Episode 43: Mohammad Rasouli and Philipp Reineke of AIx2 on Improving Investment Processes with AI

Episode Transcript

 Shiv: All right, Mohammad and Philipp, welcome to the show. How's it going?

Philipp: Hey, how are you?

Mohammad: Thanks, Shiv, for having us here.

Shiv: Yeah, excited to have you both on. Obviously, AI is a very hot button topic in the world of B2B technology and also in investment companies and firms and how they're looking at targets and where they're deploying their capital. So excited to hear your perspective on this. Why don't we start with Mohammad, your background, and then Philipp, you can go right after that just to give the audience an idea of where you're coming from and what you're bringing to the table here.

Mohammad: Sure, so I did my PhD in electrical and computer engineering. It was dual degree with economics. My research back in the days was on AI for marketplaces, especially two sided matching markets. I continued that work and I worked in Microsoft. I worked in McKinsey for three years off of the New York office where I was managing the AI transformation of some top private equities. I’m back in Stanford about 10 months ago as a researcher, where I do continue the same research. And most of my time, majority of my time is spent on AIx2, which is AI for alternative investment with a group of Stanford PhDs and industry experience on the same vision of transforming alternative investment with AI.

Shiv: That's awesome. Philipp.

Philipp: Cool, yeah. So I'm Phil. I'm a graduating PhD at Stanford. Much of my research has been in kind of how to use AI to make managers and researchers more effective. Did some business degrees, degree in computational social science before that. Worked in consulting, lots of startups, including a German unicorn and an OIC company. Did some work in real asset tokenization. And that's how I got into alternative investment and joined Mohammad at AIx2 now. And yeah, we did a study with AIx2 in Stanford where we interviewed more than 250 fund managers on how they're thinking about AI transformation, what they're doing. So yeah, I got lots of learnings from that.

Shiv: Yeah, and this is such a broad topic, right? AI in the investment world, like there's so many different applications and places to use it. So, Mohammad, can you narrow that focus for us a little bit and talk about where funds or investors can likely deploy AI even today and see a return on it or see operational efficiencies or improvements overall in terms of their regular course of business?

Mohammad: Exactly. So when you think about this area of AI impact and alternative investment, as you said, it's a broad set of impacts, especially with the omnipresence technology that we have these days. The way that I would like to bring it to fund managers to structure the way they look at AI coming from many conversations we had is to put it in two camps. Number one is AI used for internal use cases, being a better investor. And number two is for their portfolio companies. And I would say like 2023 when I was in McKinsey, it was 50-50, like the amount of questions we got from private equities for internal AI transformation versus their portfolio company transformation of AI, which is both very important. Now, when it comes to internal transformation with AI, which is using AI to become better investors, the first camp, that itself contains of two category of use cases. The first one is the operational efficiency that you also mentioned, Shiv. It's about how I can write my investment memos faster, how I can run a due diligence in a more timely basis and more efficient way. How I can have my LP reporting, LP queries, my portfolio monitoring of the KPI monitoring of my portfolio companies, market trend analysis, all of these things, doing them in a better, faster, higher quality way with AI is use cases on the operational efficiency. And there's a second camp within the internal use case of AI, which is finding the good opportunities in the market. Finding alpha. It could be a good asset, good deal, or even a good fundraising opportunity that you are looking for as a GP. Or if you are an LP, finding a good GP that you can deploy capital to. Now, if you look at this lens of use cases as a structure, it can help the managers to decide which camp they want to start from because nicely matches to the underlying technologies.

Shiv: And where do you see the bigger opportunity? Because obviously internal efficiency is something that is important, but the entire business of PE firms or investors in general is to deploy capital and find deals and win deals, right? So is it more so in the winning deal side that you're going to see more of this kind of investment or is operational? I know you said it's 50-50, but do you see that continuing?

