Most investors don’t lose great deals because they lack conviction.
They lose them because they see them too late.
That’s the quiet problem Ali Dastjerdi is obsessed with. And it’s why he left Insight Partners, one of the most sophisticated growth investors in the world, to build Raylu, an AI-native platform designed to help investors think faster, not louder.
This episode of Ignite isn’t about AI hype. It’s about a structural flaw in private markets that almost everyone has learned to live with.
The invisible tax on investing
Imagine trying to track 30,000 companies.
Now imagine that every 90 days, something meaningful changes in half of them.
That’s not diligence. That’s cognitive overload.
Most investor tools pretend this problem doesn’t exist. They pile on more companies, more filters, more dashboards, and quietly push the real work back onto humans. The result is familiar, constant motion, very little early insight, and decisions made just late enough to hurt.
Ali’s core insight is simple and uncomfortable. Private market investing isn’t starved of data. It’s starved of synthesis.
Why “proprietary deal flow” is overrated
Early-stage investors love to talk about inbound. Later-stage investors quietly panic about it.
Ali breaks this down cleanly. At the extremes, deal flow is abundant. Pre-seed investors are flooded. Mega-funds see everything that matters. The real battleground sits in the middle, where timing, preparation, and conviction decide who wins.
In that zone, proprietary deal flow isn’t about secret access. It’s about who shows up first with a sharper understanding of the company, the market, and the why now.
That’s not a networking problem. It’s a workflow problem.
AI that thinks like an investor, not a spreadsheet
Raylu doesn’t try to replace judgment. It tries to earn the right to inform it.
Instead of static databases, investors teach AI agents what they actually care about. Founder backgrounds. Business models. Go-to-market signals. Ecosystem integrations. Even oddly specific heuristics, like which customer logos matter or which hires signal momentum.
Those agents monitor, score, and map companies continuously, not once a quarter when someone remembers to update a CRM. Timing becomes dynamic. Context becomes default.
Ali draws a hard line here. AI should never write the final memo or make the final call. Those moments aren’t outputs, they’re thinking processes. Automating them would feel efficient and quietly destroy decision quality.
Founders aren’t convincing investors, they’re matching frameworks
One of the most honest moments in the conversation comes when Ali says the quiet part out loud.
Founders don’t change investors’ minds.
Investors recognize patterns they already believe in.
Pitching, in this light, isn’t persuasion. It’s search. The real job is finding the investors whose mental models already align with your worldview. Everyone else is just intellectually curious.
It’s uncomfortable advice. It’s also freeing.
Why AI favors new entrants, not incumbents
There’s a popular belief that AI will entrench incumbents. Ali disagrees.
Building truly AI-native products often requires ripping existing systems down to the studs. Most incumbents can’t do that without breaking what already works. Startups can.
That’s why Raylu exists. Not as a feature layer on top of legacy workflows, but as a clean-sheet rethink of how investors actually operate.
The real future of investing
Ali doesn’t believe AI turns private markets into slot machines. He believes it removes friction so judgment matters more, not less.
When discovery gets cheaper, thinking gets more valuable. When access equalizes, insight compounds.
If Raylu succeeds, it won’t be because it automated investors out of relevance. It’ll be because it gave them back the one thing they’ve been quietly losing for years, time to think clearly before everyone else does.
And in a world where being 5 percent better often means winning the only deal that matters, that difference isn’t incremental.
It’s everything.
👂🎧 Watch, listen, and follow on your favorite platform: https://tr.ee/S2ayrbx_fL
🙏 Join the conversation on your favorite social network: https://linktr.ee/theignitepodcast
Chapters:
00:01 – Ali’s background, machine learning roots, and joining Insight Partners
03:40 – Why investing felt broken from the inside
06:15 – Early startup attempts and the pull back to company building
09:10 – The original Raylu idea and why it failed
12:30 – ChatGPT as a forcing function and the reset moment
15:20 – From infrastructure to vertical SaaS for investors
18:45 – Private markets as a sales and timing problem
22:10 – Why proprietary deal flow matters less than investors think
25:30 – Teaching AI agents what “good” actually means
29:40 – Replacing databases with adaptive investor workflows
33:15 – AI as conviction acceleration, not decision-making
36:50 – What investor work should never be automated
40:20 – How better context changes investment outcomes
44:30 – The future of venture in an agentic AI world
Transcript
Brian Bell (00:01:14):
Hey everyone, welcome back to the Ignite podcast. Today we’re thrilled to have Ali Dasjurdi on the mic. He’s the co-founder and CEO of ReLU, an AI company helping private market investors think faster and decide smarter. Before building ReLU, Ali was on the investment team at Insight Partners, where he backed companies like Weights and Biases, Landing AI, DNS Filter, and some others. He’s now on a mission to redefine how investors make decisions in the age of AI. Very cool. Thanks for coming on. Yeah, thanks for having me. Well, I’d like to start with your origin story. How do you tell us about your background?
Ali Dastjerdi (00:01:42):
Yeah. Absolutely. Starting all the way back in school, kind of classically studied machine learning and data and kind of ended up in a weird role for someone right out of school, which is I started off as a large venture fund called Insight Partners. I joined a fairly technical team there, which was a great fit for me. So I spent most of my time investing in machine learning ops companies. Now it’s maybe called AIOps, but at the time it was called MLOps. A lot of developer tools, developer infrastructure companies is what I was covering. Insight’s a unique place where a lot of the model of a junior investor, when they joined that team is to meet a lot of companies, speak to a lot of founders and kind of really get a lot of reps in through the whole history and cycle of building businesses. But after a while, I knew that my kind of end goal was to get back to actually kind of company formation. And in college, I had spent a lot of time with my college roommate, starting little things and trying to build companies. And we never did it in college. But fast forward three years later, my co-founder now, and at the time, my ex-college roommate, Nathan, had shut down his first startup attempt, looking to build something new. And it was a perfect moment to start something with him. So yeah. I left my role at Insight. Nathan and I joined forces with our third co-founder, Sam, and we started Raylu. Our company is a story of a lot of pivots and change. You know, I’ll be honest with you. When we first started Raylu, it had basically nothing to do with what we are today. At the time, we were three very nerdy people deep in machine learning land. And so we set out to kind of make it easier to allow traditional software engineers to build models in their applications so that you didn’t need both a software engineering team and a machine learning engineering team to build a feature that leveraged AI. So that’s what we were working on when we first started the company. Lo and behold, ChatGPT happened. And every product team and engineering team we were working with at the time said, hey, this traditional machine learning, this prediction system, this ranking system, the scoring system is cool and all. But with a little bit of work with the OpenAI API, I can do XYZ amazing thing. And so it was a big reset moment for us as a company. And so we went through all sorts of pivots trying to kind of find product market fit again. Funny enough, we basically took everything we knew about building agents, building LM based systems, all the work we spent on in that world at an infrastructure lens, we kind of translated into a vertical SaaS app. So today, we basically build AI agents for investors. And our goal is a pretty simple one. It’s, you know, in the world of private market investors, and that’s everyone from VCs to late stage middle market private equity funds. It’s really difficult to go from, I as an investor have a thesis as to where I want to deploy capital and the types of companies that I want to invest in, to actually finding and executing and building relationships with the exact companies that let you do that. And so Rayleigh’s goal is a pretty simple one. We want to make it so that AI agents fill that gap. You tell us what you’re interested in, we find the companies, we build the relationship, we prepare you for the conversation, we help you with the research so that really the process of thesis to company identification can be as simple as possible.
