Why Airtable’s founder is betting on teams of AI agents

0:00

Howie, it's great to have you on the show.

0:01

I think what you did uh with Air Table was one of the most interesting deals this year, selling it to to Bending Spoons.

0:09

And obviously, you're the the founder and CEO of Air Table and you've got a new company called Hyper Agent, uh which we're going to get into, but first, can we start there with with the sale because >> uh at least in the tech world, everyone paid attention to that.

0:22

It was super interesting.

0:24

I mean what bending spoons is doing generally is interesting but also your decision to to sell and then and then now be doing what you're doing with hyper agent.

0:31

Walk me through that decision process.

0:35

>> You know I ultimately think that companies are built um you know at either the beginning middle or end of a certain technological era software companies at least.

0:43

And you know when air was being founded when we started this company back in 2012 it was really kind of at the beginning of a lot of trends beginning of PLG like up to that point like Dropbox and Evernote were probably the only known products that were pure PLG.

0:57

Slack had not yet launched.

0:59

Uh you know the whole B2B SAS like PLG movement had not um you know happened yet and and really the advent of like the browser becoming this really powerful platform for you know kind of uh rich desktop grade applications right?

1:12

So you know um back in you know 2012 through 15 is the first time you really started seeing like robust single page app experiences that could you know kind of replicate desktop grade experiences.

1:22

So you know Figma, Air Table, Mirro etc.

1:24

A lot of these companies kind of were were formed of that era because of this new technological capability, right?

1:29

And you know, obviously there was a lot of like business thinking that went into the original Air Table idea, but I think it was the right place, right time.

1:38

And you know, if you kind of fast forward a decade plus to today, this was back in 2012, maybe through 15 to now, a lot has changed, right?

1:47

changed, right? I mean not not just AI but like the whole SAS revolution happened you know kind of that ushered in you know sort of economics and like you know kind of a way of thinking about that type of business you know and yet I think we've obviously entered into this new era where AI is clearly the most potent kind of technological you know uh

2:06

foundation to any new business and it's really kind of rewritten as with every era uh the rules and and kind of the structurally you know kind of good or bad properties of a business And so all that is to say, you know, when I look back and see Air Table today as a, you know, kind of a great business, we got to nearly half a billion in revenue scale. We were generating a lot of cash

2:27

We were generating a lot of cash flow.

2:29

Uh, we were still growing and had a really high quality product and, uh, you know, customer base, etc.

2:32

And yet, you know, even though it's got really good bones, I saw it as being in a very different uh moment from, you know, at the beginning or even the middle of the the kind of like SAS golden era, right?

2:46

And so, you know, when I see the opportunity ahead uh in AI, I think it's it just redefineses a lot of the the um the rules and you know, the pricing power expectations, the way that you think about distribution uh in SAS.

2:57

And, you know, ultimately we had a choice in running Air Table.

3:02

We we could either continue to try to execute optimally on this, you know, fundamentally SAS business, right?

3:09

You can obviously add AI on top of SAS, but ultimately, you know, a lot of the underpinnings of a SAS business are rooted in the SAS era, right?

3:16

Seatbased pricing for one, right?

3:18

And a seatbased product, it's kind of all about employee productivity enhancement versus, you know, kind of potentially highly variable token spend or complete autonomy uh completely automating, you know, certain processes with AI.

3:29

And um you know and secondarily like there's there's just a different stage to air table partic uh in particular as a business because we had reached this kind of you know more mature point than obviously when we were starting out and just building initially a great product that could get some traction.

3:43

And so in retrospect I think there was a fork in the road.

3:46

We could have continued running you know air table and and you know building it into the best possible SAS business we could have in the AI era or you know we could find a buyer for it.

3:56

Um, and you know, in this case, Bending Spoons is a great highquality buyer.

4:01

Uh, I had a huge amount of respect for them and their team as I got to know them and you know, and how they thought about things.

4:07

They're extremely high integrity.

4:07

I think they uh really care about buying quality businesses and holding on to them forever.

4:12

That's that's their model.

4:14

Like they're not just here to like, you know, flip it and make a quick buck.

4:16

Um, and and ultimately, you know, it's like we could let them run the business, you know, pay out investors and pay out shareholders.

4:25

shareholders. um you know everybody makes something right from the exit um but ultimately allow them to run the business as ultimately it could be optimally run right as a again like kind of SAS business now in the AI era and meanwhile you know what we decided to do with hyper agent which I had actually

4:41

incubated within air table is to you know spin it out uh and effectively give a lot or all of our air table shareholders a piece of upside in also hyper agent as kind of a a bet on the purely AI native uh uh you know kind of um uh future opportunity uh that we had conceived with product. So long-winded

5:00

So long-winded way of saying you know I think there was a big uh you know kind of uh fork in the road to decide what do we want to do with the business that we had created of a different era and now you know still a high quality business but a lot of the assumptions and kind of tailwinds that we had built the the product and business originally on were no longer you know there or you know things had changed.

5:21

This episode is brought to you by Jira by Atlassian, where teams get the context, visibility, and control to move work forward. Try it free at jira. com. That's ji. com.

5:34

Thanks also to Granola, the AI notepad for people in backtoback meetings.

5:37

It works everywhere you do and lets you focus on what matters. Try it at granola.

5:42

ai/sources and use the code sources for 3 months off.

5:46

This episode is also brought to you by Mercury, AI native banking that's loved by more than 300,000 entrepreneurs, including me. Visit mercury. com to learn more.

5:54

Mercury is a fintech, not a bank.

5:56

Check the show notes for details.

5:58

notes for details. I think a thing I've heard from investor friends, people who work in tech after that deal got announced was, well, if Air Table still had all that cash and was generating all that revenue, why not just use that to

6:10

to build hyper agent, keep it all together, use the cash flow of the core business and reinvest into the future versus, you know, >> separating the two, selling Air Table, and now you're the CEO and you have a new team and you're you're doing hyper agent. And you mentioned the equity

6:25

And you mentioned the equity piece which is important but but why not do that?

6:28

do that? So two things one is you know as it turns out I don't actually think capital is the primary constraint in building a great zero to one product and finding product market fit right in fact you know rarely has it been the case in you know I think corporate history that the best pocketed either startup like

6:48

you think about all the kind of like disastrous startups that raised too much money up front and then never launched a great product right uh you know or just large incumbents you know the really deep pocketed incumbent who had all the money but couldn't go and innovate on the new thing, right? Like Google with

7:02

Like Google with their plus product when they tried to compete with Facebook.