Mohammad: So that has been within the internal operations, within the internal use of AI, internal to the fund. Five years ago, majority of the funds who started using AI, they started from finding alpha in the market. It's predictive AI. It's using a lot of recommendation systems. It's using crunch base and pre-coin and other data sets to search for good deals in the market. That's the nature of the work they were doing. And that was a very tempting idea. Like five years ago when big, big funds, all the funds about $50 billion asset and other management started this, promise was that if hedge funds can do quantitative trading, if Amazon can have a recommendation for consumers, if Netflix can have a recommendation for media, why can we not have a recommendation for private investment, for private equities, for alternative investment? That's a very promising, tempting idea, right? But now, this last two years, just because ChatGPT, Anthropic, Generative AI has become so commercially available and has been a U-turn in the AI generally, but also in the application of AI for alternative investment, most of the funds have started from operational efficiency. They have tried to focus double down on operational efficiency, which is writing their due diligence, running their investment memos, doing their LP queries, market trend analysis, monitoring of the KPIs of the portfolio company. And that's because it's now more available, low cost, easy win, which is important not only because of cost and time, but also because of cultural readiness of the fund. If you have one, two use cases that are priority and you can win them quickly and have organizational adoption at high level, that sets the path for the future AI use cases, including the predictive ones that I mentioned.

Shiv: So let's talk about each of those things and let's start with the deal sourcing side and maybe Philipp, this would be a good place for you to jump in. On the deal sourcing side, how can you leverage AI to find better investment opportunities?

Philipp: Yeah, exactly. And I think there's like, there are different ways in which you can leverage AI and different ways in which you can leverage the different types of AI. So we're talking about AI, but AI is actually like you have different categories of it. You have the predictive AI and predictive AI is basically taking structured data and predicting something from it. And then you have generative AI and generative AI is kind of working a little bit differently. They're basically predicting words. They work really well with unstructured data. And the nice thing that you can do now with a generative AI that we've gotten in the last couple of years is you can really take this unstructured data and the structured data together and use both of these to figure out what deal you should get. And that's really giving us some advantages that we haven't had before. So the whole predictive AI that's been used in the last couple of years, that is basically limited by the data availability and the databases like Preqin and so on. And now what we can do is we can take all the available data and really, really help finding deals. And so if we've talked to a couple of VCs actually recently, who are trying to build completely, completely data-driven VC funds. So VC funds that don't do even for really, really early stage deal sourcing, they completely rely on data rather than any kind of like manual processes. And people have been trying this for many, many years. And now it actually seems to be happening. So there are some funds that claim they have success with that automated deal sourcing because of the things that Generative AI has brought in the last couple of years.

Shiv: What kind of factors are those models looking at? Is it just core business metrics like revenue and growth rates, EBITDA margins? Like how are they determining the viability of an investment and how is it doing that accurately? Help us understand that.

Philipp: Well, I think there are different things. So first of all, of course, they will look at all of these metrics. I mean, for the earlier stage you go, the less these metrics are actually available. But for kind of like later stage deals, they will look at that. And then they will be able to use kind of data that is unstructured, like random stuff online, basically, or not so random, but like newspaper articles, like different social media, and integrate that to infer missing data from this structured data, but also to kind of to enrich the data and to even kind of calculate parameters that you wouldn't even think about. So if you have like a big machine learning model, it will calculate like parameters from the data that is being fed into it. And sometimes these parameters, they respond to something that you can actually like interpret and that is like kind of human interpretable. And sometimes it will just make, make some kind of random features and random variables that are very difficult for humans to determine, but that sometimes are super like predictive in how good a deal is. Yeah. And then once you have all of this data together, the other thing that you need to do is to basically match it to the firm's, to the firm's kind of portfolio strategy, look at like, you know, what does it do to diversification and so on. So just having the data and just using AI is like one step. And then the second step is doing a recommendation system, which is like one of the things that we're building at the moment for that reason, because like you need, yeah, you need both of these things.

Shiv: How close is that? And I'm asking for my benefit and even the audience's just, how close is it to, let's say a human were to look at that same data set of structured and unstructured data versus the AI looking at it. How different is the decision making or how much faster is it? And if it's fast or fair enough, but like how accurate is it to how a human might make that decision or is it even better than that?