Brian Bell (00:04:36):
that’s really interesting so it’s almost like almost like a marketing platform for private equity in vc where like hey i i want to go find these customers which are founders and help me go find them like help me find my icp yeah absolutely funny
Ali Dastjerdi (00:04:50):
enough is a lot of you know some folks in this universe call this business development they call them like dd professionals a lot of funds it’s just like a hybrid investment team vd role but it’s exactly that it’s like a sales function And a lot of times when we meet investors, they actually have kind of filled their stack with sales tools as kind of replacements for some of the problems that they hit in here. And we end up replacing with those sales tools for them. But yeah, it’s a perfect analog. It’s as if they’re doing sales.
Brian Bell (00:05:15):
Yeah. I mean, it is like we’re as VCs, I guess we’re selling to people, right? We’re selling LPs and giving us money and then selling founders and taking the money. And yeah, you got to fill that funnel. How do you kind of react to the sort of... organic deal flow that VCs kind of built up with their network versus I need to go out and be proactive and find things like in AI and for ops or whatever that I’m looking for?
Ali Dastjerdi (00:05:39):
Yeah, that’s a great question. You know, one of the things we say whenever we first meet a customer is how much does a proprietary deal flow matter to you and how good do you feel about it? Apparently, Brian, we meet some, especially early stage VCs where they’re like, look, I invest at the earliest stage. There are a thousand people knocking on my door to even get a meeting. And so that dynamic is like, they don’t need this. But especially when markets change a little bit. So hey, if you’re an investor that says, I want to invest in a company that you know, has 5 million of revenue, it’s doing 1 million of EBITDA, and it’s growing 50% year over year. And there’s a, you know, founder who wants to sell a portion of the business. You know, if you ask a room of investors, how many of them would want to invest in a company like that, you’ll get 100 out of 100 hands raise up. And so the dynamic changes later on in the market where it really does become who can get in front of people first, who can present the most research, who has the most conviction. It’s a little bit scale dependent. And funny enough, it’s a barbell. At the very early stages, more than enough deal flow. And at the very late stages, it’s the same things. If you’re a flagship bulge bracket investor, there’s like five private equity firms on planet Earth that will do a $10 to $50 billion take private. And there’s not that many assets out there for them to even take private. And so that universe is also very small. There’s nothing to be done there. But in that middle is where the majority of ballpark of call it 15,000 funds globally exist, where they’re investing in that intermediary journeys of companies, where it is often, especially for the best assets, a market where investors of capital are sellers, not buyers, essentially. Yeah.
Brian Bell (00:07:18):
Yeah, and it’s a pretty big problem if you put some numbers to it, right? There’s something, and you probably have better numbers than I do, but there’s something like 20,000 pre-seed fundings and 20,000 seed fundings a year worldwide. Is that kind of the number you have? Right, so there’s like 30,000 companies that you should be tracking if you’re a Series A investor, because sometimes pre-seed companies just skip and go right to NA. Sometimes they raise two, you know, three flavors of seed, pre-seed, seed, seed extension. But, you know, somewhere there’s about 30,000 pre-series A fundings every year to track and understand and get into the best deals and which ones are breaking out and which ones are on thesis and which ones help my portfolio construction. And that’s sort of the problem that you’re solving for really kind of the series A investor, it sounds like, and maybe B.
Ali Dastjerdi (00:08:01):
Yeah, absolutely. And that’s usually our target demographic is everyone kind of in that growth stage of investment. Yeah, that growth stage, yeah. Yeah, exactly. And it’s not just that you have to, you know, it’s one thing to know a list of 30,000 companies that could be interesting to you. It’s a really challenging focus and timing problem. Because one, you as a fund can’t meet all 30,000 of them to decide which ones you’re most interested in. And two, timing matters a lot. So if you meet a company at X time, three months later, they might have had an inflection point, something may have changed, it might be that they weren’t interesting or now interesting, they may be pivoted. And so it’s 30,000 companies, not that you get to chew through over two years of time, it’s 30,000 companies that every three months or so, something fundamentally could have changed with one of them that matters for you. Yeah, love that.
Brian Bell (00:08:51):
So how does it differ from other automation tools that you said you’re ripping out some of them? What is your ICP usually? What kind of systems are they running on? And then what do you replace? And do they still use PitchBook? And is it kind of you’re ripping out sort of Affinity or maybe Adio or HubSpot or name your favorite CRM?
