In the ever-evolving landscape of business, the integration of Generative AI stands as a transformative catalyst. From reshaping operational paradigms to addressing intricate challenges, the journey with Generative AI heralds a new era of innovation and strategic evolution.
This week we have the Co-Founder and CEO of AltaML, Cory Janssen. Cory is a visionary entrepreneur and co-founder of Investopedia, a pioneering financial education site established in 1999. Under his leadership, Investopedia became a go-to resource for individual investors, gaining widespread popularity and ultimately being acquired by Forbes Media in 2007. Beyond his successful ventures, Cory has played a pivotal role in fostering entrepreneurship, serving as the past-President of the Edmonton chapter for Entrepreneur’s Organization (EO). Actively engaged in the Alberta entrepreneurial community, he contributes as a member of the A100, participates in the Venture Mentoring Service (VMS) program at the University of Alberta, and serves as an associate for Creative Destruction Labs (CDL) Rockies. Cory’s influence extends to various board roles, including Brace Tool, a leading private oilfield tools company, the Canadian Brewhouse, a private restaurant chain, and Edmonton Global, a novel economic development corporation owned by municipalities in the Edmonton region.
AIM: Could you provide us with a brief overview of your company’s perspective on AI before we delve into the evolution of businesses in the Gen AI era and its impact?
Cory Janssen: I think we’ve got a bit of a contrarian bet in terms of the space where a lot of the venture money has gone into. We see this repeatable pattern where there’s a massive seed round or some significant capital that comes in. The organization often doesn’t get enough traction in terms of a unique edge in data and those workflows. All of a sudden, they kind of become a platform or try to pivot to that to save the investment.
The issue when you’re selling AI into the Enterprise, and our take is, it’s a messy business. It’s not the same as going out with a SaaS application where everything just works right off the bat. There are no good data sets. Now, that will change over time as we’re seeing more and more unique ways of getting data and kind of building on top of the right platforms. But we think that the platform battle has been won by the hyperscalers like Databricks and Snowflake. So we think that the opportunity is really in building vertical applications, getting an edge around data, and deeply understanding the workflows of a specific industry, but not trying to compete against the Microsofts, Googles, and Amazons but rather saying, ‘Hey, there’s an underlying foundation here, and we can build on top of that.
AIM: As we discuss the enabling of businesses through Gen AI, I’d like to hear your perspective on business evolution. Given your extensive experience in the industry, what does “business evolution” mean to you?
Cory Janssen: I guess we try to cut through when we talk about digital transformation or business evolution. Often with our partners, we’re trying to look beyond the buzzwords and probably just saying, ‘Forget all this. You’ve got your strategies. Technology doesn’t change strategy. You’ve got your strategy. How do you enable that with data?’ So even if that might sound weird, ban the word AI. It scares people. And in reality, when you talk about optimizing and improving business processes or reinventing business processes with data, it just so happens that we’ve got these really cool new building blocks that we might call under the umbrella of machine learning that can be used. But it should be that we’re using that technology in the data in service of the strategy and not the other way around.
So often I think we see this evolution or digital transformation almost like ‘Okay, I have to do an AI project.’ No, you don’t. What you need to do is embark on your strategy, but can AI/ML be used in that service rather than as a standalone piece.
AIM: It’s intriguing that you emphasize the importance of not implementing AI or Gen AI for its own sake but to solve specific business problems.
Cory Janssen: The taping of this is in Q4 of 2023, and we’re still in this peak hype cycle with Gen AI, which is good because it’s generating conversations. I mean, if you’re in my business, that’s a positive. But we need to always educate the business side that Gen AI actually isn’t even a type of AI. I mean, you’re still using the same underlying technologies; it’s just an overarching term. When we talk about it, it’s still a prediction under the hood. You’re still using the same underlying technology, and it’s being viewed in a different way, but it’s not necessarily new. It’s Deep learning and all this packaged in a different way, but it also isn’t all of AI. So, most organizations, if they haven’t started yet, started with the Gen AI project, are probably on the wrong path.
AIM: Are companies showing interest in integrating Gen AI into their operations, and if so, how are these discussions taking place? What areas of their business are they looking to evolve and where is the greatest potential for impact?
Cory Janssen: So I think there are two sides, as I alluded to initially, saying, “Let’s have that conversation around, yes, let’s look at use cases that might have a Gen AI aspect to them. But understand that that’s only a percentage, whether that’s five, ten, twenty, or fifty percent of the use cases out there, depending on your industry.” There’s more out there, and you should use the simplest technical solution to get the job done. So there’s always that conversation. But there’s still a misunderstanding from the business side and the average executive in terms of how LLMs are built and the origin of hallucinations.
