Ep 150: How AI Is Transforming Diligence, Decision-Making & the Future of Investing with John Melas-Kyriazi

AI is a technical co-pilot for diligence is actually one of the most interesting

use cases for AI as an investor right now. If you come across a really interesting company,

you can get really, really deep within the matter of hours and then pull in human experts to help

you to go the last mile. The mission for our business is to accelerate innovation in the

private markets. If people had access to better data, they could make better decisions. Is it

changing how venture capital firms operate when they're using tools like cloud and chat GPT

to poke holes in them? What am I missing? Where's my argument weakest? Investors for decades have

not had access to the kinds of tools that they need to really operate their businesses in a data

driven way. But let's talk about how AI is changing things right now. You can go into Excel,

hook it up with cloud, hook up cloud with standard metrics and then ask you to do a discounted cash flow,

it will just build it for you. It's amazing. Keep going to be able to spend more time on what they're

truly passionate about. John Milas-Karyazi is CEO of standard metrics. Over 10,000 innovative

companies uses technology with hundreds of venture firms to do their work and figure out what's

going on, how businesses are doing, how to interact with them. John is on the leading edge. He's

running a really high growth SaaS company. He's becoming an AI company. It's really interesting

to hear from John about how AI is transforming business, how it's transforming finance and what's

going on in the innovation world. Welcome to American Optimists. Today we have my friend John Milas-Karyazi.

John, thanks for joining us. Thanks for having me. John, you're a co-founder and CEO of standard

metrics. I think it's a good example of what's happening AI with agents right now with all sorts

of other things. So I want to talk about that. But first, tell us a bit about yourself. Where'd you

grow? I grew up in Brooklyn, Massachusetts right outside of Boston. What were you into as a kid?

What kind of stuff? So I was really into reading fiction and then over time more and more science

fiction, fiction and then science fiction because you studied physics eventually. Eventually, yeah.

I think science fiction was maybe a precursor to that. And your master's was a material science.

That's a different area than you're working on now. Is there cool stuff happening in these areas

you studied? Is that something you eventually want to go back to? You know, I don't know if I'll

go back. I loved it though. And it definitely inspired me a lot, I think, as a kid. Being interested

in science, I got really into electric guitar in high school. And then that drew me into electronics

and circuits. And I think I freaked out my parents because I ended up setting up a soldering iron

and printed circuit board etching station in my basement. That kind of was my first foray into

getting into engineering. What were you trying to build in your basement? I built a tube amplifier,

a vacuum tube amplifier. And I also built a bunch of guitar pedals to help with reverb and distortion

and stuff like that. So you're an engineering kid, a physics nerd, went to Stanford, you started

a PhD, you dropped out of a PhD. How come? That's right. You know, I was really lucky in my sophomore

year in undergrad. I ended up at this material science labs, Mike McGee, who's now a professor at

CU Boulder. And he was researching next generation photovoltaics. And I had this incredible PhD mentor,

this guy named E. Kong Ding, who kind of took me under his wing. We ended up publishing a bunch

of papers together. And it was this really cool experience. Papers on what? The first one was around

a new manufacturing process for a new type of solar cell we were working on. And the second was more

of a device physics paper. So it was understanding really the loss mechanisms in the solar cell. Like

when the sun shined on the solar cell, what happened to all the electrons and holes, which are

like the kind of part to electrons? And where were they getting lost and what was degrading the

performance? So it was a really cool experience. And I thought, gee, like, this is awesome. I should

keep going. I should get a PhD and become an academic. So that was where I was originally kind of

pointing myself. And then, you know, I had a bunch of friends that went and worked at startups.

I went and worked at a startup for a summer. And when I got back to my PhD, I realized that I

wanted to be doing startups instead. It's interesting. You know, I think this is a really common thing in

our society right now, where I think it used to be maybe 50 years ago, 100 years ago, the smartest

people who wanted to pursue the most intellectually demanding and interesting things would stay in

academia. Because it was just more interesting there. And there's more challenges there. And there's

things you could do there. And now I feel like the last 30 years, especially the last decade,

startups have become so much more compelling. And there's like so many smart people there. And you

can build and solve so many problems. You actually get like more of your, you know, Aristotle would say

that like the that one of the things is most satisfying for man is to use the highest use of

their faculty applied to successfully. And I feel like that's now more in startup world sometimes

in academia. That was definitely your experience. That was my experience. And look, there are

aspects of academia that I absolutely loved. I think that there was an element of intellectual

freedom exploration. That was incredible. It's actually something I love about being an investor too.