Philipp: I mean, ideally, ideally, it's going to be better than that. Of course, like, you know, lots of engineering is going into that. So I don't expect there to be like a database that is like, you know, perfect in the next three months or so. But then, you know, there's, there's improvement happening. So at some point, I think it's going to be, it's going to be of higher quality than what humans would do looking at this, just because you can take so much more data into account. And because at the moment, how we look through these databases is basically in a hierarchical manner. We define three or four of these parameters and we're like, okay, so we've got like, I don't know, AUM of the fund if we're a private equity or an LP who's looking for a GP even, or AUM of the fund and how long has the manager been active and so on. And these are these hierarchical filters that you can use to basically narrow down the deals. And if you do this with an algorithm, you can just filter on like much more on kind of parameters that you can't even define that are just being inferred by like, you know, how your previous deals like the previous deals that you invested in and so on. So that alone makes it kind of potentially superior. Yeah.

Shiv: Yeah, and Mohammad, this would be a follow up for you is just, you know, one of the places where my skepticism or the way I think about it is that there's a definitely an element of science, but also definitely an element of art here of trying to figure out which companies are going to be more successful or which companies are actually going to break through the noise and be able to get the kind of traction you need to succeed. So how does AI factor into that and how can humans leverage it better to kind of fill that gap? Because that element of the art component is kind of hard to duplicate, it feels like.

Mohammad: That's a very good question. Even if you talk to an investor and we have talked as Phillip said, the 255 managers, and you ask them, describe your thought process on how you choose a deal. Some of that is the hierarchy, like high level categories like AUM, geography and others they can describe to you. And part of that they cannot even point out is purely intuition. They cannot say exactly what was the feature. And that's basically the beauty of the deep neural networks as well. If you feed in enough data of the decision of the investor in the past, the neural network is going to find that complicated pattern that cannot be described in a simple term or simple metric. So the way it works in the language of algorithms and engineering and machine learning is that if you get enough historical data from an investor or have targeted interviews with them, target, show them like multiple sanitized deals, historical deals, and see if they pass or take it. From that amount of data, you can have pattern recognition with deep neural networks that in a black box way can replicate their thought process. Again, they may not be able to point out the simple metric on how they use it. And the machine will figure that pattern and what that metric is, like interpreting the machine bringing that is another set of engineering work to now go back and see what was that metric that the machine picked up. But that's the beauty of deep neural network if we give enough data.

Shiv: Yeah, I guess, and just as a follow up, some of that data, in order for an AI to be trained on it, it actually has to be available, right? And to actually be able to learn the patterns so that you recognize it in the future. I feel like a lot of the components that come from expertise or intuition or like the art side of this is not in any data set. So how does the model get around that? Because it's not even in the training data.

Mohammad: Yeah, that's a good question. Some funds do have a lot of historical data, like each UT has applied, they claim like 140K historical data, right? That's a lot of data. For the other funds who don't have it, there are two ways to do that. One is you can literally look at the history of all the successful and failed investments, even out of your fund, and sanitize them and feed them to the machine, so the machine is gonna have a pattern recognition. Then use the property data, the limited number of historical data the fund has or the limited number of expert interview we can have right now with their investors, and use that to retrain that model to their preference. That retraining, taking the elements of meta-learning, transfer learning requires less data because you have used a massive amount of historical data that is not just for that unique fund you're working with to train a generic machine for investment that picks up patterns that are not describable by simple metrics, then retrain that to transfer learning and meta learning and few-shot learning. These are just engineering methods, engineering like AI methods with few samples from that fund to now have something unique for that fund. And then as they start using the machine through a closed loop feedback, they can retrain the machine better and better. They look at a deal that is picked by the machine. They say if they like it or don't like it, and perhaps what was the reason they liked it or didn't and the machine is gonna take that into account, retrain itself.