The ability to explain to someone that an LLM out of the box does not understand ground truth can be explained in layperson terms. It’s not lying, but it blows people away. They don’t get it because everyone’s tried working with the popular tools, tried ChatGPT, and had amazing results drafting an email or editing in a bunch of these use cases where a human is still in the loop, and it gets 80 or 90% of the way there.
When you look at the industry, we only see use cases or the vast majority of use cases using an LLM on your private cloud. So in your own instance tied to that source of truth, whether that’s a Knowledge Graph or some multiple databases. In that way, we’re seeing LLMs. I’m not trying to just use LLM as synonymous with Gen AI, but in that way, really, Gen AI or LLMs are just a fundamental building block for software development. If you think of it as part of that horizontal layer and say, “Now you’re exposing all these other aspects of a solution. You’ve got this new tool in your tool belt.” I think that changes the way you can think of solutions.
AIM:How does the integration of AI models like Gen AI impact the immediate stakeholders involved in certain tasks, and what effects does it have on the broader life cycle and supply chain of the business processes it’s applied to?
Cory Janssen: I like the mental model of taking any job and breaking it down into as many specific tasks as you can, and then understanding: is there a solution here that can take out a significant percentage of time, often with a human still in the loop? Again, as you’re embarking on this, there’s a whole lot more to it in an industrial setting or in a real corporate setting than just the technical side of it.
Thinking of not replacing, first of all, getting rid of the fear and saying we’re not replacing a job. We’re going to optimize all the tasks within a job. Like, I do not believe that AI will ever replace Radiologists. I completely disagree with those who bring out that perspective. That doesn’t mean that the tech can’t get there. That doesn’t mean that you can’t have a computer vision application that is actually as good as a human. But tell me, who’s going to sign their name to the liability of making that cancer diagnosis, and are the regulatory bodies going to write themselves out of a job?
There are aspects to this in terms of putting this print into production. There’s more than just the tech. So understanding those pieces and where that fits in, in most cases, it means a human is still going to be in the loop for the majority of use cases.
AIM: Do you envision Gen AI continuing to address individual problems in businesses or evolving to become deeply integrated into the entire business process, influencing everything from strategy to execution?
Cory Janssen: In the enterprise, we’re seeing a ton of use cases around semantic search. Any Fortune 1000 company will have so many documents in the backend. You’re seeing a number of use cases there again. Some of them would be Gen AI. Someone would just be using LLM in different ways. Because we invest in early-stage AI businesses, there are a few legal tech spaces right now which we are very excited about. Maybe a couple of other examples that would demonstrate the ability to use Gen AI in terms of building intelligent agents from the sales process. I think that there’s going to be many businesses created there. We all kind of see these chatbots pop up as you go to any e-commerce platform right now and for the most part, they all suck, and people hate them. We actually think that there’s a way if you understand and have deep domain knowledge about a specific industry, there’s a way to actually build intelligent agents that people actually want to interact with. So, we’ve got a number of use cases there, especially one company in stealth mode as well that is in that space.
A lot of time they come from use cases or industries that you might not otherwise know of unless you’ve been in that industry. So, we’ve launched some early-stage investments in one company that’s actually using Gen AI to automate the reporting that engineers go through for utility companies. So if you think about a power line corridor, there’s work that’s been done in terms of modeling how those power lines change over time with vegetation overgrowth. There’s a whole bunch of factors that come in. As you are being built together and look at that from a Gen AI perspective. There’s actually you’re actually building out and trying some industry-specific software and how you’d actually model this out and build the visual representation. And then how you actually fast track and take out a significant amount of cost in terms of what it takes to do that reporting.
So that might not be something that you read about, that common use case that’s out there. But again, it comes back to what’s the business problem. This is just one element of the tool, no solution will have Gen AI aspects to it, but it’ll also have rules-based aspects and they’ll have just an overall software scaffolding that makes that you need the application and it will have elements that are many, models that are Gen AI in nature.
And so going to those almost blue-collar, almost more boring use cases, you’re going to see more things like that happen that maybe aren’t quite as sexy but actually get to how work is actually done and how, coming back to the point I made at the start of the interview, really, this is just another tool in the toolbox when you think about that horizontal layer here. What Gen AI and LLMs have done is they increase the solutions that we have beyond what was there beforehand.
AIM: As a company offering Gen AI services, how do you approach the identification of problem statements and use cases for Gen AI, and what defining parameters are you considering for successful implementation that enhances business processes rather than replacing human involvement?