Like post academia, I spent a couple of years helping run this early stage startup accelerator

called startups. And then became a VC. And I was a VC for six years. You're a spark. I was at

spark. And that was one of the things that I loved about being at spark was this idea of

being able to explore many different things, being able to go very, very broad and then choose

when you went deep. It's very different from being a founder, which I also love for very

different. It was interesting that because you like that about academia, but then in spark,

you're able to do even more, even more things that sounds like and go even broader, right?

Yeah. I think as an academic as a researcher, I felt like I had a lot of freedom to explore

different areas within a certain field. And part of the goal was to become the world's expert

in that field or in a subset of that field. As a VC, and I think it's actually something that

AI is now changing, which I talk about with you. It's possible to have a much, much broader scope

in terms of the types of businesses that you're looking at, but then kind of gear up and start to

get really deep technically where needed. The AI thing is actually interesting. I feel like I've

always had a little bit of advantage because I do go pretty deep in a lot of areas, maybe not as deep

as the PhDs, but I'd go deep and lay deep. But now with AI, everyone could just pretend, right?

And by the way, it's useful for me too because I'm like, oh, how does that? How does that really

nuance deep tech thing work again? And you just like ask it and then all of a sudden you're

like asking really good questions. It's, yeah, dangerous right now. It is. Yeah, I think like AI is a

technical co-pilot for diligence is actually one of the most interesting use cases for AI as an

investor right now. I don't think it's a replacement for human experts necessarily, but it certainly

is a huge help. If you're, if you come across a really interesting company in a field that you

know a little bit about, but you're not an expert in, you can get really, really deep within the

matter of hours and then pull in kind of human experts to help you to go the last mile. Whereas,

you know, five years ago, it would have been this kind of mad scramble kind of calling people,

trying to assemble folks that might have taken days or weeks. That can be done in much faster now.

I actually, one of the hacks that I really enjoy when I have free time is to ask the AI,

like, what are the most interesting and important or cited papers or new papers or controversial

papers in a field? And then explain these to me as if I only have an undergraduate degree. Like,

so dumb it down a little bit from like the most sophisticated PhD level of whatever it's saying

and explain it how it works. And you actually can pretty quickly kind of learn what's going on

in pretty much any area. For sure. Yeah, I think that as a VC having the ability to tailor the

output of a model to meet you where you're at from an understanding perspective is one of the most

useful aspects. Opening up a 10, 15 page long academic paper full of citations is extremely

intimidating. It's so, it's so annoying these papers and it's like they purposely, I think it's

like they purposely sometimes make them so you can't understand this. I always felt about

a manual cons and some of these philosophers, I mean, pretty convinced they're purposely doing it

to like, so I'm not one of the initiated. There's like a proof of, there's a proof of work there.

You need to put in the time and effort and they like that. They're purposely putting the barrier in

so the AI can remove some of that period. It's not anybody. It's dangerous because you probably

are missing a few things still. Yes. So you were helping people build things. You were an investor

and then you went on and I was involved, obviously, in founding what's called standard metrics now,

which is this awesome, really quickly growing company. You have over 10,000 companies now that

standard metrics report on for people. That's right. What was the idea behind the company? Yeah. So

when I was at Spark, you know, when I joined the firm in 2014, the firm had already been

fabulously successful. It's like taking Twitter public, Wayfair, you know, Oculus Assault to Facebook.