Shiv: Got it. And what about X factors like, and that's not in the data set. Like, for example, I'd say there's, and Philipp, maybe you can take this one is, the two of the most critical factors that go into helping a business succeed. One is just timing that has been identified. Like if you having a business at the right time in a market is far more likely to succeed than at the wrong time. And then the other is the founding team or the founders or the particular creator of the business and their understanding of a particular market and what's actually going to be a legitimate solution for that market. Also, that doesn't exist in a data set. So how do you account for both of those things? I get the historic side and extrapolating from that, but what about X factors like those two in particular?

Philipp: I mean, I disagree a bit with the idea that the data on this is not available. There are two ways. One is that there's some data available on this. The timing and so on will probably be indirectly be expressed through some of the metrics that this company has and data on the overall market. If you feed both of these things in, the data is probably going to be available and the machine is going to figure out, well, you know, there's going to be a parameter that is the match of like certain things like from the market data and the corporate data for the founders and so on. You can put in their LinkedIn that is online. You can put in their CVs. And if you don't have it available, what you can always do is you can also kind of like feed in your own subjective evaluation, right? So you can go and put in some kind of like data that the investor integrates in this like, well, we have this feeling about the person or here's like a text that is describing like how I think about the deal. We've also kind of, and there we get to the human in the loop. So I don't think, and we don't think that machines are going to completely replace humans in the next couple of years. What is more likely going to happen is people will take this output from the machine or, and also like in different spaces than deal finding. And then they will work with that. So what we'll do is just going to get them 80% of the way, but in like 20 times faster. And then you do the last 20%, but now you have way more time to do it. You can focus on it really. And so investors are going to take this, and then they're going to apply their expertise. And they're going to work with the things that they get. So yeah.

Shiv: Yeah, I think that's a great point, which is just that it doesn't have to entirely replace a human. It's that if we can help investors do their job better, they're going to be better at finding opportunities and investing in the right kinds of deals by having this data available. So I think that's definitely a great point. As you see firms that start to adopt this more, does it become harder to drive alpha for the funds? And maybe Mohammad, you can take this one. As more funds are competing with AI models and having more of this data at their fingertips to make better decisions, I would imagine premium assets going at even higher and higher valuations where now you need to turn it around and grow it much faster to actually generate a return. How do you kind of see that?

Mohammad: So the question you asked about the social welfare impact of AI when multiple funds or majority of the funds start to use that and how is that going to change the industry? That's a great question. There's a lot of research going into that. The vision that we have is that AI fundamentally is going to reduce the illiquidity in the alternative investment. What it means is that now the deals are going to be made faster. For example, if you can do your due diligence faster, or you can find the deals faster, right, the investment, everything just in a more efficient and timely scale, we have a more liquid market. In the language of economy 101, if you believe in the theory of perfect economy with less friction in the market, there is gonna be better decisions and higher stake for everyone, which means the matching between investors and assets is gonna be better done. The price is gonna be discovered faster. And considering the element that in private investment, investments are unique. A fund has unique values with an asset. If it matches its thesis, its capabilities in turning around the asset, especially with that lens on, think generally AI goes is going to have improved the social welfare, better economy, less illiquidity in the market. Naturally, there are some people who lose, but the majority of people are going to win. Now, this is in the stationary state, in equilibrium. Getting there, the funds who start earlier obviously are going to have a lot of early move advantage, especially considering that AI has a time to embed, time to learn, time to even culturally understand it, organizationally understand it. Let's back to the question you asked me earlier, which use case they should start. Even the work we do in AIx2 between the two products we have, the products we have for generative AI, which is faster due diligence, faster document analysis, faster intelligence from data rows. We do offer that first because they can take it, they can quickly learn it and invent it. Then they go to the second product that Philipp was describing in extents about finding that alpha and predictive models and all the questions you ask. So starting early, think has a benefit, but in the stationary estate, majority of the funds are going to be the winners.

Shiv: Yeah, that's a great point that with more of this data at their fingertips, you can kind of invest with more confidence that brings more liquidity to markets that currently doesn't exist because deal cycles can take two months at times and sometimes even longer depending on the type of asset that you're talking about.