Cory Janssen: In the enterprise, we’re seeing a ton of use cases around semantic search. Any Fortune 1000 company will have so many documents in the backend. You’re seeing a number of use cases there again. Some of them would be Gen AI. Someone would just be using LLM in different ways. Because we invest in early stage AI businesses, there’s a few legal tech spaces right now which we are very excited about. Maybe a couple other examples that would demonstrate the ability to use a Gen AI in terms of to build intelligent agents from the sales process. I think that there’s going to be many businesses created there. We all kind of see these chatbots pop up as you go to any e-commerce platform right now and for the most part they all suck and people hate them. We actually think that there’s a way if you understand and have deep domain knowledge about a specific industry, there’s a way to actually build intelligent agents that people actually want to interact with and so we’ve got a number of use cases there especially one company in stealth mode as well that is in that space.
A lot of time they come from use cases or industries that you might not otherwise know of unless you’ve been in that industry. So we’ve launched some early stage investments in one company that’s actually using Gen AI to automate the reporting that engineers go through for utility companies. So if you think about a power line corridor, there’s work that’s been done in terms of modeling how those power lines change over time vegetation overgrowth. There’s a whole bunch of factors that come in. As you are being built together and look at that from a Gen AI perspective. There’s actually you’re actually building out and trying some industry specific software and how you’d actually model this out and build the visual representation. And then how you actually fast track and take out a significant amount of cost in terms of what it takes to do that reporting. So that might not be something that you read about, that common use case that’s out there. But again it comes back to what’s the business problem. This is just one element of the tool, no solution will have Gen AI aspects to it, but it’ll also have rules-based aspects and they’ll have just an overall software scaffolding that makes that you need the application and it will have elements that are many, models that are Gen AI in nature. And so going to those almost blue collars, almost more boring use cases, you’re going to see more things like that happen that maybe aren’t quite as sexy but actually get to how work is actually done and how coming back to the point I made, at the start of the interview, really, this is just another tool in the toolbox when you think about that horizontal layer here what Gen AI and LLM’s have done is they, increase the solutions that we have beyond what was there beforehand.
AIM: While you see Gen AI as a tool in the toolbox, there are others who believe it has a more significant role to play. What’s your perspective on it?
Cory Janssen: I just believe they’re wrong. I mean, and you can go and, again, the professor, Yann LeCun, has an amazing Twitter feed and he’s probably the most pragmatic of it. And so, I would argue that, and we can debate the whole coming of AGI in general intelligence here and there, reasoning back and forth on it. You still need to get over the hallucinations issue, and to get over the hallucinations issue, I haven’t seen anybody do that out of the box in most industry applications. You still need to either fine-tune models and bring in other sources of truth.
AIM: Given the perspective on the evolving role of Gen AI, how do you foresee the job landscape in businesses being impacted in the near future, especially in the context of intelligent automation and the digitization of processes?
Cory Janssen: Those who use and build applications on this technology are going to thrive and disrupt those who aren’t. So, in the legal space, if the average lawyer makes half a million to a million dollars a year in the United States, I think what you’re going to see is a lot more lawyers that are actually compressed down. The average partner isn’t going to make 500. They’re going to make 350, but you’re going to see a whole bunch more that are making in the seven figures. I think you’re going to see massive increases in productivity for those who adopt and use these tools, but I also don’t think it’s going to be the end of any job.
I think that we’re going to see an increase in productivity. Anytime you reduce the cost of a specific service or good, usually, you get that demand curve right and there’s more of it that’s utilized. And so if you decrease the cost of medical tests, you’re going to increase the number of tests that are done.
So, we always talk about how this time is different, the laws of Economics don’t apply. Mark Andresen, one of the most famous VCs, has written very well. I can’t say it better than he does. But this whole idea that we’re going to eliminate jobs or eliminate work, I heard the same things at the start of the internet and it didn’t happen. I don’t think it ever seems to happen throughout human history.
What this does is it creates more opportunities for entrepreneurs than any other time. There’s no better time to be an entrepreneur than to start a business. There’s no better time to be in that space. It’s not great if you’re maybe mid or end of a career and it’s difficult to learn and adapt to these new technologies, but for those who are looking to build and those at the beginning of their career, it’s actually more exciting than any other time.
AIM: Certainly, that’s an interesting perspective, and it opens up a whole other debate about the potential for a post-work economy and the search for meaning in life when traditional jobs may no longer be prevalent.