It was a really interesting time to join the firm because it was a group of very technical people

who were really interested in investing in cutting edge products. At the same time, the firm

didn't really have core software that it was running to operate day to day. And you know,

when I got to the firm, I was actually surprised by that. I went to talk to friends in the industry

and started realizing that very few investors in the private markets really had access to great

software. It was this kind of cottage industry that had suddenly grown a lot and become very

competitive and very global, but software hadn't yet cut up. And so I started going on this quest,

actually ended up meeting Ray and Shubi from Affinity as part of this. I'm looking for software for

our firm that we could use. We ended up buying that software and using it to try to run and improve the

way we were doing relationship intelligence for data. So kind of discovered firsthand and some

of the stuff that I went through when I was there was around, you know, for example, I have an

LP meeting coming up. How to collect data from over a hundred companies, clean all that data,

have it be apples to apples, use it for a bunch of modeling exercise and reporting exercises.

When you say apples to apples for people, it's basically you want to show off the metrics where

you actually compare the company's progress to each other. That's right. For example, let's examine

revenue growth rates across the entire portfolio. Let's examine, you know, growth profit margin

across the entire portfolio. And so going through those exercises, even things like, hey, let's

keep a list of the companies that are pretty low on cash to make sure that we're jumping in to help

them. Those things were incredibly manual and time consuming. And those experiences really inspired

me, kind of gave me a prepared mind. Then when I ended up meeting you Alex and Danny and the team

eight VC who are also working on this problem, kind of gave me the inspiration to go and start

this business and try to build a platform to automate and improve portfolio reporting, reporting

management for the private markets. It sounds like pretty SOTAR, great portfolio reporting,

management for private markets, but there's there's literally hundreds of thousands of the smartest

people trying to solve these problems. And so you're kind of supporting how many customers do you

have now? How many people use this? Over 150, 150 firms. 150 firms, yeah, and well over 10,000

companies now on the other side of the platform, you know, and it's neat because if in any,

obviously we built to solve one part of the problem. This is like a, this is like the other

probably biggest part of the problem for all these firms. And it started in venture. Are you doing

private equity now as well? We're doing some private equity now as well. I say, especially in growth

equity, which is kind of an interesting in between looks venture style in some ways, but then also

they also do buyouts. You know, we see, you know, the mission for our business is to accelerate

innovation in the private markets. We think that ultimately a lot of folks spend a lot of time

doing things that ought to be faster and more data driven and more automated. And we also

think that if people had access to better data, they could make better decisions. So portfolio

companies, for example, when they report data on state or metrics to their investors, they get

access to aggregated and anonymized benchmarks that are relevant to companies at their stage and

their sector. The 10,000 companies actually in some ways are kind of your customers as well.

Absolutely. Absolutely. Yeah. And from our perspective, if we can help those companies to build

better relationships with their investors, have a better dialogue with their investors around

their numbers and also have access to better data, we think that they're going to improve their

probability of success. And then on the other side, you know, investors for decades have, we think

not had access to the kinds of tools that they need to really operate their businesses in a data

driven way, which extends from the, you know, the back office and processes like, you know,

working through the audit and valuations and LP reporting all the way to, hey, our portfolio

companies coming in, they're raising another round. Should we invest more? Should we invest less?

Should we participate a lot of my stuff right now? It's like I was with the CEO last week and

send a new AI company that's, you know, a couple of years old. He's like, Joe, we're not just

doubling revenue this quarter, we're tripling revenue this quarter. Yeah. And that always sounds

crazy. It's like, really good. But I'd be interested to know, okay, well, I'm sure there's like

another 50 that are doing that too. Like, how does it compare it based on the stage it's at?

What are those raising rounds out? Like, so like, you're based on the algorithm, like, what is

this worth? Like, actually, don't even know. Yeah. Like, it's all changing so quickly. Even for

me in the middle of it, probably as much as anyone, like, what is that worth? And it's like,

you would know a lot better than me because you have all the data. Like, yeah, like, like, how does

that workflow work? Yeah, for sure. So I think it's a couple things. I think one is,

um, getting all of the data in one place. And I think the way we think about it is that

firms need a mechanism to collect and store all their information on their portfolio. As we said

before, kind of in an apples to apples format, that's clean, that's auditable. Um, that's,

that's kind of traceable back to where the data came from. Um, they need a way to access that data.