Mohammad: And one more thing to add to that, Shiv, is that the dynamic we observe for AI is that mega funds, as I said, like any fund about $50 billion as under management, they have started the AI journey five years ago and they do have some results. The small funds are at disadvantage, small and medium size because data science and having these data sets and state of the art R&D is challenging. It's not the bread and butter for investor for investment. So if they don't find a way to democratize AI, let's say, or everyone has an access to that, then we are going to have a more consolidated like industry generally, because those who have the power of AI and can harness the power of these powerful tools are going to basically take the advantages and all the alpha in the market, right? So it's pretty more consolidated market. But if it happens that these tools are available to everyone in the SaaS space, for example, that can reduce and cut the cost of R&D across multiple funds, that is going to be the stationary state that I mentioned to you, which is higher social welfare for everyone.

Shiv: Talk about the diligence side and the deal process side, because we've been talking about finding deals and sourcing deals till now, but how does AI get leveraged there to improve this process and make it more efficient and at the same time move faster as you find a deal that you want to invest in?

Mohammad: So like what Phil said, the power of generative AI is going through a lot of structured and unstructured data and come back with synthesized insights, targeted insights. That's the power of ChatGPT, Anthropic, and others, right? If we can build a product on top of the operating system of ChatGPT and others, because ChatGPT, Anthropic, and others, as is, is definitely not a product. It's not secure. It's not compliant. It's not even taking all the large documents that you need to take. It doesn't support all the functionality that an investor wants. It's not going to be there. I always give this example, using ChatGPT for due diligence is like using a dictionary for translation. It just takes so much time. Having a translator for translation on top of dictionary or in the language of investment, having a product for due diligence that uses the operating system of ChatGPT and others and suddenly brings a working product with high quality answers to investment, that can be immediately used for due diligence. How? Plug in directly to your data room, deal room, get all the documents, everything you have, and then you can ask targeted questions. Like the way I used to do my due diligence in McKinsey, the way I manage the teams. I sent my analysts and associates a list of 150 questions and asked them to read it, find them against the deal data room that we had, like all the files, 500 PDFs and Excels and others that we have, read them, find this. Now obviously, a human, an associate, an analyst that they had at the time, going through them, reading them, and just learning the context takes time, right? But the machine can just do that much more efficiently and bring back those targeted insights in a format that you want, in a final report format. Obviously, there's a human in the loop that's going to check it. Like the way that the manager in the McKinsey language worked, like the way I was working, I was checking the work of my analysts and associates, right? So they did the majority of the job. I had to just check and see if there are other questions I should ask, where I should double click and everything like that. In the same way, the human in the loop, the machine is going to be that analyst and associate, do all the job and the human in the loop is going to validate the job quickly and answer follow up questions and target and tailor the outcome for them.

Shiv: And Phil, this is a good time to switch to you just to help bring this to life a little bit more. As you're kind of going through that, how can you get to those insights that you would want to get through? Because yes, the machine can get through that data room faster, but then there's a ton of context that is layered within that data that a human needs to interpret. So I guess somebody is looking through whatever the output is. But how do you get through the diligence process faster with the assistance of AI tools?

Philipp: Exactly. So it's building on what Mohammad said. So I mean, what you can do is you take all the data that you have on an asset, you have on a deal, and then you ask a machine, for example, to go through all of this data and then provide your answers to whatever you want to evaluate on it. So this may be a diligence questionnaire. It might also be that you have data on, for example, if you're a GP, all your portfolio companies are giving you KPI reports and they're unstructured and you're asking the machine or can you please like tell me from all this unstructured data, what is the you know 15 KPIs that I actually want to look at like what is it? And so so in that case the machine will tell you well, they're missing information for three of them And here's the the best answer you could come up with for the for the remaining 12. And that whole process if you if you have a kind of a product that's geared towards doing this that is going to take maybe three to five minutes after you set it up first, because you can parallelize all these questions and ChatGPT doesn't take any time to read all these documents. Then you get a response back and that's basically then where you have to do the work to make this 100%. So likely this is not going to be as good as what you would get if an analyst is working for two weeks, but it's going to be like 80% there. And then you spend the rest of that day basically going through the documents, looking at the sources of what the machine provided you. So in order to make this faster, it should actually tell you what documents it used, what sources they are, so you could look at this and refine the answers. And then you basically spend one day doing something that would have taken two weeks. And in the end, you get the same result.