Ali Dastjerdi (00:09:12):
Yeah, absolutely. Usually when we meet folks, they have invested in some form of private company data provider. That could be PitchBook, that could be things like Grado or SourceGrab, that could be things like Harmonic. And all of those vendors are ostensibly the same underlying technology, underlying premise of the world, which is we have a... look at certain data feeds. We have a BPO somewhere in the world. They clean up our data. We have a data set of X million companies that we cover at any given time. And you, as an investor, get access to our database. And realistically, they’re not even much of a software product. It’s a UI layer on top of a data asset is what they provide. And so usually that’s the first kind of conversation point we have with the fund. We say, look, that’s great. This is an awesome database. And some funds decide to replace that database. Some funds keep the database. Either way doesn’t make a huge difference to us. And we say the real value of ReLU is the work here from I have a thesis. Not only do you have a thesis, let’s say you’re an investor, you want to invest in next generation AI companies that are changing various back office automation tasks in finance, let’s say. Problem one is just getting a list of companies that do the thing that you just asked for is nearly impossible in these tools. They have some filters and some capabilities, but you’re going to miss half the assets that matter to you. The second problem is there is some really unique things that you as an investor look for that make a deal more interesting to you versus another investor. I’ll give you a simple example. Some of our competitors, some of the kind of traditional data assets in this market have a label on them that says like, is a good founder or is like a credible founder, essentially. And that means drastically different things to different funds. Some investors are looking for founders that have a background in the same industry. They come from certain types of X, Y, and Z. And that matters to them. Some investors think that the best founders on earth are ex-consultants that can kind of think analytically through problems. Some investors think that the best kinds of founders come through untraditional paths. And that’s kind of the second nuance of ReLU is we’re not like a static data asset where the same signal you get as one investor is the same signal every other competitor has. You essentially teach our AI agents what makes companies interesting to you. So you can tell us, Hey, I really like founders that have an unusual background, but have one signal of excellence to them. You can tell us, hey, I really like companies that have an enterprise sales motion where it’s a seat-based model. So I know they have that consistency of revenue. Hey, I really like companies that are in the Salesforce ecosystem. So if they integrate to Salesforce, I think that’s like a critical landing spot for them to be in. Hey, I want companies to really be AI native. I don’t want to waste my time on AI as an add-on type company. And so it’s really easy to codify that to us. And then our agents score, rank, identify, and thematically build market landscapes, basically chew through the data for you. And you can tell us what to monitor the company on over time. So if they can ask us to monitor, if they pivot, if they hire someone new, if the new logo was up on their website, we had asked all sorts of things. A little while ago, someone once said, could you just track whether or not Ramp is a customer? They said, if Ramp is a logo on their website, we think that’s a signal to us. And I was like, yeah, sure. Easy. Our agents can certainly identify that. And if Ramp ever pops up as a logo, we’ll let you know. And so that’s the work that we try to automate for our investors is we’re not a data asset. We’re a workflow automation tool where you tell our agents what matters to you. They chew through the data for you. We do do a little bit more than that too. We do the sales automation quote unquote for them too. So we help them write emails. We help them send LinkedIn messages. We help them get prepared before the call. Investors speak to a lot of companies back to back in different segments. So it’s just hard to get up to speed. So everything that we can do in the day-to-day workflow of an investor to make their lives a little bit better is our North star as to what we’re building.
Brian Bell (00:12:47):
That’s awesome. How much do you think your experience at Insight eventually landed you on this idea? Put your investor hat on for a second. You were just finding the founder market fit that was inherent in who you are and what your background is?
Ali Dastjerdi (00:13:01):
You know, funny enough, for about, let’s call it a year and a half of my startup journey, we really tried to not build something for this industry.
Brian Bell (00:13:10):
Right. That would be my, that would be my toughest thing if I ever went back to be being a founder or an operator, just not working on something in venture capital. Right. Cause it’s like, I’ve been living and breathing that for five years and it’d be hard like to not go build something in that, in that industry.
Ali Dastjerdi (00:13:27):
Yeah, absolutely. Thinking back to it, there was all sorts of mental blockers, I think, to even considering it for me. Some of it was just putting my investor hat back on and being like, hey, at the time, every fund has different heuristics they build. And one of the ones that was built into me at Insight was, hey, tools that sell into investment management and buy side are just not interesting. TAM is too small, not enough spend, not enough mobility, whatever’s there, we’re not going to spend time on it. And so I was like, yeah, like, is there a venture backed asset to be built here? Like, is it exciting to build in this space? And then I think the other kind of hesitation to an extent was, you know, I had left venture, I wanted to do something new. And so a lot of hesitance, but you know, I think in the end, there’s actually a very specific moment that i remember that we decided to kind of find ourselves in this journey um i can’t remember exactly what podcast it was but in various places people regurgitate pieces of yc advice that various folks have given and i guess i we were never a part of yc but i guess it’s a yc piece of advice which is as you’re pivoting get closer and closer to home and you know the basic idea there is like Get closer to the things that you know really, really well. And at some point, I had this aha moment when I heard that that was like, okay, fine. We still haven’t found PMF. Let’s go back home. And home to me was Insight. It was sourcing. It was investing at the growth stages. And I pinged a bunch of my friends. I pinged, not only do I know a bunch of people at Insight, but other growth funds that I had built relationships with over time. And I was like, hey, if I built an agent that did X and Y and Z, would you buy it? And a lot of people were like, not only would I buy it, it would be like the single most valuable thing anyone has put in front of us in a long time. And so there was a big aha moment that I was like, okay, I’ve avoided this for a long time, but I think it’s the important place to go. And as we approach this market, it actually kind of opened up my mind that the TAM question in this market is actually a really big one. It’s just a little bit sneaky, in my opinion. Yeah, I’ll just throw out some numbers. I mean, these are both private companies where I don’t know the exact numbers, but I know ballpark that both AlphaSites and PitchBook are half a billion of ARR plus businesses at this point. Wow. Yeah, they are not small, to say the least. Those are the two largest players in this market. Prequin was acquired by BlackRock for $2 billion and change. There is a lot of money to be made in this market. And in the end, it also makes sense. If you’re selling to a sales organization, you make them 5% more effective. You’re hoping that you make their overall top line revenue 5% better. in our universe, if I can make an investor 5% better at what they do, we’re talking about a very large quantum of returns that economically you get access to. And so it’s both a willingness to pay question, but I also think that there is more than enough funds in this market. And there’s also unique alternative assets that aren’t just pure funds. So today we have a customer that’s an investment bank that their BD function to transact is the same thing as an investor. Same thing, yeah. You have two customers, same exact thing. Two customers that are corp dev teams, I mean, they also have to do M&A. They have to provide GFL. So it’s a much larger universe than you’d think at the get-go. And the last thing I’ll say here, and this is, you know, we’re not there today, but this is where we think we’re going. And frankly, we think the market will go. I kind of think about this as a distinction between Cursor and Cognition. Cursor is a co-pilot that sells seats to people, and Cognition is a tool that sells an engineer as a product, essentially, as a full self-driving car experience. And we really do think that like today we’re a copilot, like you use us to make your work more effective. We really do think there’s a point here in the next year or so where we can be quite autonomous as to how effectively our agents can fingerprint what you as an investor are interested in. And then very little work interventions necessary from you for us to kind of highlight what are the companies you should be spending time on.