We think about it as, you know, we have this deep commitment to interoperability at our company.

So we think that when an investor has their data, it's their data. They should be able to use it,

you know, with their own AI agents that they're bringing to the table or with software that they're

building on top of standard metrics or in a spreadsheet or on our platform. Um, and then lastly

is, uh, kind of robust tools that help them to leverage that data as well as the corpus of data

that exists in an aggregated anomaly. Yeah, because everyone asks what that's worth. I need to use

your aggregate data to see to compare other things. Absolutely. So we think that the most powerful

mechanism is if a firm has great intelligence on their portfolio company, they understand all

the numbers. They also have, uh, written, uh, artifacts, notes, partner commentary. Um, and they

also have access to what's going on broadly in the market and how that's changing over time.

They can take all of that data to bear to make the right decision for their firm. And that's

something that we, we think is an opportunity for firms ultimately to generate more alpha,

not just creating like having really smart initial investment decisions, which is obviously

mission critical, but then also, you know, many firms take 20 to 50% of the capital they're fund

and reserve it to invest in their existing portfolio companies. And we want to help them to do

that better. So let's talk about like how AI is changing things right now for firms. Obviously,

you guys are seeing what's going on. You guys are seeing the numbers. Yeah. Is it changing how

venture capital firms operate? Like, like, like, how are people using AI in general? Yeah. It's

changing how firms operate, um, all the way from the very, very earliest stages of sourcing and

identifying companies all the way through things like, you know, portfolio management LP reporting

on the other side. Um, it's actually really fun right now. I say, when we started standard metrics

with you guys, you know, there was in the, in the initial phases, a lot of what we were doing

was evangelizing to firms like, Hey, you should use software. Like software will help you to run

your firm better. And, um, you know, now that's still happening, but there's also a lot more

that we're learning from firms. They're experimenting a lot. So for example, on the sourcing side,

for the last decade plus, there've been a bunch of firms that have built data-driven sourcing

engines. They're looking at things like GitHub stars, a number of employees on LinkedIn,

social media followers. We're now seeing people use large language models to do a lot more qualitative

research that compliments that quantitative. For example, like, what's the tone of what people

are saying on Reddit about this company? Um, or much more nuance is a good if they're attacking.

I mean, I mean, those guys are pretty crazy. Maybe. Maybe. Well, there's, there's also, there's

also a lot that people are doing around the analysis of people. So for example, if you take someone's

GitHub profile for their kind of public facing work and you take their LinkedIn profile,

can you start to build a map of the type of founder that you want to identify it back early?

That matches with your investment thesis. So we're seeing a lot of really interesting experimentation

there. Um, on the diligent side as we discussed before, um, you know, AI is a technical

super powerful. Another thing we're seeing is that firms are writing investment memos and then

they're using tools like cloud and in chat GPT to poke holes in them. What am I missing? Where,

where's my argument weakest and then using that to help to guide the diligence process and then

feeding that back in again. So you start to get into iterative mode on the portfolio management

side. I mean, we have customers that are doing wild things right now. For example, we have one

customer in Singapore that's using an agent orchestration tool and it's building this wide

variety of AI agents that interface with standard metrics, grab data and fulfill a bunch of really

critical internal reporting workflows. Um, for example, an LPS a question, an agent automatically

gets triggered reaches into standard metrics, goes back, composes an email, flags it for a human

and then it sends it off to the LP. Things that you wouldn't have even considered a few years ago

are now starting to get built. Just all automatic basically. Yes. People are talking to agents

more than they realize. Oftentimes with humans in the loop. And I think human in the loop has

been a really important part of our business, particularly on the data processing side and making

sure that data is correct. Um, and then yeah, when it comes to, you know, the sort of technical

diligence side, this another area where I think the AI is really helping people to push deeper into

into deep tech. Can you not teach the AI to like use the common sense check to make sure the data

is correct? Like, like, like, it's like how long do people going to have an advantage in that?