7:05

Um, you know, even though they had arguably at that time more money, more distribution, they couldn't just go and like brute force their way into it.

7:12

And so I think one lesson I have or one belief I have is that, you know, capital alone doesn't um, you know, doesn't make you win, right?

7:21

And in fact, like it can sometimes be a distraction from the finesse and the obsessive, you know, kind of attention to detail, especially for a, you know, kind of a product that's meant to be uh self-s serve adopted.

7:32

Uh, and in a sense like almost be a crutch to to, you know, kind of not do the right and hard work to get the product right. Um, so that's one.

7:39

And then secondarily, I think, um, you know, what I came to appreciate is, you know, these types of businesses are just so structurally different in almost every way, right?

7:50

way, right? uh again the the kind of pricing model and the long-term you know kind of uh margin and economics of a SAS business is just quite different from AI businesses right like I mean AI has much worse margins by default now you know as we've seen with like the anthropic S1

8:06

like turns out you can actually generate pretty good uh positive contribution margin on on even like training and uh and selling inference on a frontier model but you know I think um in you know in reality like the the economics of a true AI native business typically look very different from SAS. I think

8:21

I think the distribution looks quite different.

8:24

So with hyper agent as an example, unlike Air Table which had scaled really into enterprise and getting into like you know very high quality Fortune 500 type type accounts, hyper agents express focus early on is to actually depprioritize enterprise and go after SMB owner operators, right?

8:38

Um because that will just allow us to have a much faster iteration cycle.

8:43

We can deploy new product changes very quickly, learn quickly from our early customer base, etc.

8:49

And so all of the typical reasons why you would think like, you know, incumbent distribution, incumbent, you know, kind of cash, etc.

8:54

would really help us go and win at this new thing, I just don't believe are actually as as uh synergistic um as you know, uh in practice.

9:06

>> On the bending spoon side, obviously you're not there.

9:07

You've gone on to hyper agent, and I don't want you to speak for them, but when you look at what they've done, it's a very interesting acquirer for a company like Air Table.

9:15

I mean, they've got Evernote, Vimeo, We Transfer, AOL.

9:17

They've collected a lot of these assets that >> great brands.

9:23

>> Well, they're they're amazing brands.

9:24

They're, you could argue that they're from a different era.

9:26

Like you were talking about the era that Air Table came up in, and what was it about them?

9:31

I mean, I'm sure you had options of where you could sell the business, but like what was it?

9:34

You mentioned their integrity and all of that, but I think they're not a company that's super well understood, especially, you know, even in the United States.

9:40

And I' I'd love to kind of know more from you on why that was the right home.

9:46

>> For one, I think they are becoming better understood now, right?

9:48

Obviously, they're public uh you know, publicly traded in the US.

9:51

And, you know, I think as they've kind of gained more scale and started to buy, you know, kind of more and larger and arguably higher quality assets, uh, than when they started out, I think they're they're making a brand for themselves.

10:03

And I think, um, their model, you know, you can read about it in the S1, like I think they're very transparent about how they they run their business.

10:10

And so like you know I think there there is um something very pure about the fact that they know who they are.

10:16

They know what they're good at.

10:17

They have a playbook and I think they're pretty disciplined about adhering to the playbook, right?

10:21

Like they're never going to go and pay a, you know, crazy multiple for some exotic, you know, kind of uh far out bet, right?

10:30

They're going to buy a highquality business that they understand that they can deep dive into the, you know, economics and cohorts of and in their models they can make it work such that they have a fair price they're willing to offer that, you know, um, gets the deal done, right?

10:44

And they bought public companies as you mentioned uh, where, you know, that offer was enough to basically, you know, get the company's board to, you know, agree um, uh, you know, in their fiduciary duty to sell because that was the best outcome for the business.

10:58

And I think they're, you know, they've done and are doing the same in in the private uh markets case, including with Air Table and, you know, obviously you've seen Miro and some of the other companies they bought.

11:05

Look, I think, you know, there's a lot of value I ascribed to a high integrity, transparent actor that, you know, you get exactly um what you expect, right?

11:18

Um you know, because the the truth is like there's also a lot of uh you know, kind of shadier uh you know, types of uh actors in the M&A world.

11:24

actors in the M&A world. like you you have sometimes acquirers who you know may may lead you on may may take you um all the way to a term sheet or even beyond right and the worst case would be like you know then it just uh becomes a diligence or kind of a data mining exercise and doesn't actually go

11:39

anywhere doesn't you know or or yet even worse than that if you have antitrust risk etc which you know thankfully we didn't uh in this case you know that could be even more complicated right we obviously all saw that with Figma and Adobe which you know that that was just a long distraction I think for both companies that ultimately did not culminate in a in a deal. So, you know,

11:57

So, you know, my my opinion is um you know, it's it's kind of like there's a fair price these days for SAS.

12:05

I think you can see the public markets and how they value SAS businesses.

12:09

Um and you know, you have a very high quality, high integrity actor that is exactly what they claim to be and you know, they can transact very efficiently and I think uh for you know, our shareholders, for board, etc.

12:21

uh you know that was you know the the best option for us.

12:27

>> How do you guide the team through a moment like that?

12:29

Because I mean Air Table's over a decade old.

12:30

I'm sure you had people who have been with you maybe since the very beginning still but certainly many many years and people have come over to hyper agent. Some have stayed behind.

12:40

How do you figure all that out?

12:42

That seems like it could be messy.

12:43

messy. It's it is messy and you know I uh I learned a lot uh in my experience of of uh you know building and scaling air table and and I think one of my lessons learned is sometimes leadership comes down to making the right call on

12:58

everyone's behalf even if you can't make every single person 100% happy right and look I'll I'll say it as it is like obviously you know the outcome was not you know maybe at peak uh air table what everyone you know really wanted like you

13:14

know we we sold for ultimately a lower equity value than our peak valuation you know many companies were were valued uh very highly in the ZER era and like you know obviously uh either you know ended up selling for less or are still independent and I don't think will ever

13:29

reach that uh that number again and and yet at the same time you know I think from my standpoint there's a you know there there's a high risk that we won't get a better price even if we stick it out for longer, right, as air table, right? Uh because there's so many

13:45

Uh because there's so many unknowns in terms of where SAS is going, right?

13:49

It could get worse for SAS and SAS multiples, right?

13:51

This onslaught of AI could accelerate.

13:54

In fact, like you know, I'm obviously very AI pilled myself now.

13:58

And like I think AI is coming for everyone and everything, right?

