell us a little bit about your kind of early career and how you got into this area.

Speaker A: Sure. Happy to. So I'm a technologist and marketer, and by education I have an engineering degree. The first time I came to the US was back in 2003 to do my Master's in Telecom Engineering at University of Pennsylvania. I graduated in 2005 with my master's. I was lucky to land a job at Comcast in an engineering role. I was there for about four and a half years working in the CTO office. In my last role, I loved my job. I, uh, used to work on technologies which were three to five years out. So AI was a distant dream at that time. This is, I'm talking about in 2005 and 2009 timeframe but we were doing some really cool stuff where video calling from your landline. We had tested devices which you could do that. So I was always the futuristic. So 2009, I started my entrepreneurial journey, uh, with building a platform which brought high quality Indian movies or Bollywood movies for Indians in the us and it was in some ways similar to what Netflix is today.

Speaker C: Mhm.

Speaker A: Netflix at that time used to do DVDs.

Speaker C: I remember those days.

Speaker A: Yeah. So we did that for the Indian audiences. Between 2009, 2011, I was able to raise a, uh, small round of seed funding. We licensed over 6,000 Bollywood movies, over 10,000 Indian TV shows. And then in 2011, YouTube jumped into the game doing the same thing. Okay, um, interestingly, the bandwidth, the Internet bandwidth in India at that time was very, very low. So nobody was uploading user generated content at that time. So YouTube wanted to get more traction, so they started licensing the same movies and TV shows we had on our platform. Now, uh, at that time, my marketing team had done a brilliant job promoting this platform. And m, the companies from whom we were licensing movies and TV shows, they started asking us to help out on their social media marketing. We had crossed a million likes on our Facebook page in 20.

Speaker C: Mhm.

Speaker A: So we started social squares primarily to help out our media clients grow their social media presence. And we had a technology team, so we kept developing social media software. In 2015, we started working on an AI platform which could predict how many likes your next Facebook post can get based on what you put in in the content and the image you attach with that. And we were doing that in partnership with a Wharton PhD and a Wharton professor. They built proprietary algorithms around that. So we partnered with them, we licensed their, uh, algorithms and built this prediction engine so brands could use that to improve their content to get more likes and comments on the next Facebook post. Those were really early days for AI. Accuracy levels were really low.

Speaker C: And so what year roughly are we, you know, talking about what year this is?

Speaker A: 2015.

Speaker C: Okay. Mhm.

Speaker B: Yeah.

Speaker A: And then that product, even though it was working as expected, we were not able to get the edge in terms of the product market fit. Yes, it would give suggestions that add these words, make those changes. But the incremental increase in engagement wasn't that much. It was Nothing like what ChatGPT has been doing for the last couple years. So really early days. But we kept on, uh, developing other tools as well. I remember in 2017 we had built a face recognition platform to gratify the followers our clients had on Facebook. So we Work with a lot of malls, shopping malls, and they have a m million plus people liking their Facebook page and following them on Instagram today. So what we did in 2017 was we built an application on their Facebook page where anybody could upload three of their selfies. And when they walk into the mall, we had our um, selfie, uh, camera. We called a deal fee. They just look at the deal fee camera. And we would match at the back end. Okay, is this a registered follower? And they would get a discount coupon on their phone within a couple of minutes. We were using AI for the facial recognition at that. So we've been dabbling with AI long before it became what it is. And along the way using technology for marketing and growing our uh, clients, businesses. So that's in short my story.

Speaker C: That's a fascinating story. I want to thank you for that and I uh, want to focus on. So our audience is primarily CFOs, controllers, financial professionals, and there's a plethora of tools out there. And I know when you gave the presentation you kind of did a little survey at the beginning. What tools are people using? ChatGPT or code perplexity, Gemini, Copilot, uh, others. What are all the different tools? And I confess I find ChatGPT, the paid version, really useful for general research. So what I do, I'm an M and a professional and CFO of our firm. And what I find it really useful is researching all kinds of buyers lists and information about different companies. And obviously you have to check the sources, but that's what I find really useful. But I haven't gone any farther than that really. I mean the idea of uploading a, uh, client's financial statements to a tool to then do a forecast or something seems a little terrifying to me. And obviously concerns about confidentiality Kind of walk us through these different tools and what's your opinion on how they might be useful to financial professionals? In which ways?

Speaker A: Sure. Before we get into how they're useful in the finance space or any uh, vertical, we need to think about what the underlying technology is.