Isn't it like a common sense framework you can give another agent to do the check? I think

there will be. I think there will be. And we're certainly seeing even just the, the, the,

the kind of core models, the reasoning models get better and better and better at identifying

issues. Um, I think that when we think about the future, when data is really well organized,

um, for example, the internal AI agent that we've built, which we call our AI analyst, um,

when the data is really well in the analyst that's part of standard metrics that a firm could use

if they have standard metrics. That's right. Effectively you have a name for it. We call it AI

analyst. Not the most creative name in the world. I guess standard metrics, there's no like

little, like, standee or something. You know, like, you know, like, you know,

adipore has Addison, right? So that's right. I tried to like, yeah, standard metrics. This has,

AI analyst, we just, yeah, we decided to not anthropomorphize the AI. It's not an agent.

It's an agent. That's fair. But we, maybe we could do some branding help there. But the, um,

the AI analyst, when it has extremely structured context with really clear harness in guardrails,

it's unbelievably accurate at pulling data across large amounts. So for example, you could have

hundreds of portfolio companies and ask a question like, um, make me a table with my top 10 fastest

growing companies where I own at least 5% that have at least $10 million in revenue and include,

like, these metrics. And it'll produce it with very, very high fidelity. We do want our customers

to always go and like, double check the data. But that works really well. What's tougher is feeding

in a completely unknown document and having AI extracted well out of the unknown documents as

well. We do, we do that currently, but we use humans in the loop to make sure that we queue.

And that's something that's just like, it's not, it's not a fully solved problem. Obviously,

it is not a fully solved problem yet. We think it will be in the future for now, having humans in

the loop dramatically improved. So this is where a lot of kind of cutting-edge stuff's happening,

like one of our companies, like with real, like hand-drawn landscaping documents, they can get

it right, even though it's like different every time, because like it's so smart enough to like

figure out that they're applying me to plan, given what they're doing. They're applying me

to this type of plan. But it's a very hard problem, right? So you have to have a person check for

most of these things. Yeah. By the way, Chris thinks Stan is a good name for us. Stan? Okay.

Cool. We'll take that into consideration. So internally on your team, this is where you

were all living in AI world. I think there's like probably a few hundred thousand of us that are

part of this in a lot of ways. And obviously everyone in society's part of it, but a lot of people

are listening and they're not building things they are right now. Like, how's it different for you?

Like, tell us about this. Like, you have a team. Obviously, you have a lot of technical people on your

team. Like, how is the AI changing how your company works internally? Yeah, dramatically.

And it really feels like the rate of change is increasing. Every day, the sort of the product

releases, the conversations we're having, it's super exciting. And it's a lot to stay on top of.

I see the first place where we saw it change the most on our team had to do with data operations,

as I mentioned before, we sort of ended up building a human team to go and work with our customers

and help them to process our data. And the first place we launched AI agents in our product was

to assist the data operations team. Now, AI is doing the vast majority of the work. So that's

been a big, big, big shift. Have you cut people or you could do new things now? How does that work?

Because our teams grown a lot, we actually have continued to grow that team, but we're just handling

a much larger volume of data. So it didn't need to grow the team as much as you would have. Exactly.

Do do the same thing. And also what we tend to do with this with this team called data solutions

team is they're oftentimes focused a little bit more on breaking new ground. So we're really good

at processing data, data from financial statements. We just launched a board deck parsing with AI.

And so our humans are like much more involved in running that. So someone sends me a board deck,

like right now I have agents like helping us with emails and stuff too, of course. And like,

you know, in the loop, obviously, when you hear back from me, it's not the agent, but it's

healthy. So if the board deck comes in, that's actually smart. Like there should be like something

where our agent knows the gift, make sure Sandra metric automatically gets it and just automatically

goes in. That'd be that'd be a good work. And the data is extracted, but doing that from a board

deck is really hard, right? From a financial statement, it actually is quite hard because

the range of different formats and currencies and fiscal board decks, even harder in some ways,

but at least this usually should be labeled like usually, but oftentimes data is in charts. So

then you need to use AI to start extracting approximate values out of those charts. So there's

a sort of a different set of technical issues that crop up. Well, super annoying when it's a chart.