Brian Bell (00:17:14):
Yeah, and you might imagine a future where your agent that you’re building for the VCs is talking to the founder’s agent and doing that first screening and discovery. Tell me about your traction. Tell me about why you’re doing this and whatever the questions that VC cares about. How do you envision the future of VC? And you’ve probably done some deep thinking on this and you’re building an agentic platform in the space.
Ali Dastjerdi (00:17:37):
Yeah, it’s a great question. I think in any transition, there’ll probably be a weird intermediary period where some people will have superpowers and using AI to reach out. There’ll be some people who have superpowers and how they respond back to investors. And there’ll be kind of a lot of folks that are in between of kind of different old ways. But I do think that... end result here and what the results should be is you know in the end like you want to get to a relatively perfect market where this friction and discovery is reduced as much as humanly possible and that it is much more like a public market where you know today there’s a lot of reward to be had and kind of call it the arbitrage of what deals you get to see versus not that doesn’t exist in public markets like the way you get to be better than anyone else is that you’re smarter like you make bets that are simply better than others and i do think there’s an end point here where ai makes it so that the jumble of millions of companies the incredible amounts of unstructured information are pretty easy to chew through and consume and so that private markets start to look a lot more like public markets in the way that they transact
Brian Bell (00:18:41):
Right. Yeah. And that would be great for everybody, giving everybody more liquidity. And so you’ve actually said the future belongs to investors who combine judgment with intelligence. What does that mean in practice?
Ali Dastjerdi (00:18:52):
Yeah, that’s a great question. You know, I think a lot back to when I was an investor and deals that happened, that didn’t happen, that maybe should have happened, that didn’t, that should not have happened. And I do think a lot of it comes down to not the fact that judgment was wrong, but rather just like information was asymmetric or unavailable there. And that is, in the end, I think one of the hardest tasks as any private investor is actually having enough information to make really good judgments happen. And that’s kind of another thing that we think about as a North Star a lot is how do we make it so that at any given time when you’re in the room making a decision, either in a single pane of glass or in a memo or whatever it might be, you have the right contextual information around you to make the right kind of judgments. You know, one of the things that is a weird aha moment we’ve seen with a lot of our early customers is that we’ve basically made it, I mean, like trivial, like one click of a button trivial to have like a pretty perfect market map of any deal you’re looking at. So if you’re looking at a company, like we will tell you every competitor, every adjacency, like you know the multiple conceptric circles around the company with the click of a button. And it’s actually been really interesting to see how some investors have given us feedback back being like, this has changed some of our decision making on deals that we otherwise wouldn’t have. And the reason is just like most of the time people don’t have this context. Even as a founder today, I don’t 100% know the universe around me. Part of it’s just like a focus thing. I don’t want to spend time thinking about them. It’s also not my job. And as an investor, you do some of that diligence, you try to find what’s adjacent and what’s around it, but it’s hard work to do it well. And we’ve just gotten feedback from both investors that it’s like, this has changed our decision. And then funny enough, a lot of investors have started to push this data back to their portfolio companies because they’re like, hey, you guys don’t even have a good enough grasp of your competitive landscape to be making the right decisions. This is the right data you should be paying attention to, which has been an interesting evolution to see. But just going back to your question, it’s if you don’t have the right data, you can never make the right decisions. however good your heuristic human machine learning model is in your head. If you don’t put the right data in, you’re not going to get the right answer out.
Brian Bell (00:20:53):
Yeah. Yeah. Yeah. As an early stage investor, I mostly do precede. I kind of take the search engine category saturation strategy, which is like, you know, if... You’re an investor in the mid to late 90s investing in internet tech. And, you know, you already invested in AltaVista and Yahoo and, you know, whatever other ones I’m forgetting, Excite, Lycos, there’s a bunch of others. And Google comes along and it’s like the 17th search engine. You’re like, oh, crap. really good demo really good team i guess i’m investing in another search engine right and that would be the right decision because of course that became a multi-trillion dollar company right and i don’t know how the others did yahoo did okay i guess but um so that’s that’s kind of how i think about it it’s like if there’s a category and somebody’s gonna win most of the economic realities of that of of that and i better get in i better get into like the two or three in that category uh that could potentially win
Ali Dastjerdi (00:21:44):
Yeah. It’s funny you bring that up. I spent a lot of time talking to investor friends and I once brought up this point where I was like, I think eventually private markets will look a lot more like public markets. And they brought up something similar to that, which is they’re like, look, it’s my job to pick players in a market and pick a particular company. It was like, I really wish I could early on say, I believe in agent decoding and just buy a basket. Like I know my returns will be lower because I’m buying a basket.
Brian Bell (00:22:12):
Not really though. If you think about it, if you dollar cost average into the category and you buy every agent a coding agent, maybe there’s 10 of them, maybe there’s 20 of them, right? But one of them is going to go a thousand X. Yeah.
Ali Dastjerdi (00:22:22):
Yeah. And you know, I guess lower return than investing just in the one that goes in the one. Yeah. Yeah. Yeah, but you’d take that deal 10 out of 10 times. And I think most investors would. If you could.
Ali Dastjerdi (00:22:35):
Exactly. It’s an interesting question. And maybe it’s a cultural one as to why you can’t. Access to everything and give everybody access and float the shares that early. I think that’s some of it. I think it’s also that you as an investor can invest in competitors. And I know this is like a thorny issue and I used to run into it. And I’ve been on both sides of the aisle now. And my transparent take here is I think investors should 100% be allowed to just invest in competitors. Like you should just information share and put up walls.
Brian Bell (00:23:00):
I think if you’re on the board, it’s a little different. Yeah, if you’re on the board, I think there should be... Yeah, at least the same partner. Like you could have the same firm invest in competitors, but the same partner obviously can’t invest in the same.
Ali Dastjerdi (00:23:12):
100%.
Brian Bell (00:23:12):
I think you have to...
Ali Dastjerdi (00:23:13):
Different people, information sharing needs to be segregated.
Ali Dastjerdi (00:23:16):
But I do think very large funds procure themselves for investing in competitors because it’s just like a... It’s because of... I mean, to be fair, it’s founders. It’s that founders say like, we refuse to take your money because you’re invested in our competitor. Some of that’s an emotional take. Some of that... It’s happened to me a couple of times.
