Ep. #8, The Promise of Generative AI with Josh Furstoss of Incued

Rachel Chalmers: Josh Furstoss is a FinTech founder, startup advisor and startup community leader. His background is a combination of sales, strategy, software engineering and expertise in the intelligence and security technology industry. While studying at Northeastern Illinois, Josh worked as employee number one of Supply Line, a SaaS based ERP system for hospitals and critical care facilities.

After three years at Motorola he has become a strategy advisor to half a dozen startups, a venture partner with executive counsel network, the executive director and head of the Chicago chapter of Founders Network, and maybe most importantly, founder and CEO of Incued, automating financial reporting while gathering insight and intelligence for VC, private equity and banks. Josh, thank you for coming on the show. It's great to have you.

Josh Furstoss: Thanks, Rachel. I'm happy to be here.

Rachel: As a startup founder, what does the current mania for AI and machine learning mean to you? Do you plan to incorporate, say, gen AI into your product?

Josh: So you were talking about Supply Line a second ago, Supply Line was an AI company actually. We were doing AI for the supply chain back in... this was probably 2014, 2015. So one, we are interested in putting AI into Incued in various different ways, but it's really interesting to me to see that AI is popping off in this big way. But it's been there for so long, and maybe the question is generative AI right? It's the thing right now that everyone is getting crazy about.

Rachel: Yeah, exactly. I think because ChatGPT just hit this inflection point and became able to pass the Turing Test, everyone is equating that with AI. But as you say, we used to call this big data and before that we called it stats and before that we called it Bayesian Inference. It's not something entirely novel. It's the Gen AI that has brought it front and center to the conversation.

Josh: I think it'll change the ecosystem a lot, but I also am hesitant to think of these as fully formed companies in a way, that are making generative AI.

But the more interesting companies to me are the picks and shovels companies that are enabling it from the base of it, and then I think the companies that are going to make the biggest profits or be the biggest changemakers are actually going to be companies that gather unique data services, read them out, and then leverage generative AI on top of it as a standalone feature.

Especially in the way this is moving with OpenAI's app store. It's more a feature in my mindset than it is an entirely new product. OpenAI is the product, or LlAMA is the product. This is just a new way of interpreting data that I think will become the market standard, almost in the same way that we had software before the iPhone came around, but the way that you interacted with the software dramatically changed.

Or like when Google Home and Alexa, that came out and we could interact with software through voice, this is just a new paradigm or a new medium by which we expect the user experience to work. I don't know, I think I'm pretty excited about it, but I'm pretty excited about things that haven't happened necessarily yet, or are just on the bridge of happening.

From a feature set standpoint, auto GPTs are really, really interesting to me and I want to see where those go. But for Incued, what we really want to do because we take in all this financial data is figure out how to do scenario analysis as well as sentiment analysis. What I mean by sentiment analysis is, okay, so you're going to take in all these Twitter posts and all the news articles on the web around a company, all this kind of qualitative information and then you can squeeze down all that qualitative information to gain unique insights.

That's really powered by the fact that these models can access the internet and that feature by itself is basically entirely displacing or will eventually displace entire companies like Grand Watch who have built a reputation on sentiment analysis. I almost think this is going to empower really, really well made SaaS solutions more than anything.

Then the companies that try to bill themselves as just a GPT with a twist, I think that in much the same way that kids were making Flashlight apps for the iPhone or things like Flappy Bird, I actually think that's going to be like an entry point for software development. Not necessarily something that will gather a magnificently large amount of funding because there's no moat to it, the data is the moat.

Rachel: Yeah. I like that analysis and I hope you're right because to me the real promise of gen AI in particular is to democratize access to what has been pretty arcane levels of data manipulation. If the value really is in the differentiated datasets, then it gives scrappy founders like you a chance to pull together a really differentiated dataset as you're hoping to do and build something really valuable on top of that.

Josh: Yeah. I think it helps existing large enterprises as well. Imagine this, imagine you're like Blue Cross, Blue Shield and you're trying to figure out which plan you need to have, PPO and whatever or whatnot, what premium I should have and all these different coverages. By talking to something and it telling me, "Here are the best options," instead of looking in... Honestly, Healthcare.gov could use this.

It would be probably great for them just to have the average person find these things, I think that all helps small and big companies alike. I think that small companies seem to just look up on the data tree of being like, "What have people not exploited yet? What do people not have?" And then make access to that information really easy, but paired with something else.

I'm starting to believe in this paradigm of all software is, is taking two things that haven't been smooshed together and smooshing them together, and then making it more accessible than it was before, and then connecting it to other things.