There's not label perfectly or something. That is also like to your guess. But so data operations

is the first area. The second big area for us is engineering product and design. And that's been like

a, what do you guys use? What do you use? We use a bunch of different tools. So we use on the

engineering side, you know, cursor, we use cloud code, we use this tool called AMP, which is,

you know, in many ways, it's also a GenTX similar to cloud code. We, what we've ended up seeing is

engineers adopt some of these tools. Product managers and designers say, hey, that looks cool.

I should try this too. And suddenly designers and product managers are doing very different things

than they used to. So for example, we're redesigning kind of a core part of our app. And one of our

designers built a fully functioning prototype using cloud code. I love it. I love how you can just

vibe code. It's something that actually works with a designer. It's amazing. And it's not just like a

mock that you can look at. It's actually something you can play with. You play with and get feedback

on it. You can start to identify edge cases and corner cases that aren't being met necessarily.

So that's changed a lot. Velocity has increased overall and also the ability for people to be creative

and contribute to areas outside of their traditional core domain has dramatically changed,

particularly in product and design, being able to reach in a little bit more, contribute a small

PR pull request, for example, for maybe something looks visually off and there's a little bug in

the platform. That might be something a designer could identify and fix and then engineering needs

to review at the end versus adding it to the stack of all the different tasks that the engineering

team is working on. And then of course, there's a bunch of other places in marketing. One fun thing

that we've seen is the way people find standard metrics is also changing. Interesting. So are there

agents finding it now? Yes. Well, I don't know what extent it's agents, but certainly it's people

using, you know, chat GPT and Gemini and cloud. How do I do this workflow? Oh, well, this thing out

here does is like our friends at Versel, Guillermo was I guess this the power like tens of millions of

things, like the vast majority of the time it's used as infrastructure now. It's an agent that knows

it's the best thing. And then there's like a really interesting question of like how do you put

things online to make sure the agents like totally identify and know it. So you're you're thinking

about this? Absolutely. Yeah. And where I actually vibed coded my personal website with Versel.

And and it was it was a ton of fun. So yeah. So we we actually use software that helps us. It runs

hundreds of prompts per day across four or five kind of major LLMs. It helps us to identify exactly

where standard metrics is getting mentioned, which pieces of contents that are linked to standard

metrics, the LLM is referring to and helps us identify gaps to improve the rate at which we're

getting. So you're actually creating some content out there to make sure that they can see it.

We're creating content specifically for LLMs that's also human readable and useful for humans.

But it's really it's really engineered to make sure that LLMs understand what we do. I love

it. It's really helped by the way. It's a brave new world of marketing to the computers now.

Yes. Please excuse me. And then you know, so people talked about the SaaS

pop clips was a big term last month. Yeah. Where I mean, this is obviously this is a high growth

SaaS company. Yeah. But it is like, but also you're moving into AI and your AI company as well.

Yeah. So should people be afraid of the SaaS pop clips? Like what are the certain companies? You're

actually your bearish on because of it. But then like, how do you think about it? Yeah. I think

that the SaaS pop clips like most things there is there's a there's a lot about it that's correct.

And there's certain things that are probably overblown. I think that the thing that the most correct

about the SaaS pop clips is getting us to question all of our fundamental assumptions around competitive

barriers to entry for software companies. Historically, the number one, you know, competitive

barrier to entry for software companies has been switching costs. You buy software,

you build a bunch of workflows on it, you get a bunch of users using it, you get a bunch of data

into the system, and then it's very painful and expensive to switch to another platform.

Because of AI, that's changing. It's cheaper for competitors to build software. It's cheaper for

people to vibe could their own software. And it's also easier to migrate data because of AI agents.

So I think switching costs have gone down. And so I think it's right for the way that software

companies, the way that software companies speak, value to change in light of that. It turns out

though that there are other competitive barriers other than switching costs. So for example,

for our business, one of the major competitive barriers that we have is around network effects.