Brian Bell (00:23:32):
No, but I have a very vast portfolio. It’s almost 300 startups. So they could like go through it and go... They could usually find somebody like, oh, that... They’re...
Ali Dastjerdi (00:23:39):
And I get it. But if I were to snap my fingers and change one piece of culture between investors and founders, it might be that, which is, you know, as long as it’s not the same person on the board, you don’t share proprietary information between it, you should be able to make multiple bets in a sector and support them equally. I think that’s very viable to exist. It’s just kind of culturally taboo. Yeah, I love that.
Brian Bell (00:23:58):
What kind of decisions should always remain human?
Ali Dastjerdi (00:24:00):
Yeah. That’s a really good question. We get asked a lot by our customers to help them automate some of their bottom of funnel workflows. So making a model or writing an investment memo, maybe monitoring their portfolio. And, you know, and I know there’s startups that do that. And we very strongly say like, we don’t do that. And I genuinely as a philosophy think that AI should not be doing any of that. And the reason I think that is that the work is not a product, it’s a process. So in my mind, when you’re writing a memo, yeah, like at the end, you write a slide deck or a long form memo and that is memorialized and it’s important, it’s presented to people. Realistically, the investment team that worked on that memo did all of their decision making, their thought process, their real analytical work to understand, are they excited about this company or not? As they long form wrote out their thesis, wrote out the TAM and whatever else. And I kind of look at investors as like, yeah, I can automate this for you. But the AI is biased mostly towards confirmations. It will just paint you really rosy pictures of companies. And you haven’t done the kind of critical loops of, Are you going to actually do this deal? Same thing about building financial models. Same thing about how you want to strategically approach your portfolio companies.
Brian Bell (00:25:15):
I’ve actually trained my prompt now is on like version 153 or something for a chat GPT. And I keep backtesting it and backtesting and backtesting it with successes and failures. It does a pretty good job of good, better, best, but it doesn’t make my decision for me. It still gets things wrong sometimes, but I think it’s getting pretty good. It’s getting pretty good. I almost always agree with it now. 95% of the time, maybe. But it’s still like, there’s still a human in the loop kind of reviewing the output and saying, yes, I do agree with the traction score of five. No, I do not agree with the moat score of four. You know, it’s more of a moat of two, right? Stuff like that. That’s sort of how the little sausage making on Team Ignite, how we kind of work.
Ali Dastjerdi (00:25:54):
Yeah, no, I totally believe it. You know, I think that’s a... In my mind, what you’re doing there is you’re using an AI agent as a sounding board. It’s a sounding board. It’s helping you get to an end result.
Brian Bell (00:26:03):
It’s my investment committee, basically. Because I’m a solo GP, right? It’s like, here, what do you think? And let’s talk about it.
Ali Dastjerdi (00:26:09):
Yeah. Exactly. And I think that’s a fantastic use case for it. But sometimes we meet folks who are just like, I want the final product to be made by AI. And we’re kind of like, that is a recipe for disaster of all sorts.
Brian Bell (00:26:20):
I mean, I think as we hit super intelligence, right? And like, we already have like AI that’s 150 IQ. I bet if you throw on ChatGPT Pro and throw a data room at it, it could do a pretty good job. And you know, what’s going to happen when it’s a 200 IQ or 300 IQ? You know, it’s going to do a better job than humans, right? And it’ll be the humans just meeting the humans like, hey, like my AI thinks you’re cool. You know, like... Yeah.
Ali Dastjerdi (00:26:40):
And I think maybe the end question there is just like, what’s the data in? When I was at Insight, we also saw all sorts of deals that came through intermediary banks. And there’s all these little things that you know to look for in the kind of data and representations that an investment banker has put together at a company. I, as a human, I’m very critical of and very, I’m like, hmm, this isn’t past some of the sniff test. And like, you just know they’re painting a rosier picture than some aspects that then maybe you think as an investor would. And so that’s kind of, you know, I believe in the super intelligence to chew through the information available to it smarter than anyone else can. Today, as an example, we recently released a feature around expert network calls. Investors do tons of hours of expert calls, often, especially at those later stages. And it’s actually really hard for humans to process this information well. Like I, as a person, cannot consume 40 hours of interviews in a particularly coherent way across like 40 different people. Actually, humans are not well built for that. But AI agents are, even today, with a little bit of tuning and finagling, pretty good at doing it, even superhuman in their objectiveness as to how they analyze that information. But the one thing they’re very bad at, and they cannot replace today is that, you know, if I’m interviewing you as a customer, there’s a whole bunch of different ways I could ask you a question, each of which result in a different way of you answering me back how much you liked the product, basically. And each of those have a kind of human perception here as to like, am I pulling teeth to get that answer from you? Are you really excited in the way that you’re mentioning it? And did I lead you on in some context or another? And that is the kind of data fidelity that I think gets is lost in some of that. And so my take here on the super intelligence is like humans need to be in there either in the data gathering, the human aspect of understanding the affect between people. Yeah, there are tasks like 40 hours of interviews that AI agents are just going to be way better at.
Brian Bell (00:28:21):
It’s a really interesting question about the super intelligence. What does work look like after like super intelligent agentic swarms, right? Is it just like agents and robots building and doing anything that we kind of desire and it’s like
Ali Dastjerdi (00:28:34):
kind of a wally kind of world it’s a good question i think a lot about the i think the end result here is easy to to think about in the sense that like maybe you know there’s creative work to be done and there’s some interpersonal work to do but there’s probably a real reality where good swath of humans maybe get to pursue non-economic pursuits long story short is a world that we get to so you know like
Brian Bell (00:28:53):
i’m really good at making i just love making coasters
Brian Bell (00:28:55):
Yeah, absolutely.
Brian Bell (00:28:57):
I put whales on them, you know, like I’m just really into it.
Ali Dastjerdi (00:29:01):
Yeah, it’s like, sure, a robot could also, you know. I’m making that up. I don’t want to actually do that, by the way.
Brian Bell (00:29:06):
But you get my point, right?
Ali Dastjerdi (00:29:08):
Yeah, no, absolutely. I do think a lot about what happens in that intermediary, in that transition. What are the pains that happen? We’re already seeing a little bit of that pain in some of our engineering hiring we see today, which is long story short, like the market has decided they don’t want to hire junior engineers. And we totally get it. We also have a hard time hiring them because just coding agents are better than a junior engineer. It’s just from day one and they’re just astronomically cheaper.