14:01

And you know, it's kind of like do you want to take the overunder bet on on that?

14:05

And ultimately, it almost feels irresponsible to make a gamble on the other side, right?

14:11

I mean, you could say like selling is a is a call you have to make uh on on one hand, but to not sell, even though that's kind of the default from inaction is a different and also possibly riskier bet uh if it doesn't go well, right?

14:26

Um and in our case, the fact that we still did have so much cash on our balance sheet and we'd actually been replenishing the coffers with our cash flow meant that we could uh we were able to clear our prefac right.

14:36

So, we had raised a ton of money uh over 1.

14:38

4 4 billion uh over the multiple funding rounds, but because we had the vast majority of that in the bank account, right?

14:45

Uh we were able to pay back all the investors, clear the prep stack and return dollars to common, right?

14:50

Which, you know, I think a lot of SAS companies may not actually be able to uh even do.

14:57

Again, you know, there's no telling where things are going, right?

14:59

I'm not an oracle of the future, but my bet is, you know, at least that there's a lot of uncertainty ahead in terms of outcome value and certainly common share outcome value for uh even like quality SAS businesses.

15:12

businesses. This is a really important point and it's one that I think people listening to this especially if you work at a company where you have common and you have a lot of value writing on that people even people you think would understand they don't understand how that works what you just described you know the prep stack investors uh you know they're getting you know preferred shares they're cleared first and you're

15:33

an employee and even though you've put your blood sweat and tears into this company ultimately like you may not see a dime because you're you're a common shareholder >> and so It's important when you work at these startups to understand these mechanics and to understand what has the leadership done in terms of capital raised the the dilution the the pref stack like what actually has to happen for your equity worth something. I wish

15:55

I wish that that was something that was I don't know about taught but like at least the the leaders of these companies certainly especially in this era I wish they I wish more of them communicated it more clearly even internally because I think there's a lot of employees who are totally blind to what you just described. >> Yeah.

16:12

Yeah, I mean it's it's complicated, right?

16:12

Like it's not as simple as okay like you know equity value or share price of the company goes up, you make money equity, you know, like to your point like there's these kind of weird divergent states where you know common could get wiped right if if you uh don't clear the prep stack.

16:26

Um these are kind of if you really think about it like from a financial standpoint like these are complicated instruments effectively being used to fund a company right and now like some companies are raising debt to finance uh future growth etc.

16:39

future growth etc. I think of it like you know the simple heristic I use is how do you preserve optionality uh for the company and on behalf of individuals within the company right and you know what I mean by that is it's one thing to go out and raise a mega round and let's say you know it does come with pre um and hopefully it's only 1x pref not 2x pref uh which also happens right you

17:02

know there there are some companies who who take bigger than 1x pref which is you know has its own perils but you know you could go and raise a mega mega round and and yet like there's a big difference between raising that mega round and then spending all that money which means now you have to clear you have to earn back that amount of equity value just to clear the prepack and return even a dollar to common shareholders. um you know versus you

17:23

shareholders. um you know versus you raise the money and you kind of use it as a um like a war chest like a a reserve but you don't actually spend it or you spend it very uh you know diligently right and so maybe you use it for strategic acquisitions where you're confident that the equity value that

17:40

you'll create is much more than the the dollar outlay right of of those acquisitions but you know I think that's one of the things that's sometimes forgotten is like you know people just kind of assume the headline valuation is the valuation right and the reality is like it's not you uh apples to apples. >> Let's talk about hyper agent. So this is

17:56

>> Let's talk about hyper agent.

17:56

So this is your next chapter.

17:57

This is uh >> the company that you're now building.

18:00

Uh you incubated it inside Air Table before you sold the company. Maybe let's start there.

18:06

The decision to go okay this is it. This is the next thing.

18:08

Explain the vision and then how you decided to go okay I'm parting with my teenager Air Table and I'm I'm now betting the next 10 plus years on this.

18:19

You know, the funny thing is the the inception of hyper agent was, you know, somewhat more accidental uh than than uh than grandio Grandios and uh it starts with actually last uh winter holidays.

18:30

um as with many other I think uh you know kind of technical folks or founders like I just kind of fell back in love with building again because I got to spend the whole holiday period uh building with uh in that case like cloud code and uh you know some cursor agents and uh and really kind of using the frontier models at the time like Opus 4.

18:50

5 I think was the the one uh that really felt like a breakthrough to just build and not just vibe code because vibe code always to me implied like it's you know it's a very shallow you know, kind of veneer of like it looks like it works but it doesn't really.

19:04

But, you know, I'm I'm fairly technical.

19:06

Like I thought about a lot of, you know, I thought about and worked on a lot of the architecture of Air Table, including like the back end and, you know, how the real-time collaboration layers, etc. work.

19:14

So, I appreciate not just how things look, but how they work structurally underneath.

19:20

And I think what was really impressive to me in that moment was these agents actually got to the point where they they could they could do more than just make something that looked like it worked.

19:28

They could actually make it work, right?

19:30

And you could actually ship like real product that structurally was pretty similar to what a high quality engineer would go and build themselves, right?

19:39

And obviously much faster, much cheaper than me to go and like hire or have a team of people work on stuff.

19:43

So initially that that holiday period was just me going and like kind of being blown away by what you could do now as you know a solo person, right?

19:50

Like you know for all intents and purposes I could have been a college student again, you know, just uh burning.

19:55

I mean probably I couldn't have afforded the amount of credits I um I used on uh on these models but like you know but but the fact is like you didn't have to have like a big team.

20:03

You didn't need to go and like you know uh formalize a project.

20:07

You could just go and build right and so that was really kind of a rekindling of you know the excitement of early days air table early days like product building for me.

20:15

product building for me. Um and you know going back to the the metaphor I used earlier of like I think we go through these technological cycles and in fact Benchmark um used this uh this term called inside uh inside out versus outside in product thinking right and uh it really stuck with me um because outside in is effectively you know kind

20:34

of what you think of as like classic like MBA driven jobs to be done product um you know kind of building right uh and there's nothing wrong with it like you can build a great you know quality business especially in the SAS era with But you effectively don't care about the technological implementation details until the very end, right? You kind of

20:51

You kind of like say here's the product that I'm trying to build.

20:54

Here's who I'm trying to solve it uh solve for.

20:55

Uh it could be, you know, I'm building software for plumbers, right?

20:59

And you know, we'll figure out the tech later.

21:01

And then, you know, inversely inside out is where you kind of start with a love of the technology and what's, you know, now possible as breakthrough capability because of the tech, right?