It's like all we want is, effectively, roads that connect the knowledge ecosystem, and we want faster cars that drive on the roads. It'd be nice if my fast car could also have the functionality of a hatchback or a pickup truck so it could do more than one thing. That's a bad analogy, potentially.

Rachel: I think you're thinking of trains. Trains can do that.

Josh: It is what trains do, yeah.

Rachel: You're anticipating my next question, which is around how might generative AI in particular or machine learning in more generally, change the way software is developed and deployed? And what will that mean for early stage founders like you?

Josh: I think it'll make rapid prototyping dramatically faster. I don't think it's just for startup founders, I think it'll start in startups but it'll move upwards. I think that we'll be able to create companies so much more capitally efficient, where actually the average raise amount I think will go down. Not just in accordance with the fact that we're in an economic plow right now.

But more or less because you don't need as many developers or even as many marketing people or sales people or customer support people in the future, so you'll have very, very lean teams that don't have a ton of hierarchy. There's maybe like one or two layers of management in a company that normally would have like multiple. I think it's probably 80 to 90% improved efficiency in these companies within 20 years.

Rachel: It's interesting that your mind went to the hierarchy in the small organizations, because I hadn't specifically thought of that angle. But I think you're right, we've already had folks on this show who are working with ChatGPT like it's a junior engineer and pair programming with it. But the idea of that VP of engineering within an early stage startup being able to do a lot of individual contributing him or her self, and then having a much smaller team and getting a lot further with it, it reminds me of then the cloud appeared. Suddenly you didn't spend your entire Series A on Sunny 10K, you spent it on an AWS subscription and it was a tenth of the size and you got way further with it.

Josh: Yeah. I think that the hierarchy is really important as organizations scale, I think a lot of the blockages to getting tasks done is effectively communication between levels. It's like how many people or pushback does this idea need to go through before we can go to a yes/no decision? If you can make executive decisions very, very, very, very quickly, then everything moves much faster.

I feel like you have a lot of employees that are just waiting around in some of these big companies because they can't make executive decisions and they're not really working on highly productive things. But if the CTO can just continuously make executive decisions around the product... Imagine this, having AI design user interfaces or do Figma mockups very, very quickly.

The head of engineering isn't usually great at design work, but by getting the design work out of the way and maybe you can take that Figma... I've seen things like this where you can take a Figma file and convert it into something you can actually place directly in software rather easily. I think it'll cut down dramatically the amount of work that needs to be done, or the amount of heads.

It's like every single person... You think about D&D, in D&D different characters, there's a bunch of skills and some skills you have proficiencies in which means you can do them better, and the other skills you can have double proficiency which means expertise. One person has like 10 skills of the 20 skills, and the other person has the opposite n skill so you overlap in a few different areas. Instead of needing 10 people that are all with unique, different skills and you're all speaking different languages to each other.

Rachel: Yeah. That's super interesting. I mean, it all comes down to rapid prototyping, but you're talking about prototyping things that we haven't thought of before, like prototyping ORM charts and prototyping go to market and A-B testing.

Josh: All the admin work. It's all the admin work inside of the large orgs and startups which I think is the biggest blocker to productive labor because it's the things that you don't really want to do. It's like when we started having more fractalized roles, like fractal CMOs and CFOs and things like that. A CFO isn't a great value add for a startup, they don't really have a full time guy yet so it may as well be on the outside and they can serve like 10 companies.

They live in a marketing agency or a design agency. I think you'll see this being more and more and more things fractalized and eventually you'll have... I see it almost as a bunch of little puzzle pieces, all the companies are little puzzle pieces and they all can fit into one another for the job that needs to be done at that individual little time, like a fractal. Yeah, I see it as tons of little micro businesses is our future.

Rachel: Yeah. I'm almost working in that mode already, so yeah, I can see it.

Josh: What's so interesting to me about it is that you have this generation, millennials and then you have Gen Z and the upcoming generation after them where people more and more are focused on freedom. That may mean freedom from work, don't have a 9:00 to 5:00, less working hours, four day work weeks, all these kind of things.

But for me what that comes down to is people want more autonomy, they go to school to become creative thinkers, education is higher among the average of every single generation we're going to have to treat everyone like an individual, effectively. As to the lot more freedom, I think that they'll engage in capitalism from the standpoint of creating their own companies to achieve that freedom and it's like this used to be just in the Gilded Ages, big industrialists, and then it's like software was the real thing that changed this in the long term.

It was like we can make a giant company with only a few peoples' labor, but now I'll think you'll be able to do everything. A precursor to that I think is the Shopify stores and these Etsy stores and all these little micro businesses that you're seeing popping off because people want to provide something themself and have a fairly unstructured life where they're in the thrill of it.