Speaker C: Mhm.

Speaker A: ChatGPT runs on a large language model, uh, called GPT, which stands for General Purpose Transformer. And Claude has their own large language model. Claude is the product of Anthropic, which is a different company. Gemini is owned by Google. And there are few other LLMs which are popular. Now as the name suggests, LLM M, it's built on large language model. So this form of AI is trained on language. It is not good with numbers, which is especially the calculation, the financial part is more deterministic. Calculation calculator, for example, would be 10 times better than using a large language model to do math.

Speaker C: Mhm.

Speaker A: When we use any large language model, whatever we type in, it converts each digit into a token and then it does pattern matching at a very large scale at the back end to predict what would be the next token. So if we ask it to do a multiplication or a complex mathematical problem, it is trying to do pattern matching and that's why it fails miserably. So for any of the AIs to work well for finance professionals, they have to use a combination of LLM with another technology, complex Excel functions which they can run at the backend and just display the output rather than trying to predict the output because the prediction part is not required here. So that's why specialized tools have been built on top of these LLMs using the strength of the LLMs, which is around content and pattern recognition and reasoning, and then using deterministic formulas which are uh, the holy grail of finance and math calculations. And bringing those two together. Uh, there are tools which are just focused on finance professionals and I talked about a few of them. For example, there's Bookkeeping AI, Shortcut AI, there is Quadratic hq, there's Julius AI. These are some cool tools. They're all early stage startups. They're not as big as OpenAI or Cloud or others. So we m have to use them carefully. You have to be careful with uploading all of that. But using them in tandem with the powers of LLM is the right approach for finance professionals. Rather than using even the paid, even the $200 version of ChatGPT. If you use it for research, that's great. There are things M, you can do really well when it comes to research with ChatGPT's deep research feature or ChatGPT's agent feature. But if you want to analyze financial uh, data, then tools like Bookkeeping AI or Shortcut or Julius AI would be much better.

Speaker C: So could you then take us through. So you've mentioned kind of zeroing in on these specific AI tools for finance professionals. Could you take us through? Uh, let's take them one at a time. Bookkeeping AI, Shortcut, Quadratic, Julius and Context. And could you take us through each of those five tools and briefly describe kind of the, uh, you know, what they are, the maybe strengths, weaknesses and applicability, what you would use, for instance, starting with Bookkeeping AI, what you'd use it for. And just walk us through these if you would Please.

Speaker A: Sure. Yeah, happy to. So first of all, I'm not associated with any of these.

Speaker C: Understood. So. No, no, got it. All right, so now that we have that cleared up and uh, so you're not an invest in any of these or whatever.

Speaker A: Right. So bookkeeping AI actually came to us uh, through one of my team members who's my finance manager in one of our companies. And usually bookkeepers don't like AI. But the moment he tried bookkeeping AI, he fell in love with the tool. And right now we are using it for one of our companies. I still don't trust it to give um, our bigger data sets, but it's working well. It does a better job than QuickBooks tagging and other things. So it will connect to our bank accounts directly. We can connect our email to it. So for example, if there is a contract which gets finalized and signed, it will pull out the details of the contract, what would be the billing terms and create an invoice. And that invoice would be sitting in my drafts to hit send. So in addition to bookkeeping, it has those features as well which make it easier for early stage entrepreneurs or early stage companies to make it easier for them. So bookkeeping AI is great for that.

Speaker C: Mhm.

Speaker A: Thought card. Yeah.

Speaker C: Can I just ask a follow up question? Does it. So you would feed in, say your financial statements, say it's end of June, you've got your six months financials, uh, you know, your P and L, your balance sheet, your cash flow. Would you feed in your existing financial statements in some historicals and then ask it to generate invoices or connect to your emails? And if you're feeding in existing data, where does that data go and how can you be certain of the confidentiality?

Speaker A: So I think the licensing terms are similar to licensing a QuickBooks account or buying a QuickBooks Online account. Similar to how you would do your setup in QuickBooks Online, you are essentially setting up uh, bookkeeping AI. So you will connect your bank statement, uh, or your bank account directly to Bookkeeping AI. It will then tag all the information and how QuickBooks Online displays everything. It will display that and then that would be exportable and you could have your balance sheet, your cash flow statement and your profitability from there for billing. If you're familiar with bill.com or. Yes, like that. Think of it like setting up a Bill.com account. But Bill.com, you typically don't connect to your email. Right. In this case, once the company is set up, you have that option of connecting your email to Bookkeeping AI and as soon as it monitors the contract or invoicing related and customer emails, so when it sees that language, which you can put in keywords for, it will create that invoice and it would be ready in drafts. It doesn't send it out on its own. You have to trust the platform how you would trust QuickBooks with your QuickBooks or anything else.