Cory Janssen: And usually that comes from left-wing professors that have no experience in real life. When does it happen in any form of technology? You could have said it’s this Malthusian argument as well. That it’s this negative and it’s all this time it’s different because it gets to thinking. Now, I mean, just like there’s an optimistic side of this that I think you can take on if you side with the entrepreneur and the builder, and I will always be on that side versus the prophets. It’s the Ivory Tower that talks about the destruction of society. But societal questions of hey, I’m just a dumb entrepreneur trying to move in this. I prefer to take the positive side of it because I think that AI is going to save humanity, not destroy it. I think that the tools we build are what’s going to cure cancer. I think that the healthcare systems around the world are going to collapse if we don’t make decisions in a better way. I don’t think we can solve climate change if we don’t use technology and have these tools.
So when you look at what we need to accomplish as a world and health and in a climate, like AI is the solution here. So, I’m going to be making bets along those lines. And hey, if I’m wrong in the world, I’m going to live my life in a positive way betting on entrepreneurs, betting on people that want to create and who believe that this is a new technology that actually transforms us for the better.
AIM: Do you believe that AI has the potential to be an innovative solution for addressing major business challenges, and if so, what are the key challenges you see, aside from technical issues like hallucinations, when it comes to effectively implementing and scaling Gen AI across large businesses in a sustainable manner?
Cory Janssen: I still think the adoption problem that you have in AI and the trust issues they exist in Gen AI just like in any other ML project or perhaps are more pronounced. So if you’re building an ML model to make a prediction on a loan default, your output and how you test that, there’s just that one number you come up with, a percentage, and here’s your prediction and it’s more about explainability, at least you have a better sense of the explainability. The fact that it’s so difficult in Gen AI to get to any sense of explainability makes it very difficult to test and validate. And so for most organizations to feel comfortable putting that into production, it is a heavy lift.
So I think you’re going to see that much more from the startups. You need to be very careful about the type of use cases or the areas that you select when working with an enterprise because of the risk-adverse nature of it. And I would say the opportunity cost, but there’s a real risk from a reputational perspective if something really bad comes out. Which I think you can argue when you look at Open AI versus the rest of the large players there, everyone had these large models, but Open AI was able to put it out because they could take on a little bit more risk in terms of dealing with not just the edge cases but the times when it wasn’t as accurate.
So the adoption issues I think are substantial. I still think there’s a scaling issue from a cost perspective and I’m not just talking about training models. It is a material cost when you look at how you implement Gen AI and LLM within an organization. So obviously there’s the big vendors that everybody knows. Even if you’re using an LLM on something like that and putting it on your own private cloud, it’s not a small number in terms of the horsepower that you need to actually run these models. So getting over that as you’re building applications, I think that’s an engineering problem it will be solved. But this is what we’re seeing in practice. It’s the adoption, the trust factors as well as the costs or the increase in cost that you see from running these types of models.
AIM: Given the challenges discussed, what message do you have for the business leaders you cater to regarding what they can expect in the next decade and any advice on their approach to navigating the journey of implementing Gen AI to solve business problems?
Cory Janssen: I’d say the most important thing is mindset. And stop thinking about it as a Gen AI problem. Yes, this is the bright shiny object today, but understand AI/ML as a broader suite of technology. This is a very important point and one that’s captured the imagination of the world, frankly. But think of this as the internet back in the mid-’90s. We’ve now had our Netscape moment. We’ve, for the first time, seen what’s potentially possible in this field.
And so the most important thing is to be able to allocate some budget to move in to understand how to use this data, new technology in service of your strategy, and that involves being intentional, making specific, a number of bets in the space. And so getting in, starting working on projects, and diving in and building this is part of the DNA of your organization is going to pay off over the next few years. But you need to think of it in the context of that portfolio of bets. Take on a number of use cases and really understand how each project, regardless of whether it succeeds or fails, helps to improve the AI literacy within your organization.
AIM: Do you have a specific framework or key parameters that you advise or consult on when businesses approach you with the intention of implementing Gen AI to address their challenges?
Cory Janssen: We certainly have our own niche tools that we work with our partners. I’d say probably Emerj is an amazing resource that’s out there that has published the most in the space. Dan Fagella is the founder of that site and it’s probably the top leader in the industry in terms of, rather than try to rehash everything we do in a much larger context in 30 seconds, maybe refer your listeners to his site where he’s put out a ton of content on this and framework for almost every situation. He is probably the best one that I would recommend people go to. But I mean this is what we do every day. And so this is what we’d sit down and really work with the team on. There’s a lot of complexity here after seeing a few hundred use cases. It’s not rocket science, but there are lots of pitfalls for sure.