21:11

And so I think about like the early days of computing. You have the Apple 1.

21:14

You have uh you know both Bill Gates and Steve Jobs seeing like the 8086 processor as this amazing technological hardware marvel of like for the first time like a CPU that could actually be powerful and cheap enough to power potentially a personal computing experience.

21:29

Right now what does that mean and what does that look like and is there a market for that?

21:31

Those were kind of the unknowns that they have to figure out from the you know starting with the technology first.

21:36

And so to me this this uh this felt like another breakthrough leap.

21:41

Like obviously we got the chatbt moment.

21:43

People were excited about that and it spawned a certain set of applications you could build with that kind of like earlier chatbt like GPD3 era you know kind of model capability and this just felt to me like it was a different era entirely.

21:55

And it just got me thinking more and more about well like what do I personally want to build with this right?

22:02

with this right? And I mean it could have just remained a nights and weekends project like you know just build some hobby projects continue to iterate on it but at some point you know I'm a very meta person like air tableable itself is you know effectively an app platform to

22:15

build more apps right like it's software to build software and you know in a very fitting way I was like well like what if you know I'm enjoying using agents frontier agents to build products what if I made a product that itself is a way to build frontier agents right it's it's agents to build agents, right? So, um

22:32

So, um that's ultimately what Hyper Agent was, you know, uh conceived as.

22:36

OpenClaw was starting to take off.

22:38

I was following that quite quite closely and, you know, it just felt like kind of the right time to go and uh you know, not just make a bet from a resources standpoint, but like personally go and learn and build this.

22:49

And you know in doing so I think I also it it helped crystallize just how much had changed from again kind of the starting assumptions of air table in the SAS era.

23:01

There were so many things that structurally felt completely night and day different in both how you built something like hyper agent the kind of team you needed the resources you needed.

23:12

again, you could do it with way fewer people than you know you could possibly build something of that complexity and and uh significance in the software era.

23:19

And it became increasingly obvious to me that also like the culture and like the business model etc of you know how you would build something like that needed to look quite different than uh you know than than the uh the kind of core business of air table.

23:34

>> And what is the the world you're trying to build with hyper agent?

23:36

I mean it's multiplayer agents that work together.

23:41

It's it's teams of agents and in its essence.

23:43

But >> why does the world need this?

23:44

I mean, there's a lot of agent products.

23:45

I mean, even in the last couple of months, you've seen Grockbot, you know, is growing tremendously.

23:50

Uh there's DOS, there's, you know, obviously Instinct, there's Claude Duet, there's all these agent products.

23:56

How does this fit into all that?

23:59

>> It's very exciting that it's a extremely competitive product landscape, right?

24:03

To me, that actually kind of just adds to the excitement of it.

24:06

like I don't want to work in a lonely bubble where it's like nobody else is working on this kind of product.

24:11

of product. I love the the element of competition and like you know I use all of those products uh products uh quite religiously right like I'm like a super user probably in like the P1 percentile or P99 like of uh of power users on all those products um because I want to stay a breast of like what do great agent

24:28

products look and feel like right and one uh you know kind of really interesting difference between how I'm running hyper agent and how I built air tableable is like with with the original SAS assumptions like you would build software and it was like fairly really hard and slow to replicate that software, right? Theoretically, somebody

24:43

Theoretically, somebody could have replicated Air Table.

24:44

In fact, there are a number of other companies that tried um you know, and ultimately like it's hard to catch up, especially when Air Table itself is continuing to evolve.

24:53

I think with the agent landscape, I mean, things are moving even faster and you have fewer laurels to rest on, right?

24:59

laurels to rest on, right? Even Curser, which obviously was like kind of the the um the rightful darling of the last what I think of uh you know the the last agent era uh you know had to completely reinvent themselves like Grockbot is like a fundamentally different product

25:14

like has almost nothing to do with cursor or even the core ICP of Cursor the developer and you know I think you have to keep reinventing yourself and like you know again like even your distribution as an incumbent like you know is not worth that much and so to me it's like you what we launched with uh for Hyper Agent 1.0 back earlier this

25:32

0 back earlier this year is almost like you know kind of a um like it's going to be a completely different product uh from what we're building now and imminently going to launch as Hyper Agent 2.

25:43

0 right by necessity because you know it's almost like that was just useful as a learning ground for us to learn the mechanics of agents get some some usage and scale and like you know learn from early customers but ultimately led us to a point of view on what's missing from the market right now.

26:00

So effectively, hyper agent 2.

26:00

0 is all about multiple agents and specifically small businesses that want to go and employ a team of agents.

26:07

Um, and if you think about all the great agent products that are out there right now like instinct and cloud duet and chatbt dots, muse etc.

26:13

They have fundamentally a onetoone agent to human model, right?

26:18

And that is a really important form factor.

26:20

Like I think that is a great battleground in and of its own right.

26:23

Like I think I do think like every human, you know, uh even people who aren't in like tech needs an agent, a personal assistant agent that just piggybacks on all their accounts.

26:32

Like I have agents that like, you know, help me like look through all my emails and like figure out what I need to respond to and don't respond to.

26:40

They're helping me with like, you know, kind of tax planning and like medical planning and all that stuff.

26:44

And I think that's a really really important form factor.

26:46

But fundamentally, it's a completely different product with a completely different set of constraints and assumptions than if you want to build a platform that allows a business owner to hire multiple agents, each of which effectively needs its own identity and its own set of provisioned accounts, right?

27:02

So, when you hire somebody, you don't let them just like use the CEO's email, right?

27:06

like you know to send all their emails like you know you actually go and like set each one of them up with their own accounts and each one of them has their own identity and like maybe guard rails and you can audit what the marketer is doing versus the salesperson and also ideally coordinate between them, right? And so 2.

27:20

0 I know hyper agent again it's it's going to feel like almost a completely different product is really about you know redesigning everything about the product around this kind of multi- aent multihuman type of interaction including the fact that you can have like you know group chats with all of your agents at once group calls with all your agents at once.

27:38

Again, you know, kind of each of your agents can have its own identity and um and even like, you know, uh you can apply like selectively policies or guard rails to each of those agents to allow them to have their own kind of tailored rules on what they can and can't do.

27:52

I spend a lot of time context switching between meetings, often with no time to process one before the next starts.

27:58

Thankfully, Granola runs in the background the whole time.

28:02

It's an easytouse AI notepad for meetings that works everywhere, even on phone calls.

28:06

I use Granola to recall what was said in meetings and create helpful summaries.