Speaker C: It's just sort of like a more intelligent and more multifunctional tool beyond QuickBooks.

Speaker A: Right.

Speaker C: So it seems to me that QuickBooks might be watching this and thinking, oh my goodness, we better add extra functionality.

Speaker A: Yeah, uh, absolutely. I'm surprised that they haven't launched anything meaningful yet. And they did announce Intuit AI Assist last month, but it's nowhere close to the capabilities Bookkeeping AI is offering. I wouldn't be surprised that if in next 6 months or 12 months, QuickBooks comes out with some of those features and we then don't have to worry about switching to another platform, maybe they

Speaker C: should buy Bookkeeping AI. Bookkeeping AI is a startup, so with my M M and a hat on anyway.

Speaker A: Yeah, and I'm sure some of this consolidation would happen in actual. In the last six months, Meta has acquired several companies in that effort of bringing in more AI talent as well as capabilities. But that's a different story. So moving on to Shortcut, Shortcut, Shortcut and Quadratic are focused on making Excel in the AI age. So think of those two platforms as AI first or AI native Excel platforms. So whatever Microsoft Excel has, they have those features, but they are all AI enabled. So instead of running formulas or typing in formulas yourselves, you can just, in plain English, give those instructions to Shortcut or Quadratic. And it will produce nice graphs, it will analyze data, uh, it will produce reports based on the raw Excel you upload in these platforms.

Speaker C: Nice. Yeah. And so again, confidentiality. You have to trust the platform just like you're trusting Excel, except it's a step beyond that to create, you know, as you say, reports and graphics more easily.

Speaker A: It's a very well funded Bay Area startup. They have reputable investors behind them. They do have enterprise plans where I'm sure they have a msa, uh, in place which ensures that the data remains in a pod which is owned by the customer. So I can't comment on that from a legal standpoint or a confidentiality standpoint, but I'm sure they have things built in around that Quadratic.

Speaker C: Is that just startup as well?

Speaker A: Yeah, they are another startup. I'm not sure if they are Bay Area based. But we tried Quadratic for some of our uh, data analysis, which is more advertising and media, not for. And both Shortcut and Quadratic were equally impressive.

Speaker C: Okay, moving on to Julius AI.

Speaker A: So Julius is similar to Quadratic and Shortcut in many ways, but it has a big differentiator. What it does is that it will write Python code for every analysis we ask it to do. And the beauty of doing that additional step is that we can export that code and then we can run it on our data on a local machine. So if you are uh, savvy with some basic Python code and do not want and to upload that data, you can ask Julius that. This is my goal. Here is a sample data set.

Speaker C: Mhm.

Speaker A: Write me a code for that and then use that code and run it on your full data set. So I think that's a big advantage Julius has over the other two and that might work better for some professionals if they have serious confidentiality concerns.

Speaker C: So once you've written your customized code, you export it and then you use it on your other files. Does Julius then retain. So would it have a library of customized codes that people have created? They're people like you, customers have created and then they'd have this library that then other folks could go into and perhaps use or adapt those or does it become your proprietary code?

Speaker A: That's a great question. And it might be a great product feature for them to add because I didn't see that. So it's a uh, good suggestion. You might want to.

Speaker C: All right, I should contact them. Are they, is Julius a startup as well?

Speaker A: Yeah, they are a startup as well, but they claim that they have over 2 million users using the platform already. So it's, I would say it's a fast growing startup. They have their GDPR compliant, they are SOC2 compliant as well. So they're ahead of what I've seen from on Shortcut or Quadratic. But it's similar to licensing ChatGPT and how our GPT chats are saved in the same way our code gets saved as well. Within Julius. I think the library could be a great value add for all users, which they can probably enable on their platform.

Speaker C: Yeah, once you create the custom code, you should charge them money back for them having it, keeping it in the library or whatever and then moving on to Context AI.

Speaker A: Yeah, so Context AI is a full office suite which is focused on Excel and not just focused on finance, but it can do all kinds of analysis on Excel Data and make PowerPoints at the same time. Uh, so think of context AI, uh, as an AI first office suite. So starting from Docs to PowerPoints to Sheets, they do it all. And they are also a Bay Area based startup.