Brian Bell (00:29:33):
Like a junior engineer now would be as good as, you know, like if you’re using some coding agent, reading the output and working with it and just supervising it, right? No, that’s not how it works.
Ali Dastjerdi (00:29:44):
No, that’s not how that works. Well, because, you know, you need someone who understands code really well and architecture and good design patterns and how bugs exist to guide the hand of the coding agent. And so if the person who’s working with the coding agent is as good as the coding agent in its understanding of maybe like great architectural coding practices, you’re going to get a lot of terrible AI coding slop from those junior engineers. So it actually has made it very hard to hire, long story short. And it’s caused two problems, like as we see in the market, just trying to hire talent. One is that the junior engineers who did get jobs at maybe places that let them use AI are frankly not as good in their experience because they just aren’t coding that much. It’s their ability to just get better at the job has been reduced. And the second problem is just like a lot of them have had delayed careers because it’s been so hard to be a right out of college grad into software engineering for the past two or three years, where today in the market, there’s like a there’s a huge shortage of engineers who have, quote unquote, three years of experience because everyone who would have been in that has had a really tough time in this kind of AI universe of their career. And I wonder what happens to that both long term, like, how do we have great, well trained mid level engineers if There’s no one hiring junior people. And the junior people who do get hired don’t get the skills because they’re just kind of working with AI or being augmented by it. They don’t learn the kind of hard lessons. That’s what’s happening with engineering, which is probably where I would say like AI has gotten the furthest to be able to replace people.
Brian Bell (00:31:07):
Well, maybe that’s the canary in the coal mine, right? Like maybe AI is not lifting a rising tide that’s lifting all boats. It’s a rising tide that’s drowning everyone. And eventually it’ll drown the senior engineers and the architect level engineers as well. And all that’s left is the two founders talking to agents, building stuff and selling stuff.
Ali Dastjerdi (00:31:25):
Maybe. Yeah. I don’t know if that... There is some pessimism I have, especially as someone who spends a lot of time in the architecture here, reading a lot of papers, following a lot of experts on how far transformer models can get as far as intelligence and whether or not we’re plateauing to an extent on their capabilities there. Where I’m skeptical enough of the, it will replace a senior principal engineer. My question is just like, how do you get senior principal engineers in five years from now if there’s just not enough junior people like going up through the ranks, learning through the job? And maybe this is a question about like what happens with higher education, where maybe a PhD is no longer about academic work and how do you become a teacher in research? And it just becomes about, hey, this is a structured program where you get, quote unquote, five years of on the job experience where there’s some structured ways of doing it because companies don’t have an economic incentive anymore to do that.
Brian Bell (00:32:14):
That’s a good point.
Ali Dastjerdi (00:32:15):
Maybe it becomes that if you’re going to hire a junior engineer, you get them to sign a four-year contract with you so that you know that year one and year two, you’re not super effective. You’re less good than AI for me. But it’s so hard to hire a mid-level engineer that I’m willing to spend year one and year two on you because I have a guaranteed contractual relationship with you for year three and year four. There’s creative structures, but what is it going to be?
Brian Bell (00:32:38):
It’s probably a supply and demand thing. I think the market will kind of work itself out because... As the demand drops for, you know, entry-level engineers, the salaries will drop to the point where it is economically viable to hire them, right? Instead of paying like this fresh college grad, 150, 180,000 with a sign-on bonus because they got a CS degree from, you name it, university. It’s like, yeah, no, you’re going to probably start at 80,000 a year, you know, like other college grads until it’s like economically worth it for us to hire you.
Ali Dastjerdi (00:33:09):
It’s a fair point. I mean, that’s the other way of solving that curve is just, yeah, I’ll always go down.
Brian Bell (00:33:13):
Yeah. I’ve become much more libertarian as I get older. It’s weird. Let the market figure it out.
Ali Dastjerdi (00:33:18):
Yeah, totally.
Brian Bell (00:33:18):
Absolutely. Amen. So what are some signals today that most investors are still missing because of their biases or outdated workflows or not using AI, or maybe because they’re using AI?
Ali Dastjerdi (00:33:28):
Question. I think a lot of the game for the investors, at least we work with at the growth stages, is how do you find the right companies early? Because by the time it’s obvious, it’s obvious to everyone. And so the game isn’t really about finding the right ones when it’s obvious to everyone. It’s just about being very early to And early indicators are all sorts of complicated to kind of get to. And some of the kind of early indications that are really easy to miss happen to be about markets themselves. So it’s easy. You know, there’s a lot of investors we meet who say like, you know, I was never I never thought I would invest in auto dealership software. Like that’s not my thesis. But, you know, I met something that was growing like a weed and that’s amazing. And we wanted to invest in it. And I think the thing that we see a lot that gets missed is that there’s not enough investors who essentially creatively research through a lot of markets to kind of understand, is there going to be an opportunity here where AI can disrupt X vertical really rapidly and then to monitor that vertical, find all of its players and see where they want to make a bet. I do think increasingly the kind of wait and see approach of, I don’t care as an investor what market it is, what vertical it is, whatever it is. I’m a generalist. I invest in all things technology. I just wait to see when headcount and revenue is spiking to have the conversation. I think that worked like five years ago or 10 years ago when there weren’t that many players who monitored that data, looked for that data. The market was a lot less competitive than it is now. But increasingly today in the level of competition that exists, like if you’re not there earlier than everyone else, you’re going to miss it. And the best way to be there earlier than everyone else is to research markets and have conviction in actual categories.
Brian Bell (00:35:04):
Yeah, there’s probably two ways to, there’s like three ways to win here in venture capital. The ways you’re describing, get really big and multi-stage, right? So you could just have sharp elbows, push people around and get into all the hottest series A. You can go earlier, right? Which a lot of VCs have done. They go seed and even pre-seed, which is where I sit. Or you go extremely verticalized niche, right? So you go, okay, I’m cybersecurity, that’s all I do. Or I am, you know, women’s healthcare, you know, that’s all I do. Yeah, what are your thoughts? Any way else you see any other strategies that are missing?