Rachel: Are you worried at all about the risks of widespread use of commercial Large Language Models? Especially in financial applications like yours. Are we at risk of baking algorithmic bias into our system?

Josh: So for our approach, I don't think so. I'm sure everyone says that, but the reason for that is I ultimately believe that humans should be the decision makers at any large financial organization like this so we're not having AI execute on ideas or execute on trades or do anything like that. Our goal is to provide people with the intelligence that they need to make decisions effectively and then we're solving for that on the AI side by having incredibly high levels of data security.

That said, I think there is concerns to be had about things where the AI is taking direct action an... I mean, it's simple enough to move past. I think that continuous automation becomes scarier and scarier and scarier, but I have pretty high trust in society that we'll overcome that and we'll figure out a way to make this safe. But it is a worry.

Rachel: It's refreshing to hear someone having faith in society these days. What are some of the ways you think we might be thinking about mitigating those risks? You've talked about human in the loop in your application, that's obviously a powerful one. What else should we be thinking about?

Josh: I know regulation continues to come up and things like that, but ultimately I think that software will be created to solve this problem. There will be software platforms that exist that check models for credibility. In the same way that you get SOC2 compliance or anything that you can put the little badge on your website, I imagine in the future that you'll have almost AI model compliance for the level of accuracy, and you'll display that on your website like any other kind of data security. We'll create almost like a private standard for what is acceptable with an AI model deployed in certain use cases, be it government grade or be it commercial or consumer grade.

Rachel: Sounds good. How do you hope that AI might improve the world?

Josh: I think there's dramatic ways to improve the world. I was speaking about personal freedom earlier and this is probably the biggest thing I think for any of us is just our time back. The more time that people have to be creative, even just to think. A lot of people who say they work in a factory or ARM or something like that, and you don't have this much time to contemplate and to maybe just be human.

I've been minimizing it a little bit, but I think it'll allow us more time to spend on endeavors that make us feel like us, and less time on endeavors that bore us to death, to be frank. Hopefully more time with our families, more time with our kids, more time making memories, and more time using our brains, and less time doing repetitive tasks.

Rachel: That sounds really good. What are some of your favorite sources for learning about AI?

Josh: There's a guy named Wes Rah on YouTube that has some very, very good AI videos. I think anything Lex Friedman is a staple, everyone knows about Lex Friedman.

Rachel: We'll put all these in the show notes, folks, so you'll be able to look those up. All righty, it's daydream time. I'm going to make you God Emperor of the Solar System, for the next five years everything goes your way. What does the world, the industry look like in five years?

Josh: I just want people to be happier. I want everyone to have a friend. I think that's the saddest thing that's gone since COVID is the lack of community and things like that, getting people together.

Rachel: That's interesting. I would say my online community has gotten way more ride or die since the pandemic. I think a lot of what used to be on the open web has retreated into private forums like Slacks and Discords. But I think those fora are relatively thriving.

Josh: I agree with you there, and I think I was well adjusted to it. I grew up playing World of Warcraft, I'm a child of the internet, effectively. But the more I talk to people that are younger than me, that went to high school during COVID and things like that or went to college during COVID, there's a lot of social anxiety out there and there's a lot of anxiety of depression.

Kind of like how Instagram affects young girls and things like that, so I don't really like that. I'd rather people be able to live freely as themselves, without the whirlwind of politics is culture. Yeah, I think I just want people to be more free. There's some other things that I want, like I'm really concerned about the state of water and fertilizer in West Africa.

That's something that I've talked to you about before, where I think there's going to be a fertilizer crisis and then specifically people in Nigeria are really going to suffer from that because their population was basically grown on having all these farming inputs that might not be available in the future. I don't know. We're in a rough time, and I just want people to be happier, I think. More holistic.

Rachel: Yeah, I think that's why you and I get along so well. One last question, Josh, if you had a colony ship to fly to the stars, what would you name it?

Josh: I was thinking about this actually, earlier. Have you seen Battlestar Galactica?

Rachel: Of course.

Josh: Okay. This isn't directly correlated, but it's somewhere. I'd have two ships and I'd call them Adam and Eve because I think it'd be an interesting way to restart humanity on another planet, where then it almost goes back to the birth of Western culture and how we conceptualized the beginning of time from that kind of perspective.

Rachel: Okay. A little heteronormative, but it's original and I like where you're going with this. Josh, thanks so much for coming on the show. It's been a delight as always, and good luck with everything.

Josh: Thank you.