Speaker B: FEI Silicon Valley is Silicon Valley's leading professional organization for corporate financial professionals and tax executives. It's one of over 50 chapters of financial Executives International, the uh leading financial network for CFOs, advisors, planners and more. FBI gives over 9,000 executives around the world a uh toolkit of ways to share best practices, opportunities and connections. Boost your professional opportunities by going to www.feisv.org.

Speaker C: one of the key things we have to do in M M and A when working on the sell side is to create sale documents which are typically in PowerPoint for the companies that we're working with. It's what I'll call a360 analysis, like an offering memorandum. And similarly, when companies are doing fundraisings too, they have to create offering memorandum with the whole company story. It has to include customers, markets, industry analysis, plus historical financial statements and then a forecast. So all of these things need to go in uh, into. It's also known as pitch deck or you know, there are various phrases. A CIM confidential information memorandum. And typically it would take an associate, you know, junior analyst, associate person, several weeks of work, uh, to pull all this together to write it. When I was a junior person in the industry, investment banking, I did this. And then now we have people, we hire young people that do it. So the question is, how can Context AI would one be able to input a company's information and information on their industry and their financials and you know, whatever and then say, please create me a PowerPoint, an um, information memorandum, PowerPoint that would, you know, has this look and feel, has, you know, not just like picking a template, which you can easily do right now, actually doing the creative work, the writing, the, you know, pull it all together and protect the confidentiality of the company's information?

Speaker A: Yeah, I believe it can handle all of that. So Context has enterprise options where your data can be deployed in a private cloud. Only you have access to that. Again, going back to the point I started with that because LLMs are only good at language. They need to be connected or hooked into other tools to be good at, for example, creating PowerPoints, Excel analysis or graphs. So what Context and Julius and others have done is that taken the base of those LLMs, some of them are running on OpenAI's GPT or clauds, and they have built these functionalities on top of that. So you get the benefits of using a ChatGPT like interface and technology.

Speaker C: Mhm.

Speaker A: Along with the PowerPoint capabilities which we used to have in Microsoft PowerPoint or Google Slides. Now both Microsoft and Google are also trying to do the same thing. M But their technology currently is nowhere close to what Context can deliver when it comes to PowerPoints or Excel analysis or even document creation.

Speaker C: Is Context a startup or are they funded or.

Speaker A: Yeah, they're a barrier based startup. They raised a 15 million Series A, that's all less than six months back. I don't know, I mean given the pace at which AI funding is moving, they might have raised their series.

Speaker C: Right. It could be way, way beyond that now.

Speaker A: Yeah, they were founded in 2024 so it's fairly new startup and their LinkedIn shows about 50 people. They're a Palo Alto based startup.

Speaker C: Yeah. Um, so it sounds like Context AI might be very useful for someone who's creating a lot of presentations, pitch decks, that kind of thing. What about Gamma? So for presentations Gamma is another. And I know you talked about this in your, in your presentation, but tell me how Gamma might be similar or different than say Context AI.

Speaker A: So Gamma is my favorite tool. Uh, I've been an early adopter of Gamma since I think the first week they opened their product. And we have a paid license of Gamma. We do a lot of presentations using Gamma every single week. It works really well. It does presentations, it does uh, now it does social media creatives as well as web pages. So adding more functionality Context is like I uh, was sharing it. It's more than presentations. Right, right.

Speaker C: It's a whole suite.

Speaker A: Yeah. So that's how Context is different. We haven't internally licensed Context yet. We are hoping that Google will improve their Gemini integrations. And um, we have been on Google Suite for, I don't know, almost 20 years since we started Ah, 15 years. So Google Suite moving from that would be a big undertaking and that's why we haven't explored Context as a team tool. But Gamma is great. It works really well. It does a great job with not just creating content for each slide but also finding relevant images, formatting the slides. They come out really well. So I strongly recommend Gamma, but not

Speaker C: necessarily if you've got a lot of financial data to input.

Speaker A: So I actually, and for when I spoke at uh, the FEI event, I had done a ah, side by side analysis of a uh, 10K report of Nvidia.

Speaker C: Ah, uh, okay.

Speaker A: And I gave Gamma the entire 10K report and I also gave Claude that entire 10K report and both of Them created a slide deck as per my requirements. But the Gamma report, um, the Gamma slide deck looked slightly better, I would say.