28:09

I use it every day to stay on top of what I need to get done with my team.

28:13

It connects to my email and suggests follow-ups for me to quickly review and send, saving me valuable time.

28:19

Granola isn't just a core part of my workflow.

28:21

It's basically my second brain. Try Granola at granola.

28:25

ai/sources and use the promo code sources for 3 months off.

28:29

As you move faster and bring in more AI agents, keeping tabs on what's actually happening gets harder. Who's working on that?

28:36

What did that agent just change?

28:37

Are we still on track for our goals?

28:39

That's why teams rely on Jira for visibility across their work, whether it's done by people or agents.

28:47

Jira is powered by Atlassian's teamwork graph, which connects work across Jira, your mostused tools and apps, and your agents.

28:54

That means your agents share the same context your team has like the goal behind the work, the decision made yesterday, and the project plan, so they work faster and more accurately.

29:01

You get oversight, track agent progress right alongside human work, spot blockers sooner, and connect your work to goals.

29:10

You don't have to choose between shipping fast and staying in control.

29:14

One of the biggest pains I've had running sources has been managing my books.

29:17

I'm not an accountant and all of the software out there is unintuitive, clunky, and detached from where my transactions actually live until now.

29:26

Mercury recently launched Mercury Books.

29:28

It's AI powered accounting that works with your Mercury banking and credit card transactions, plus external cards and payroll systems as well.

29:34

You can choose between cash and acrruel accounting and quickly generate reports for your cash flow, P&L, and balance sheet.

29:42

Like everything Mercury does, the design of Mercury Books is super approachable and clean.

29:46

I finally don't feel overwhelmed when I'm trying to understand the big picture of my business.

29:51

I love that AI does most of the heavy lifting, including for things like autocatategorizing transactions, and that Mercury's command AI agent can handle tasks end to end.

29:59

You don't need to be paying for separate accounting software anymore.

30:03

And you can even invite your accountant or CPA to work alongside your Mercury books with you.

30:09

This doesn't cost extra, and Mercury will even help you find a human accountant if you need one. Visit mercury.

30:14

com/books to learn more and apply online in minutes.

30:18

Mercury is a fintech company, not an FDIC insured bank.

30:20

Banking services provided through Choice Financial Group and column NA members FDIC.

30:25

Framer is the AI website builder that powers the sources podcast website at podcast. sources. news.

30:32

Framer brings AI agents into the same canvas where your website is designed, managed, and published so you can move faster without giving up your taste or control.

30:40

I use Framer to make the Sources podcast website be the destination for everywhere you can find the show.

30:46

Plus, you can also see recent issues of my newsletter.

30:48

Start building with agents for free today at framer.

30:51

com/sources for 30% off a Framer Pro annual plan. framer. com/sources.

30:58

Rules and restrictions may apply.

30:58

When I was trying Hyperagent 1.

31:00

0, that was the thing that stood out is just how much I could customize every aspect of how the agent works.

31:08

I got to be honest, almost too much. Like it was a lot.

31:11

There's a lot of like control you gave down to like token spend to the models to the if this person sends it this, it can do this.

31:19

If not this, like >> and it felt like I was building my own like group of Sims or something.

31:23

It was bringing you back to like Sim City or something.

31:27

>> But this world you're talking about multi- aent, you know, a company having agents that they talk to.

31:32

This is a very >> foreign concept to a lot of people.

31:35

>> foreign concept to a lot of people. I mean people are just starting to experience single player agents right dots is only gated it's still gated to the 100month uh chatbt tier instinct is still in private beta duet hasn't even generally launched the cloud version right like we're trying this because

31:51

we're in the bubble but >> like the most of the world has not experienced single player agent you're going on to multiplayer agent why did you skip single like why is why did you see multiplayer as the bet >> yeah so two things one is I think single player is, you know, a knife fight in a phone uh phone booth, right? Like it's,

32:08

Like it's, you know, it's even more competitive than I think what multiplayer will be.

32:15

And again, like it's it's not that multiplayer is a bigger necessarily better market.

32:18

I think they're both going to be massive markets, but we know for a fact that the single player like personal assistant um is a form factor that every large company wants to win, right?

32:30

Like I mean this is literally the culmination or the expression of uh uh Zach's super intellig or personal super intelligence essay right from a few years back. Like this is it.

32:38

This is exactly the product that he said he wants to make an all-in bet on, right?

32:43

And you know personally I don't want to compete against Zuck and uh you know Alex and Nat Freiedman over there.

32:47

Like they're very strong competitors and they have a boatload of money.

32:51

I mean they're actually this is one of the few cases where having like tens of billions of dollars of compute infrastructure.

32:56

They can subsidize the compute.

32:58

they can obviously go all the way down the stack and train their own models.

33:01

Like that's a scary foe, right?

33:03

And for what it's worth, I think Instinct is doing a great job at like kind of being the David to the Goliath there and making it still like an interesting race, but like but you know it's a hyper competitive uh you know kind of arena, right?

33:15

And of course like Google, of course OpenI, of course Anthropic, of course Microsoft all want a piece of that pie as well.

33:20

Um, I think the multiplayer one is interesting to me.

33:25

Uh, also because, you know, it's not as obvious of a product form factor, right?

33:31

Like there's actually a lot of like art more than a science that goes into thinking about like what is the right way to coordinate work between multiple agents, right?

33:39

What's the right metaphor to use when you break down work across these agents, right?

33:43

Um, and there's not a one-sizefits-all solution, right?

33:47

Like some people uh in the development world right have developed their own idioms with like orchestrating multiple agents with one agent what's the right way to share state between those you know uh what what's the level of parallelism you use when you spin up lots of sub agents etc and you know to your point earlier about hyper agent 1.

34:02

0 know being a little intimidating and having a lot of different options like almost too much configurability.

34:08

almost too much configurability. I think that's absolutely a fair critique and just as the consumer version of agents has distilled down the choice set right like you come into instinct there is no config you text the thing over iMessage and you don't like there's literally nothing to configure right and even muse like the configuration is like quite

34:26

limited um by intention right versus you know kind of open claw from back in the day I mean you had like infinite configurations right and it would always break and you had to like choose everything and I think similarly we have to distill a set of like um smart default choices or like just make it work out of the box but in this multiplayer uh mode, right? And so

34:44

And so deciding like you know how should these agents coordinate with each other?