Specifically, you have 10,000 companies. Exactly. We have companies that use our product when

we bring on a new investor, we're talking to a new investor, we just closed a new customer

yesterday. And they had like 40% of their companies already using. So a new customer comes on

and it's like a ton of their companies. Our data is already right there. Exactly. It's really

easy for them to onboard. It's also really nice for their portfolio companies to be able to use a

tool with more than one investor. Of course. So that's one of the KPIs that we track internally.

It's like how many companies are sharing data with multiple investors in the given month.

So network effects is one. Another one is data. So for us, I've mentioned before, we leverage

aggregated and anonymized data for the benefit of our users. We help to offer insights to our

users from that data. That's something that's also quite hard to assemble and quite hard to build.

You can't like vibe code a network. You can't vibe code a new data set. And people are starting to

use this obviously. I know we use it. How easy is it? I'm going to push you off line on this. How

easy is it to like check what valuations might be based on data? Yeah. So that's a workflow we

can make really easy. So right now our AI analysts will do that for you if you ask it to. I feel like

you should like make this like a thing that just automatic. I know. I know. We had a conversation

with an auditor recently about this too. That was really excited about it. I think that there's

going to be a lot of opportunity for innovation and automation around valuations. Leveraging AI plus

benchmarks plus proprietary company data as well as public market comps and other 100%. And I feel

like this is like super controversial because everyone gets pissed if you're like given the

valuation they don't want. So you probably have to have multiple models like under this framework

is this. Under this framework is this. It should be really nice just to have a starting point.

Agreed. You need multiple models and you also need auditability and traceability. Like so I

think that's actually an example we're having like a more opinionated deeper harness for the agent

is really helpful. There's some cases where Claude's harness that it's built is so good that

can just do a lot of stuff out of the box. Like one thing that we published recently was

you can go into Excel, hook it up with Claude, hook up Claude with standard metrics and then ask

it to do a discounted cash flow for a portfolio company. It'll just build it for you. It's amazing.

Which is of course, by the way, there's nothing to do with valuation in venture these days.

That may be the case. It might in 5 or 10 years. That's right. That's right.

But for something like valuations, you could imagine building like much more opinionated.

Like Joe is involved plus five. This other person's involved minus one.

I think it's exciting though. I mean, one of the things that we're I don't know if you're

other portfolio company. I've actually be curious. There's other examples. But one of the things

that we're starting to see is that our product development, we're also like discovering a lot of

capabilities in our product that we didn't know our product was capable of. Interesting. Because

it's pretty cool that way. Yeah. We're like, oh, we should try to do this and see if this works

in it, but then it does work. And then we tell our customers about it and they start using it.

There's probably like a lot more product marketing that can exist in an hour old that doesn't

right now. I'm seeing some of my companies because there are just like all these capabilities,

like all these things and then no one even knows. Totally. And we're like much more actively

investing there. We're doing a lot more visual content for users where we just we created a I don't

know how much like you guys other companies are using MCP sort of protocols that exists. Explain

that to our listeners. Yeah. So and we think of MCP effectively as a protocol that allows AI

like large language models or agents to to really effectively use an API. So for standard metrics,

for example, we have a really robust public API. We also have an MCP. And so if you want to hook

up standard metrics to a large language model, it can figure out how to use our product really.

And you're just letting anyone plug in. Do you have to be careful as someone's getting

with you, is going to try to plug in and just like steal stuff. So so yeah. So obviously the data

privacy and security side is critical. I'm making sure all of that works really well. There's a

there's an opportunity with MCP for large language models to be able to go and you can get super

creative with how to use products. You can also start to combine different products together.

Is it model context protocol? That's right. Yeah. We're all a context protocol. I should have

looked that up. But which is by the way, very funny. Because in the in the past, APIs are like the

super APIs. Yeah, rigid programming interfaces, super program, super like rigid. Yeah. It's like a

program. And it's defined exactly. You submit this and you get this and it works. It's exact

weights. It's a programming language, whereas model context protocol is like models are funny.

It's like you just talk to them. It's just like explains what's going on here. And what they

look at it. And they use that to kind of figure it out, which is a fun example. Yeah. So like so

we so we had a so we when we started exploring this, we had this hypothesis that we could do a

bunch of really crazy, interesting new workflows. One of them was, okay, what if we hook up with

notion calendar and standard metrics and then you ask a question, it's probably relevant to you.