The other nice thing is if you're an investor, and you're like the senior investor in the team, you can actually now check what the analysts provided you because you can also look at what the machine is generated and you can be like, hey, what are the issues in this report? What are the issues, like, given this data that I have, is there anything missing? And the machine learning tool that you're using hopefully will also give you an answer to that and actually makes it much easier to check the process of your analysts. So both these things from the both of this, like being faster in preparing the reports and being able to check them better is what's going to get you to the result faster and ideally in a higher quality than what happened before.

Shiv: Do you see this application only in commercial diligence or do you see it in other aspects of diligence as well? Like financial, obviously, but there's also technical and looking at source code and other forms of diligence like legal. Like how do you see that translating across these different disciplines of due diligence?

Philipp: Yeah, so it does. It does translate across those. There are lots of different companies that actually look at these specific use cases and lots of AI companies who are starting to do that. So far, I mean, the one thing that machine learning and AI is not so good at yet is doing Excel modeling or creating financial models. That's probably going to be available in six months to a year or so. But so sometimes like what we've heard from some of our customers, like the machines, they have some issue determining like specific accounting terms or specific like identifying specific positions and like accounting documents and so on. So that is something that they reported that they tried internally before and that was kind of a problem. But yeah, machine learning is getting better and better. And I think these use cases are going to be fully available in a couple of months and then you can basically use the same process for any type of diligence.

Shiv: Yeah, on our side, we do a ton of marketing due diligence with our clients and private equity firms and we try to leverage automation and other systems as much as possible to reduce the time it takes to get to those insights. So I totally agree with a lot of what you're sharing. One of the crazy data points about this is that on average on diligence and all the different work streams involved with that, firms spend well north of a million dollars, sometimes a few million dollars, depending on the size of the asset. So through this, you see firms becoming way more efficient, I assume.

Philipp: Yeah, exactly. I mean, I think there are multiple ways in which firms get more efficient. I mean, it's probably short-sighted to just be like, you know, how much is an analyst costing me? Because what you're doing is not replacing analyst time. I don't think a lot of the firms, that's their goal or that is something that they care about too much. But you make it way faster. That's one of the things. So what is the cost of you know, having a diligence that takes two months versus that takes like now a week or so. Well, you're going to have much less probability that you're losing a deal. You're also going to not give your competitors time to bid on it. So you might get it at a much more attractive price. So these type of things are more difficult to quantify. Then, you know, are you actually like now finding risk factors that you didn't find before? That's probably quite valuable. You do save time so like people can do like you know five times more now and then of course if you're like especially when you're saying like legal diligence or kind of like commercial financial things where you would go and pay consulting firms millions of dollars that is probably something that is going to be reduced now. So all these different things.

Shiv: Yeah. And coming back to you, Mohammad, what about other areas? When you look at things like optimizing your existing portfolio or deciding when to exit, what role do you see AI playing in those types of instances?

Mohammad: So optimizing the portfolio is twofold question again. Part of that is using this generative AI to understand your portfolio better, get the data that you have and ask certain insight extraction from that, like the risks and others. That's exactly what this generative AI tools can do. And that's basically what Phil was describing. And that's what our products also that some of the funds we give it to them are using that exactly for that purpose. Now, this other side of portfolio optimization is portfolio simulation, long-term risk analysis, and portfolio correction, like finding the deals, which one you should sell, which one you should buy, which one you should swap, maybe. So that part is, again, back to that predictive AI machine, that recommendation system that we mentioned, being able to run these different scenarios and analyze the portfolio under those. That's the second camp of algorithms. That's one. Now, when it comes to the exit side, like the exit process, like the VDDs and all the, like it's twofold. It's like getting your asset ready, the reverse side of the buy, right? The selling side, getting the VDDs ready, the CIMs ready with the investment bank if you work with them, and then response to the bidders if they have any questions on your asset. So that part is very similar to the dealings that you do on the buy side, right? So again, the machine, the generative AI, those inside extraction from targeted, targeted inside extraction from those deal rooms and data rooms is something that you can do it in the same way, but for the sell side and then finding the potential good, good funds that may want to buy from you. That is exactly the predictive AI again. That's the second machine. As you see, like through this investment process, like endpoint, you can see the power of both generative and predictive AI by defining very carefully and a sharp way to use this. In every step of the investment, you can see how the generated AI and how the predicted AI can come to the picture.