Speaker C: Okay.

Speaker A: So multiple tools can do that. They can take in huge hundred page, 200 page PDFs which are full of text and convert it into a slide deck. Huh huh huh.

Speaker B: Uh-huh.

Speaker A: I'm sure context can do that as well. But Gamma did a pretty decent job.

Speaker C: Okay. Yeah. I was, I do have your presentation here. I was trying to find it, but it's perhaps not that relevant.

Speaker A: I think we ran out of time that day.

Speaker C: We ran out of time.

Speaker A: I couldn't walk everybody through the whole thing, but yeah, it should be in that video.

Speaker C: Yeah. Let me ask you. Let's switch gears a little bit. And I want to ask you something else. You mentioned there are AI assistants and AI agents. And I wonder if you could just shed light on distinguish the difference between assistants and agents for us because we hear these terms thrown around a lot.

Speaker A: Yeah. And now they are sort of getting blended again. So AI assistants like ChatGPT is a great AI assistant. They are great at doing single tasks. So if you ask them to research something and give you the output and AI assistant can do that. Gamma also has an AI assistant as well as an agent. So if I just gave raw content to Gamma and asked it to create a slide deck, that's an assistant's job. Okay. Uh, agent comes in is where they have to take multiple decisions. It's a multi step process. If they get stuck, if they are not getting the output, they are autonomous enough to make a decision and try another route. So for example, I use this platform called HeyGen for creating AI videos.

Speaker C: I was going to ask you about that.

Speaker A: HeyGen recently launched an AI agent to create videos. So we wanted to create an ad for one of our products and I gave a very high level idea to Heejin that this is my target audience. I'm looking to create a video ad which will communicate these following three things. And then it decided that okay, it will take first the agent will have to come up with the script, pick a model, find other creative assets, pick a voice and then create the video and then edit it as well. Right. It showed me the steps that these are the steps I'm going to follow. Mhm. That. Okay, go for it. And within 10 minutes it created that entire video, including all those steps. Now an assistant wouldn't be able to do that because it requires individual decisions at each point. Right. Picking the model, picking the voice. So the assistant can do parts of that. But the agent orchestrates the whole thing and brings it together. 2025 has been more about AI agents. So a lot of companies have launched their AI agents this year. Even ChatGPT has its ChatGPT agent, which actually browses the Internet like a user. So you could ask it to research 10 websites and then create a slide deck based on what it found and email it to you. ChatGPT agent can do it all.

Speaker B: Mhm.

Speaker A: But ChatGPT assistant won't be able to do all the steps and then email it to you, for example.

Speaker C: Got it. You touched on heygen, so I'd like to turn to that. And for uh, people who are listening, hey gen is spelled not like hey that the horses, you know, eat. It's like hey G E N for generative. When you were giving your presentation, you showed a really cool video, an avatar of yourself that you created in heygen that was like a little sample marketing video and it looked like you, it sounded like you. And I think people were really wowed. So tell us a little bit more about heygen.

Speaker A: Sure. So hey Jin is another startup and I think they are maybe based out of Singapore. I'm not 100% sure, but I've been using hey Jin for 18 months now. Okay. The avatar, the first avatar I created on Heygin was maybe June last year. And the initial thought was just the wow element. Right? Wow, it can do this.

Speaker C: Oh cool.

Speaker A: Yeah, I can do that now. And then we thought about, okay, how can it help our clients? And on social media, especially LinkedIn and professional platforms, you have to put out content to stay up and center in front of your network, in front of your referral partners, in front of your clients. Right. So allowed us to create that video avatar who could speak your words. Right. Creating a video, it takes a lot of uh, time and you have to have the right setup, you have to have the right tools, and then you have to use professionals for editing and so on. And with these AI twins or AI avatars, all that essentially is taken care of. All you need to do is feed it the content you want it to speak and you can create that video, uh, in minutes using hey Jen. Hey Jin is not the only tool out there.

Speaker C: Okay.

Speaker A: There are other platforms. There is a popular Bay Area startup called Tavas. It's spelled as uh, T A V U S IO. And there are at least three other platforms which are comparable. Added me to their. They have this AI pioneers list and I still don't get any incentive or anything. I just get access to their features like The AI agent we got. Being an um, agent AI pioneer, I got early access to try it out and give feedback. Um, Monday I put out an AI news flash on my LinkedIn and my substack and I put it in both text as well as video curated, uh, by me. And the video is powered by hey Jin. So I would invite the listeners to check it out on my LinkedIn. They can find if they search for AI rav. I'm sure uh, it will come up.