34:48

agents coordinate with each other? How autonomous do we want that to allow them to be at the risk of you know if you get it wrong they could just spin a lot of idle cycle or spin a lot of wasted cycles burn tokens and not accomplish like the business outcomes that the the uh company wants right or you know

35:03

conversely like they could go and do things that have negative impact right like accidentally leak data or do the wrong thing or you know kind of send awkward messages and so you know I think the multi- aent uh product is to me uh really interesting because it requires a lot of these you know, kind of non-obvious creative product choices to make it actually work well, right? And

35:22

And uh and that's a lot of what we're kind of obsessively focused on.

35:26

of obsessively focused on. Well, at the same time like I think any agent product that wants to compete single or multi- aent needs to have like a baseline of just you know kind of frontier capabilities right it has to be fast performant um you know cheap enough

35:40

right uh especially as you know kind of the cost is coming down and like the aggressiveness of you know kind of a competitor subscal baseline level uh while also you add your own special sauce on top of it to make it you know kind of differentiated product for uh for our use cases. So,

35:58

So, you know, nothing's guaranteed, but uh but I think that the multi- aent um category is going to be a lot more winnable for us and and you know, something that I personally am excited about building in >> for people who have not experienced what you're talking about multiplayer agents, maybe haven't even experienced single player agents, but certainly not multi.

36:19

Um, can you point to a thing you've seen with like either internally or an early hyper agent user where you're like, "This is crazy.

36:26

I can't believe that this company is using agents this way."

36:30

Just something for people to to latch on to. >> Yeah.

36:33

I mean, what I love about our focus on small business owners is I mean, they often have very tangible businesses, right?

36:39

I mean, we're talking like we literally have like small insurance brokers, right?

36:43

Like they sell you insurance.

36:45

you insurance. like it's not like you know super abstract like we're building software to make more software right like like air table itself um you know or like landscapers we have like designers we have like creative agencies like these are all very I think groable businesses and you know ultimately these

37:02

business owners I think if you talk to any small business owner like 99% of the time if you ask them like what do you want um like what what if you could wave a magic wand and get anything to help your small business what would it be right I think most often you would hear them say like I wish I could just have another hire, right? It's not, you know,

37:16

It's not, you know, not usually like, oh, I wish I had like a CRM that had slightly smarter automation or, oh man, I wish my email marketing software like had more accurate pixel tracking of my customers or whatever.

37:29

Um, not to say those things are important, but like, you know, ultimately all of these tools and, you know, kind of like the SMB, you know, kind of um, uh, productivity suite is very secondary, especially for these like less technical small business owners or less tech oriented small business owners than I just want like another employee, right? I need another marketer.

37:48

I wish I had a hire that could just deal with client intake, etc.

37:49

And so, and that's exactly what they're doing with Hyper Agent, right?

37:53

Um, interestingly, they're not replacing employees.

37:56

It's more like they're augmenting their uh their current staff and saying, you know, we we have um you know, these owners who will say, I was able to get the work done of 10 incremental hires that I couldn't afford and that itself allowed me to generate just more revenue for my business.

38:12

Like I'm actually able to turn a profit, right, and generate more revenue for my business.

38:18

Um, you know, to make it very very uh crisp, it could be, you know, uh, a very real example, like if you're a landscaper, like you can literally have an agent uh, taking every client inquiry and doing research on the client, right?

38:30

So, this would be a full-time job for someone, right?

38:31

Very important because that's how you win more business.

38:33

But instead of just doing kind of a a you know, kind of a basic like surface level, you know, kind of Google of that person, these agents can go deep, right?

38:41

They can go and look up everything about this, you know, client inquiry, like what's the address?

38:44

Okay, let me look at satellite imagery of that location.

38:47

Let me see any real estate photos from past listings on Zillow, etc. about this property.

38:52

Let me combine all of that and actually use, you know, kind of um uh image models to mock up what this property could look like.

38:58

Let me use some of my skills to actually figure out like how much do I think this realistic quote for this landscaping job would be, right?

39:06

And ultimately, like let me produce a really beautiful and accurate and faster client proposal, right?

39:12

right? when my competitors are maybe taking five weeks to even give a call back and then asking like 50 more first day questions like let alone like coming to do an on-site visit to even like get to that point like I'm going to undercut them and show them show the client like a really beautiful and realistic uh depiction of what I can do for them and

39:30

then by the way if I close them then I can use agents to help project manage the whole thing right like every week every month like okay here's the scheduled jobs here's the the the notes left from the field team okay like we got to remember to actually come back the next week and like bring you know kind of these plants or these uh you know these tools etc. Like all of that

39:46

Like all of that is typically like the operational detail that makes or breaks these businesses.

39:53

Like I mean I'm experiencing this ourselves like we've gone through a couple landscapers in our uh you know in our home and you know it's like the the difference between the one that we keep and we pay more to and the one that we don't is if they get all the details right. Right.

40:05

And I think um what we're seeing with agents even in this hyper agent 1.

40:09

agent 1.0 know product is there are these small business owners they're not you know just tech founders uh they really are kind of entrepreneurs in every uh category and you know they're not like exactly what you would expect right it's not always like somebody who is a CS major you know or whatever they just are very enterprising and they're willing to go and like roll up their

40:29

sleeves and tinker and they in fact get an excitement and delight out of like tinkering with uh with agents now they can't go like super deep into the technical configuration which is why like making the product more self-serviceable is uh is a key, but they're able to come in and like build these agents that ultimately immediately create real business value for them, right? And for them that's everything,

40:50

And for them that's everything, right?

40:51

Like if they if they can make another few or you know a few dozen hires um that actually generate new revenue for their business at low enough cost that they actually produce more profit.

41:02

I mean that's everything.

41:02

profit. I mean that's everything. So, um I think it's a huge huge market and you know what's really exciting is that like the economics really are more oriented around like what what would the world look like if you know and again I don't think about this as like replacing

41:16

humans so much as if every small business could augment their current staff with 10 of their handpicked best possible genius level you know um uh white collar you know kind of uh hires for you know real money but still much less than what it would cost them to go and hire those incremental people, right, which they can't afford. And how

41:34

And how much more GDP would be created?

41:37

And therefore, how much, you know, are they willing to pay up to?

41:41

And we're talking already like a massive like multi-trillion dollar opportunity of which even if Hyper Agent as a player here captures a small percentage of uh and you know kind of also gives a lot of surplus back to the uh the customers, you know, it's a huge huge win for everyone involved.

41:58

>> You raised over a billion for Air Table. What about Hyper Agent?

42:00

uh do you need to raise capital?

42:02

How are you thinking about that?