Shiv: Right. And do you see that on the optimization side and on the exit side, with investment banks and how they play a role in this process, how do you see their role evolving as data is being leveraged more and AI is being leveraged more to figure out when to exit, what value to exit at, like finding potential buyers? There's a lot of redundancy there in terms of what they do.

Mohammad: Yeah, that's a very good question. Yeah, we were at the Finovate Conference and I'm going to give a keynote speech in their fall version in New York and I guess Phil is going to talk on another session on AI for them. And that's exactly the question they ask, like those investment bankers. Like what is our use of AI right now? And what is our future role if AI is going to come to the place? Right? It's a very critical question that those people ask. And part of that was asked from even it was McKinsey days as well, like having vision for the industry. Now, the first question is easier to answer. What kind of AI I can use as an investment banker internally? Exactly the same as you mentioned, find the good buyers for an asset, like who is the potential fund to approach in a more targeted and individual way rather than just sending the CIM to everyone, right? And just spam all the inboxes. Getting the CIM ready by itself is a lot of work that an analyst do in those investment banks and that's something that the generative AI can do the same way that we're talking about document generation and due diligence and targeted insight extraction, all of these things. And this process like the investment bank does is like, is very similar and doing that in a faster way, like in a targeted way, faster way result in a better matching of the market is what the investment bankers can do with AI, right? Now there's a price discovery as well that you mentioned. It's a very interesting set of algorithms like price discovery for an asset, which is predictive AI, lot of interesting engineering work that can happen there. Now, this is just a short version of the answer. Like what can AI do in this business? In the long run, I don't think investment bankers are going to go away. I think they have, there's a lot of human elements to this business, like matchmaking and deals and convincing the parties, the both parties. That human element for the entire alternative investment is going to stay.

We were at SuperReturn like in Berlin, there were 5,000 mix of GPS and LPs just a couple weeks ago. And we had a dedicated session to AIx2 which represented all of this AI, case and others. And you can see in that environment, in that conference, just people meeting each other and making handshakes and making like connections. That is important for investment, especially for alternative investment. That is not going to go away with the machine. But now having a more targeted conversation, faster deal processes and others what makes the entire process more efficient.

Shiv: Right, right, totally. Yeah, so in essence, the ecosystem stays intact, but it becomes way more efficient, there's less waste, you can move through things faster, cheaper, and get to better answers and have a better chance of driving alpha within your investment. So that's a great way to summarize it. So I think with that said, I think it's a good place to end the episode as we're coming up on time here, but before we do that, what's the best way the audience, if they want to learn more about what you're up to, how can they learn more about you?

Mohammad: Yeah, definitely happy to as part of the things that AIx2 is trying to do is to stay in educating the community, the firms, the fund managers. Also, we have our ongoing survey, which measures the pulse of the market and we see how funds are changing. And there have been massive changes in the last nine months, massive developments in the community. So if you're available on LinkedIn, if people want to reach out, I'm there, Philipp is same. AIx2 web page is there, AIx2 dot ai. Just feel free to go there and connect with us. There's info at AIx2 dot com as an email address. There's also our personal, which is mrasouli at AIx2 dot ai and preineke at AIx2 dot com.

Shiv: Awesome. We'll be sure to include all of that in the show notes. And with that said, Philipp, Mohammad, thanks for coming on and sharing your insights. I think there's a ton of great takeaways as the investors and founders and other bankers that listen to this episode will take a lot away from in terms of how they run their respective businesses. So thanks for coming on and doing that.

Philipp: Thanks so much for having us.

Mohammad: Thanks, Shiv.