Speaker C: I'll look for that because now we're connected on LinkedIn. And does that mean automatically I should get a notification in my feed or email or something that hey, Rob's just put out a new agent video.

Speaker A: Yeah, it should show up in your feed on Mondays. But the algorithms of these platforms continuously evolve. Right. So unless you follow there is a bell icon you can click on and then I put out something. Thank you.

Speaker C: Okay, I'll look for that. So a question on mechanics. So you basically feed in a picture of yourself and then do you actually have to write the full script of what the avatar, what the video is going to say? Or can you just say, you know, here are the key points, here are a few bullet points and it makes up a nice sounding paragraph and whatever.

Speaker A: Yeah. So similar to what we saw in Context and other platforms, Pages also is built with LLMs in the background with the power LLMs. So yes, you can like you would prompt ChatGPT to help you with the script. You can prompt heygin or Tavas or the platform of your choice to come up with the script as well and then define scenes. So for example, when I'm creating the AI news flash video, I include the main story or the source in my background or I'll include a clip which plays within the video. So I create scenes within that hey Jin video. So you can do all of that editing as well if you have believe uh, it or not, the Arizona Supreme Court. I'm um, shifting gears and talking about the legal field.

Speaker C: Yes, please do.

Speaker A: Because even the government and the legal field is adopting AI avatars like the one we are talking about through hey Jin to deliver the court verdicts or the court updates on video. So a lot of the. I think a lot of courts put out their code proceedings on YouTube anyways. Or Arizona Supreme Court, they created two AI twins M who deliver all the code updates. Uh, I think they are using hey Jen. I'm not sure.

Speaker C: Interesting. I didn't. Did not know that even the, even the courts and the government. That's shocking. I mean we usually think of it so far behind in technology that, you know, they're not on the cutting edge.

Speaker A: And um, yeah, they've been doing that for past six months now. They, I think they in March, so almost.

Speaker C: That's cool. Um, I wanted to ask you kind of one last area and I know we're a little bit short on time, but custom GPTs for finance professionals. And I wonder if you would just help us. I know you did a sort of a golf caddy example which was really kind of fun as a custom GPT. I love watching that little video. But tell us how kind of what custom GPTs are and how they can help us as financial professionals.

Speaker A: Sure. So think of custom GPTs as your personalized version of ChatGPT.

Speaker C: Mhm.

Speaker A: You can provide it your specific knowledge, your specific instructions and tune it to do certain tasks based on your industry. So you can teach it your style, your workflows, you can upload your internal documents. You can say, this is my brand tonality, this is how I present my information or write my information. So it is a version of ChatGPT that already knows your business. Custom GPTs are free to use. Those who use ChatGPT on a regular basis, they would see a uh, GPT store icon on the left menu. Currently There are over 200,000 custom GPTs which are publicly available. Ah.

Speaker C: Uh, so like a library. Back to the other the library question I was asking earlier about Julius, I believe.

Speaker A: Yeah, you could think of it like a library. I like to think of it like an app store.

Speaker C: Okay.

Speaker A: These are all applications in a way, built on ChatGPT. Now if you want to create your own custom GPT, you do need a plus version of a plus license of ChatGPT.

Speaker C: And is that's the $20 a month version. So it's not the $200. Even someone like me with 20 bucks a month version could do it.

Speaker A: Yeah, absolutely. And you don't need a software degree

Speaker C: to do that, thank gosh goodness. Finance degree only. Okay.

Speaker A: It's really straightforward. You just need to enter instructions, upload the files, and you could start using the chat and the custom GPT right away. You also have an option to keep it private. So uh, okay, you can access it. If you want to give it restricted access to team members, then you can have an unlisted URL created. For example, I have my custom GPT called airav and my team members have access to it. In my marketing side of the business, we have a custom GPT for every single client we have. Because when we are writing content, we want the content to come out in the tonality of our client. So we feeded our clients blogs and the client's website and the specific things which we wanted to understand when it's producing content for our clients. So custom GPTs come in really handy over there. Yeah. So custom GPT is. A lot of finance professionals have already created custom GPTs which can be used by others. There are a few which I noted down and shared it with the group. The first one is super cfo. So if you just go to the GPT store and CFO you'll find it upfront. I think they have over 10,000 users and you could actually leave ratings for these custom GPTs as well. So they had a pretty good rating of 4.4 stars.