What are all of my upcoming board meetings over the next month? Create a notion page for

each of them. Go into standard metrics, pull all the latest data and build me a like quick

summary and an agenda for each of those board meetings. And it'll go and it'll understand what

a board meeting is. It'll go into calendar. It'll look at your calendar. It'll pick out. Okay,

these are the six board meetings coming up over the next month. It'll go create those notion

pages. It'll grab the data and they'll build those reports. Now that might need to be tweaked by

someone on on your team or by you to get this would say my chief of stuff a lot of time. It could

save the chief of staff hours through a single prompt. So the ability to be creative like that,

I think is one of the really new fun things about building software. And how quickly is everything

getting better? Like it's given us a sense of this. Like there's some metric we're like clawed

now as like the agents are able to work twice as long. Yeah, interference every three months,

which is kind of insane if you think about it because eventually there's going to be able to work

forever. And I guess they're also getting faster. Like what does this mean? Like over the next

year, how does your business change with this? Yeah, for sure. Yeah, there is that metric that's

sort of yeah, the it's like what there's there's a task that has a 50% chance of successful completion.

How long does that task look like for a given model? I think clawed is now like 20 hours or something

like that. And it was at like, you know, one hour a year ago. Yeah, it's pretty remarkable.

I think for us, one of the biggest areas where we see the models getting better improving our

business has to do with the extraction of more and more and more structured data from a longer

and longer tail of different types of written communicators. You're still to bring everything

together. Everything together, centralizing everything and giving it access to the AI

analyst as context is is actually relatively straightforward now because of the size of the

context windows. How much data the model can kind of be fed. What's trickier is the extraction

of structured trustworthy data from that. So that's going to be an area where we think they'll be

to be rapid improvement. And then we also think that the tooling around everything is going to get

better and better. It's going to be easier to orchestrate work. Some of that orchestration will

happen within applications like hours. And some of it will happen outside of applications like

hours and we'll reach into these apps as a key data interface. Some people are referring to this

as like a headless model. And I think it's really important for companies to be thinking and building

with that in mind and making sure that if their users are extremely sophisticated with how they're

thinking about building their own internal AI agents that your company can be a partner for them

and not be sort of in a combative position. Sounds like work is going to get a lot more efficient

and effective. You know, it's like a dangerous subposition, you know, his canes when he saw

all the proctivity and wealth coming over the next, I guess he was 90 years ago. Now he thought

we would just only work 10 hours a week because he's so wealthy, we wouldn't need to work. And it's

like it seems even more so that way right now, although I guess I guess there's like reasons why

people might want to work harder and still make tens of millions of dollars. But it's just he's

saying that you're not going to need to be doing as much work as much done, right? Maybe. Yeah,

it feels like the opportunity cost of not working is so large right now though because there's

so much happening. That's the air side of you with a little bit of the little bit of time you

can create so much. The opportunity cost is like, oh, you should actually work more, which actually

is my life. I'm obviously working really hard because every little bit of time I could do something

to help cure disease or help fix this thing that's broken over here. And so every little bit of time

work, you're kind of compelled to work more because you get so much done. For sure. I think the other

side of it is something that gives me like a lot of excitement about the future is I do think that

the infrastructure around running a business, a lot of the infrastructure that's almost

frustrating and annoying to deal with, things that most small business owners don't want to spend

time on, they wanted to be done right. But I think like, you know, accounting or setting up a

website or, you know, customer support queries that are coming in. It just happens now. You

make those things can just happen now. So I think what we're seeing now, which is really exciting,

is this explosion of small businesses getting started. And I think that people are going to be able

to spend more time on what they're truly passionate about, probably spend more time on things that

directly interface with other humans, whether it's producing products for them or working with

them or helping them with something versus, you know, jobs that were more about paper-pushing

that are going to become more automated. So that is something that gives me a sense of excitement

about the future. I love it. Well, it's a very optimistic case of the future pushes back

against some of the businesses we're seeing from others. John, thank you for joining us. Thanks for

having me.