42:05

>> We we have raised a small round.

42:05

So we we raised a small round uh pretty much immediately after spinout from uh mostly insiders.

42:11

So we just want to be efficient uh and you know work with people that I've personally already had you know now the benefit of almost a decade of experience with um and in the like many tens of millions of dollars uh in funding.

42:23

We haven't yet formally announced it.

42:25

Um, but really I mean it's like that's not what um, you know, we're we're tooting our horn about, right?

42:32

Like we know that uh, and I personally know that, you know, nothing will matter with this business unless we build a breakthrough product that, you know, especially with the 2.

42:41

0 launch that really just uh, differentiates and, you know, creates a magical new level of delightful kind of value for these SMB business owners that we're targeting, right? Like Hyper Agent 1.

42:53

0 is getting some traction. It's scaling, etc.

42:55

And like, you know, arguably you could build a business with just that 1. 0 product.

43:01

But for me, it's all about, you know, like redefining and basically placing an all-in bet again on, you know, building a almost new product from scratch now that we've spun out and we're we're uh aiming for this 2. 0 launch.

43:13

And, you know, we have to operate like we are truly a seed phase company, right?

43:16

like we're not, you know, resting on any laurels of having been part of Air Table or myself, you know, having founded, you know, a much larger scaled business, you know, we have to go back to the trenches and that's exactly uh what we and what I am doing uh pretty much every day.

43:32

>> I'm curious about reflecting, you know, on your journey with Air Table and what you were saying at the start of the conversation about it feels zerpy right now.

43:39

I gotta be honest, like as someone who was there in the trenches in Zerp when companies were raising aggressive valuations like Air Table did, like many did.

43:49

It was the twilight of SAS.

43:49

It was pre-AI and like everyone was doing super well.

43:54

Feels like that kind of right now with a bunch of the AI companies I meet with and my other hat as an investor.

44:00

They're all they're all going from, you know, a hundred to a billion next year.

44:04

They're all raising at great multiples. Everyone's winning.

44:07

It feels too good to be true.

44:09

Uh, and I I be I mean you have such this unique journey you've been on.

44:14

Like I'm sure you talk to some of these founders too, but when they ask you for advice or you just give it like what what analogies can you draw to these previous eras that you've lived through as a founder and what we're in right now?

44:25

>> I lean less on, you know, kind of a literal analogy to the prior eras and more and more think about it as a first principles problem, right?

44:33

You know, I do think there's a lot of capital flowing into AI, right?

44:38

Both the capex buildout, but then also at every layer above it, right?

44:42

I mean, at the inference layer, right?

44:44

Like at the like applications layer or the the kind of agents products layer, you know, I think a lot of that to me is that the rational, you know, kind of basis for that is this is an unprecedented technology.

44:56

I think it's, you know, far more potent and it's certainly growing uh and improving faster than any prior tech cycle, right?

45:04

Like if you think about mobile or internet or like CPUs, personal computers like I mean AI both uh improvement is making Moors law look flat by comparison and the rate of proliferation into the world is just so much faster.

45:19

much faster. I mean like the the the um rapidity with which chatbt reached a billion users makes even Facebook and before that Google and before that Yahoo and before that like you know internet and computing like look like again like very slow and flat and so um I think we are dealing with like you know kind of

45:34

an unprecedented set of fundamentals to AI and I mean you know like I'm obviously I appealed but like my belief is that you know this is going to effectively increase GDP like if you think about like all of the contribution of intelligence from humans in the world and that's like tens of trillions of dollars, right? Um easily per year that

45:52

Um easily per year that is going to be possibly doubled, possibly tripled, possibly 10xed, right, by AI.

46:00

And so you're just dealing with like such crazy circumstances, it's really hard to like, you know, kind of uh reason in a very precise way about valuations.

46:09

And to your point, that is going to result in some sloppiness in one direction, right?

46:13

like you're going to have an overinvestment into companies that don't actually have, you know, kind of the mechanics or what it takes to be one of these like massive winners, right?

46:21

Um, and there is still a jury out on like how much of the massive pie gets taken by the big guys and the frontier labs, you know, versus actually is left for seed round companies.

46:31

Um, I really appreciated a post by Sarah Guo.

46:33

I think it was a few months ago talking about like her point of view on like, man, a lot of people, you know, are now like just kind of decrying the fact that like there is no real early stage ecosystem.

46:45

Like, why would you bet on anything except the big guys, right, the the Frontier Labs?

46:48

And she had a very good defense of like the the ecosystem and like startups, etc.

46:52

, and like why they will create value.

46:54

Um, and you know, I largely agree.

46:56

Uh but like there is you know kind of real questions out there around like who's going to take that the pie is clearly going to become very large.

47:04

It's already large and growing.

47:06

Um you know how big does it grow? That's an open question.

47:09

Who wins what slice of the pie? That's another question.

47:11

the pie? That's another question. Um but ultimately I think you know it's like if the SAS era's you know kind of uh Zerpy valuation peaks were predicated not only by the fact that Zer literally encouraged an influx of capital into the

47:24

equity markets right um you know but also because there was some rationale that well software just has these beautiful structural properties which are it's very high margin um you know you can scale distribution eventually up with zero almost marginal cost, right? And so effectively like you have these

47:42

And so effectively like you have these fixed costs of like building the product, maintaining the product and eventually your, you know, your margin profile will look better and better and better over time.

47:50

And these are products that can scale to a very large customer base over time.

47:53

So, you know, there are all these assumptions baked behind, you know, high SAS valuations, right?

47:58

Um, and I think, you know, like obviously some of those are true, some of those are, you know, less true now, especially in the AI era.

48:06

in the AI era. Uh but I think like you know similarly with the AI era you kind of have to just break it down on like first principles because I don't think we've ever had an exact situation that's like this right I mean both in terms of the sheer scale and pace of it but also I mean AI itself is like so fundamentally different like I don't think you know it it doesn't um uh I

48:27

don't think it invites great comparisons to any other you know kind of uh fundamental technology right even like electricity right uh or something like that it's just so But at the same time, there has to be, you know, if you were still running Air Table or when you talk to other SAS founders, I'm sure you have a perspective on at least this moment for SAS, right? Like you were alluding to

48:47

Like you were alluding to this earlier, but yeah, what would you tell companies that are still stuck in that >> seat based or just kind of PAI and that margin structure and all that stuff?

48:58

Like they all can't do what you did.

48:59

They all can't, you know, sell to Bending Spoons and have a new startup.

49:02

And, you know, maybe they don't have that in them.