Ali Dastjerdi (00:35:35):
No, I mean, I do think those are those, you know, it’s funny to think capital itself is like the least differentiated thing on planet Earth you could sell to someone. Yeah, it’s just money. It’s just dollars in a bank account. It’s just money. There’s all sorts of things people say about our value add and everything else. But at the end of the day, it’s the same kind of dollars in the same bank account. It doesn’t make a huge difference. And so you have to figure out differentiation in more structural ways. And I think those are probably the three methods that are available to you. I will say, I do think in undifferentiated, highly competitive markets, being 5% better actually means you are a lot better than your counterpart. I kind of think of it as like a tennis statistic. It’s like, if you think about the greatest tennis players of all times, they like win like 55% of their points. And that sounds crazy. And that same thing is kind of true in venture over time. If you’re just 5% better, you’re likely winning the deals that are the 100xers because it’s very binary outcomes all sorts all over the place. And so I do think if you’re like a little bit faster, a little bit better, a little bit more founder friendly, whatever it is, it goes a long way in a market where differentiation is really difficult.
Brian Bell (00:36:42):
Yeah, I mean, my favorite example would be the NBA, a huge basketball fan, right? So 35% three-point shooter or a 40% three-point shooter is a huge difference, right? That’s like average to amazing, just that five percentage points. So I totally get it. So if we’re having this conversation five or 10 years from now, you come back on the Ignite pod and ReLU is a big success. What does that look like?
Ali Dastjerdi (00:37:05):
Yeah, I think the ultimate goal here is capital and the companies that it needs to back is efficiently allocated. And, you know, that sounds nice and pretty in a kind of libertarian sense of like, we love efficiency and that’s great and markets are awesome. But I do think, you know, say what you want about venture or private capital or, you know, whatever it might be. They make an enormous quantity of the economy go around. We are talking in the context of venture and some of that’s about software and enterprise and maybe it’s less interesting, but a lot of the private equity industry are the folks that own and operate emergency rooms across the country that create the variety of different services that a public school will purchase from their transportation to their special needs learning to everything in between. An enormous amount of the economy is kind of operated by the ability of capital to meet private companies, to grow them and to work with them. And our end vision is to reduce the enormous amount of friction so that The best companies doing the most for the world have capital efficiently allocated to them with capital partners who can be the right support for them. So we just want to make that world go around in an efficient way and remove the barriers that exist today.
Brian Bell (00:38:15):
So you have a somewhat unconventional route to being a founder. How has it changed your perspective on what makes a great investor? Because you were an investor first and then a founder.
Ali Dastjerdi (00:38:24):
It’s funny. Whenever I’m talking to investors, I mean, we raised two rounds of capital at this point. I’ve met a lot of investors in that process. It’s funny because every time they ask a question, every time they look at me a certain way, it’s like, I know exactly what you’re thinking.
Brian Bell (00:38:35):
It’s like, I’ve played this poker game. I’ve been here.
Ali Dastjerdi (00:38:37):
Yeah, absolutely.
Brian Bell (00:38:39):
Yeah, you can tell when they become disinterested or when they don’t really quite believe what you said or...
Ali Dastjerdi (00:38:46):
I think the biggest thing that I’ve realized, I say this to every founder that I meet, and I think most founders don’t believe me when I say it, but I try, I hope they internalize it, is that a lot of people think that speaking to investors is a game of convincing them of something. I, as a founder, believe in something. It’s my job to convince you of that. And I try to tell as many founders as possible that I don’t think a VC has ever in their life been convinced of something net new by a founder in the conversation. Maybe that’s a hot take out of me. But I think what almost always happens is that investors have frameworks, preconceived notions, ways they view about markets, themes that they want to invest in. And the question that’s looping through their head every second as you speak to them is, does this person neatly fit into one of those frameworks, one of those thesis areas, one of those heuristics that I’ve developed as an investor? If the answer is yes, then you’re mostly figuring out some auxiliary questions on like, is this the right deal for me? Is this the kind of founder I want to work with, et cetera. But if the answer is no, there is basically nothing on planet Earth that I think a founder could say to that VC to change their mind of that fact. I’m actually pretty dogmatic that that’s true. And so I tell a lot of founders, it’s just a game of finding a venture capitalist that when you tell them your vision of the world, they are already aligned in a view of the world that is analogous. You’re just trying to find believers.
Brian Bell (00:40:05):
You’re just trying to find believers that fit into their portfolio and their thesis and their framework. I mean, that’s basically it. It’s just a numbers game. Not everybody is going to buy your product, you know, invest in your startup and that’s okay.
Ali Dastjerdi (00:40:18):
It’s like great sales. When you’re doing enterprise SaaS sales, people will tell you all the time, you’re not trying to convince someone of something. You’re just trying to discover the fact that they have the pain that you think, you believe in the same world, and that you’re showing them something that they already want to buy. Same thing is true, I think, of when you’re pitching VCs. It’s just like, you should talk to a lot of them. If the most brand name VCs aren’t aligned with you, it doesn’t really matter, in my opinion, especially at the early stages. It’s just about finding believers.
Ali Dastjerdi (00:40:41):
Exactly. Yeah, I love that. I do think founders waste a lot of cycles talking to folks who are, VCs will ask as many questions of you as you want because by nature, they’re like deeply intellectual and intrigued people who want to learn. That does not mean they want to invest in your company.
Brian Bell (00:40:56):
No, sometimes spend your time with the believers. They’re curious, but they’re not going to invest. And it sounds like buying a signal, like they’re going to totally invest. No, they’re just learning. They’re just learning about the technology. I think that’s probably an on-point characteristic of good VCs. We’re just pretty curious. Like we just want to learn. Yeah, you know? Exactly. Let’s wrap up with some final questions. What’s the biggest misconception about AI in investing right now?
Ali Dastjerdi (00:41:18):
There’s been a pretty popular take that AI will accumulate value to the incumbents. And I genuinely, both in our experience and with what’s shaking out in the market today across a lot of vertical AI companies we see, is that just isn’t true. I actually think it’s like a very classic sense of the innovator’s dilemma. which is that to build an AI native product actually kind of requires you to take your existing successful product and rip it to pieces from the bones up. And almost no major SaaS incumbent has been willing to do that or successful at doing that. So I actually think that the kind of advantages of AI, you know, outside of the hyperscalers who build LLM models or sort of the compute will go not to the incumbents, but to the new entrance of this market.
Brian Bell (00:42:02):
I love that. Which company outside of yours do you think best embodies augmented intelligence rather than full automation?