Speaker B: Mhm. Mm.

Speaker A: There are three custom GPTs which I found were there's one called Chief Financial Officer which had thousand plus users, there's one called Financial Data Analyst and there was also one called Business Finance plus Budgeting.

Speaker C: Mhm.

Speaker A: Now some of these might give you only limited use because the company or the individual who created these custom GPTs, they created them with a specific purpose. I've seen custom GPTs created by even big companies like Canva, which will get you better designs from ChatGPT, better images, but they'll give you a taste and then invite you to come register on Canva, uh, buy a license and then start using the paid version. So there are a bunch of GPTs, uh, like those as well. So I wouldn't be surprised if super cfo, for example, might give you some sample output.

Speaker C: Mhm.

Speaker A: And invite you to register on their website and use the full capabilities on their platform.

Speaker C: Yeah, and I think you know whether which of these tools can do forecasting. Now I see the fourth tool, Business Finance and Budgeting, does that. But forecasting is one of the really tricky things that finance professionals need to do both for, you know, terminal CFOs, for uh, budgeting and presenting to the board and also in the context of M and a preparing company for sale or company for fundraising. It's all about the forecast, you know, because buyers and investors, they're looking at the future, so they want to know. And creating that is often very time consuming, involves a lot of assumptions, a lot of analysis. So it seems like at least the last tool here could do, could do that, but maybe Super CFO could do that too. Or cfo.

Speaker A: Yeah, maybe. But I would always like to caution any professional using these tools, the strategy and the thinking part always be controlled by the human.

Speaker B: Right.

Speaker A: Use ChatGPT or any LLM or any custom GPT, like a really, really smart intern.

Speaker C: Uh, yeah, it's a tool so you

Speaker A: have to give it very specific guidance, what output you expect. And it does not analyze things the way we do as humans. Right. And experts in our fields. It does a different kind of analysis which may or may not be accurate. And like we talked about pattern matching, it might match some pattern which does not apply here.

Speaker C: Right, right. I think that's really good advice. And perhaps on that note, we're getting close to the hour, so I guess we should call it a wrap. But maybe we would have you back in the future for a second installment or something. This is a very hot topic of, you know, tremendous interest right now to our members, our listeners, our audience goes beyond our listeners and maybe uh, in fact get some of these specific tools, some of your favorite tools recommended by Rob Tools on the podcast. And uh, maybe I'll follow up with you and we can, you know, Gamma sounds like a winner and et cetera. So we can follow up on that.

Speaker A: Yeah, sure. Happy to come back again. And I would just like to leave the audience with some rules around AI which I learned from an expert in this space, Dr. Uh, Ethan Moleck, who's a professor at the Wharton School and he has written this awesome book called CO Intelligence. Mhm. His four rules of using AI are the following. He says, always bring AI to the table. Treat it like a, ah, collaboration. Second, always be the human in the loop. Don't leave it all to the AI. It constantly needs input and feedback. Third, treat AI like a person and assign it a role. So in finance professionals case, maybe assign it an analyst or an accountant or you could ask it to act like a cfo. But assign it a role and treat it like a person. And fourth, assume that this is the worst AI you will ever use.

Speaker C: Oh, that's only going to get better and more useful.

Speaker A: Yeah. And Dr. Mollick wrote this book towards the end of 2023 and the statement and all these four rules still stand true. AI is getting better. Every single day and every single week we are seeing new product announcements and they show that in six months from now it will be doing a lot more. So. Yes, absolutely. So assume that this is the worst AI you'll ever use.

Speaker C: So may I just refresh, Dr. Malik? M m A L I K. It's Molik.

Speaker A: M M O L L I C K. I was way off.

Speaker C: M M O L LL I C K. And the book again is CO Intelligence. CO Intelligence. Excellent. Well, I super, uh, appreciate your time and look forward to having you back again for a future installment, perhaps because as you say, things are changing all the time and evolving and all the rest of that.

Speaker A: Absolutely. Thank you for having me, Jan.

Speaker B: Thanks for listening to the Strategic CFO. To learn more about FEI Silicon Valley, go to www.feisv.org. you can find our posts on Jan Robertson's LinkedIn or the Financial Executives International Silicon Valley LinkedIn page. We'll see you next week on the Strategic CFO.