Ali Dastjerdi (00:42:08):
This is a product that I just recently started to use because my inbox is driving me crazy. It’s a company called Fixer. I actually think it’s just like a very elegant solution to what AI can be doing in your mailbox. Just that every little inch of the work that you do, there’s a little suggestion, a little bit of a nudge, and a little bit of a push to make you just slightly better at the work that you do. And it’s a great encapsulation of that kind of sense of You know, they’re not automating, but they’re just making you better at the work that you already do. It’s a good product.
Brian Bell (00:42:36):
I don’t understand why I can’t log into my Gmail. And every email has been read and cross referenced with everything in my Google Drive, everything in my calendar, everything, every email I’ve ever written. Every doc is in my, yeah, like just everything. And then a perfect reply, you know, is crafted already. Hey, based on almost like a really good chief of staff would do.
Ali Dastjerdi (00:42:56):
I actually I have a startup idea for someone, which is that someone should have the quote unquote superhuman moment again with email, with AI. The reason that no one does that is what you just described as crazy expensive, like just the sheer amount of L. I’m competing. All the tokens are so expensive. It’d be really expensive. But would you spend maybe four hundred dollars a month on a product like that? I would. The answer is yes. Then someone should build the like hyper premium superhuman of it.
Brian Bell (00:43:20):
But could somebody please listen and go build that? Because I would totally, I would pay three or 400 bucks for that. And I bet like it’d probably be, you know, $10 a day in inference costs, like in token costs, you know? So yeah, maybe it costs 600 and it’s 50% margin for now, but AI gets, you know, 10 to 20 X more price performance every year. I’m guessing Google should build this now. I mean, their costs have gone down so dramatically with Gemini. I mean, I should just be able to log in and it’s just like this perfect reply is just drafted every time. And then I should be able to give feedback like, oh, well, in this situation, I would actually reply like this and the AI learns and stores that and it’s like context window. This is basically what my EA does right now. She logs into my email and she goes through, I have about 200 emails to review. from founders and deal flow and like random stuff. And she’ll copy it into ShoutGPT and craft a response for me. So I’m kind of doing it with an assistant, but it’s just, it’s not very efficient. It costs a lot. So I’m already paying to do it basically. Yeah. Hey, someone should build it.
Ali Dastjerdi (00:44:18):
If I wasn’t the founder of this company already, I’d maybe do it.
Brian Bell (00:44:21):
Yeah. If anybody’s building this, please reach out. I will fund it immediately. Not a solicitation for investment. I have to say that now. If you could rerun your Insight partner years, what investment would you approach differently? Or how would you approach it differently? I mean, you’d use ReLU, obviously, but besides that.
Ali Dastjerdi (00:44:35):
Yeah. One thing I’ll say, you know, everyone tries to be their absolute best in every conversation that they have. And especially in the inside context, you meet a lot of founders. And maybe my biggest regret is that, you know, if you’re meeting 30 founders a week, I think I was 100% there and giving it my all and did a lot of research for maybe 20 out of the 30. And 10 of them, I was like, look, this is probably not an investment. And, you know, I didn’t invest mentally into it otherwise. And I think the biggest regret I have about that is that at the end of the day, even though maybe that company wasn’t an investment, I know that for a fact, I was still talking to a founder who was an absolute expert in the world that they operate in. And we’re spending 30 minutes together. So it was an opportunity for me to learn from a world’s leading expert on something. I didn’t use it to its fullest capacity. So that’s probably the one thing I would say to my old self.
Brian Bell (00:45:21):
I love that. Who in venture or tech has influenced you most, your thinking most?
Ali Dastjerdi (00:45:25):
It consumes so much content for so many people in this world that it’s hard to pick one. I’ll mention something that I actually think is really tangible to my company today, which is that, and this is not a person, but it’s like a document. It’s the Netflix culture. Slides that a decade ago became really popular about building elite teams, not families as companies. And I think about that every day. I think about that in the culture of our company every day. It’s like the best perk I can give an employee that works at Relo is that the person next to them is the most incredible person I could have possibly found to put in that job. It puts out 100% like you do every day. I just, we’re in the middle of hiring. All I can think about is team and culture. I go back to that Netflix culture slide that it’s like, it’s just about building an elite soccer team. And the most important thing you can do for the goalie is to make sure that the defender that’s on the same team as them is also as good as them at the game.
Brian Bell (00:46:08):
What recent AI breakthrough made you rethink your roadmap?
Ali Dastjerdi (00:46:11):
Last 12 months, the cost of LLMs has gone down pretty drastically, per kind of unit of intelligence, if you want to think about it. So much so that it’s reframed what we think we can serve to our customers. And it’s created a different ethos in us as a company and probably maybe some other startups, which is Right now, we will release features to our customers that are pretty crazy expensive for us to be doing, but we just know that they’ll get cheaper, that it has reframed just the boundaries that we’re willing to build products in because the cost will go down by an order of magnitude in 12 months.
Brian Bell (00:46:42):
Yeah, that’s what we were talking about with the email assistant. Start now. I’m sure somebody’s working on it. Maybe we can get on Reilu and go find somebody working on it, right?
Brian Bell (00:46:52):
Yeah, I’m sure. I’m sure it’s there.
Ali Dastjerdi (00:46:53):
Last question. What do you want your legacy to be?
Ali Dastjerdi (00:46:55):
This is going to sound super corny, but it is actually the thing I care about most. At some point, we will sell Reilu. It’ll be a success. And I’m really excited about that day. And it really matters to me that a lot of the people that were on this journey with us that did just an enormous quantity of work with us get to have a outcome that is awesome for them and is, you know, gives them the ability to live the life and build a family of whatever else that matters to them. I know it’s corny, but it matters to me that the people that spend 100 hours a week with me in this office in New York City and this WeWork get to have an incredible outcome with us. So yeah. A great goal.
Brian Bell (00:47:27):
Yeah. As a founder, you get to touch a lot of lives, not just your investors, but their employees and their families and your customers. And that’s a lot of responsibility.
Ali Dastjerdi (00:47:35):
Yeah. Lost some lights of sleep as we were trying to raise another round of capital on that one for sure.
Brian Bell (00:47:39):
Yeah. Yeah, totally. Well, thanks so much, Ali, for coming on.
Ali Dastjerdi (00:47:43):
I really enjoyed it. Amazing. Thanks for having me, Brian.