49:04

maybe they just want to see these these companies through.

49:05

How do you what would you make of people in that position? Like what do they do?

49:11

>> I mean, I guess I would beg the question of, you know, again, like I think um not every company obviously is guaranteed success or even an existence, right? Uh in every era, right?

49:23

I mean, like this is creative destruction, you know, capitalism at work, right?

49:27

And you know, I think when the uh you know, when the play the the playing field and the rules thereof have changed and are changing very rapidly, I think you have to revisit whether you want to be playing the same game as you were before or if you want to, you know, figure out a way to retire that game in a, you know, in a graceful way and then play a go out and play a new game, right?

49:50

And I think it it just comes down to like having the kind of humility uh to you know sort of fully appreciate that the game has changed.

50:00

that the game has changed. Maybe like you know you were very good at playing the old game right maybe your company and like you know this is like your your team was very good at playing that old game but uh you know that famous like business book uh who moved the cheese right like the cheese has moved uh I mean I think that part is like kind of a very obvious statement right like

50:18

clearly we live in a very fast changing world certainly like in technology we live in a very fast changing world where like all of the it it would be more ludicrous to say that all of the assumptions that we had from 15 years about how to build a great tech business, you know, obviously in that era like, you know, around SAS are going to be exactly the same in this new era, right? So, you know, I kind of flip it

50:39

So, you know, I kind of flip it and say, well, like, you know, if you reverse the question and said, why do you believe, you know, like what what kind of like level of either arrogance or, you know, kind of blissful uh, you know, kind of um oblivion to like how much the world has changed is, you know, is needed to, you know, kind of place a bet back on, you know, the assumption that, uh, things are not going to change, right?

51:06

And you can keep doing the same old same old, right?

51:07

and that that will even maintain equity value, right?

51:11

Because I think even the maintenance of equity value, which itself is like by definition underperforming the S&P, right?

51:16

Like I mean you you like you know to be a company that makes sense to exist especially for to attract capital and like you know kind of earn its worth in terms of like every form of equity, employee equity, etc.

51:28

Like you have to have a rational belief that you can outperform the alternative investments, right?

51:34

And you know, if if you can barely make a case for staying flat in equity value, uh, let alone preventing it from like, you know, kind of go going down over time, then I think that's a, you know, again, that's a tough proposition to stand by.

51:50

>> How are you using Hyper Agent?

51:50

I'm curious like in your personal setup, we can maybe end it here.

51:55

What is what is something that's unlocked for you?

51:57

So, I mean, honestly, I pretty much do everything with hyper agent.

52:01

Like, meaning um this is going to make me sound uh you know, misanthropic, but like you know, I so often now turn to agents to ask every question like, you know, strategic questions, product questions, like you know, riff on like new marketing ideas, right?

52:18

And like just as one tangible example, like you know, never before was it possible to have a marketing agent that actually can go and not just like give you like verbal feedback on a, you know, kind of like a campaign idea or positioning idea, but like literally can go and generate like a Super Bowl quality video ad, right?

52:38

And say like, "Wow, like here's that idea brought to life in like a form that like you could actually put out in the Super Bowl and it wouldn't be embarrassing, right?"

52:46

embarrassing, right?" um you know or likewise for like you know product stuff it's going and having an agent that you know like studies all of our competitors and like is tracking all of the competitive releases from these other products like even the solo player ones and then saying like oh maybe like we should apply that to hyper agent right

53:03

like the product and in fact like because it has access to the code base and to our developer you know kind of discussions around how to build uh hyper agent that agent can go and actually not just you know kind of tell you what the news is, uh, competitive news is, but actually spin up like some product features and not just a prototype. Like

53:21

Like it could actually spin up like a, uh, a pull request of like a real merge ready, you know, kind of a chunk of code that implements it in the real product and you can like, you know, play around with the development build of, right?

53:32

And so I just think like the it's almost unfathomable to me how many different uh degrees of abstraction or indirection you know the old world required to get things done.

53:45

Like you would go and see something and you would think about it.

53:49

You would write up like a document or a memo or a product requirements doc and then you would give it to some people and then they would huddle and have a meeting and then you would have a follow-up meeting and you might do a design prototype and then you know like more people would talk about it.

54:00

more people would talk about it. you'd have like a pricing and strate like you do all these things and then maybe like 6 months later and like many many meetings later like you would ship the thing right and it might work and it might not work and I think the fact that

54:12

now part of the benefit is that like the agents a are never sleeping right so like at any hour I actually uh know some people who uh work at you know various companies with very intense much more scaled founders than myself like the Elons etc the world and it's like you know you work for somebody like that you're getting a call at 2 a.m. and you m.

54:29

and you better respond, right?

54:30

Like if if they have something important.

54:31

Well, you know, it turns out like you could do that with your agents and the agents don't, you know, kind of stress about it, like they're not going to quit on you.

54:38

You know, just if if at 2 a. m. 3 a. m.

54:40

you wake up, you're standing, you know, you're lying there awake in bed and you're like yelling at your agents to come up with like these new uh product concepts, like they do it.

54:46

They do it willingly and quickly.

54:48

do it willingly and quickly. And uh and then also on top of that they're just so efficient and you know can be if you kind of set them set them up correctly comprehensive in their work right like they I mean they obviously make mistakes

55:03

but they don't make the same types of mistakes that like we humans and like myself like make right like if I tell this agent like hey go and like thoroughly review the current implementation of this sharing feature and then cross reference like these five competitive products look at how they do it. Do a tear down like literally do a

55:20

it. Do a tear down like literally do a screen grab and screen cap like you know kind of screenshot like walking through their products compare it and then like you know kind of come up with like your version of it here like it's not going to make like a clumsy mistake where it's

55:33

like oh like I forgot to include this step in that uh in that flow right and so they're just so like comprehensive and and also so patient like you you know agents obviously like if you have them on a recurring schedule or a heartbeat mechanism them to keep trying. They don't give up. And so, like, it's They don't give up.

55:50

And so, like, it's kind of like the super employee, right?

55:54

Like, um, you know, if if you set them up correctly, you get these agents that do work around the clock.

55:58

You can tell them to do stuff at any hour.

56:00

They work around the, you know, around the clock, right, overnight if you want them to, and they can kind of do work in a much more methodical and comprehensive, but also still creative and kind of smart way.

56:15

way. um you know and so it's just it's truly to me gamechanging how you can run a business when you are fully built around the the idea of running them with agents and you know again I think like we're both learning this ourselves by dog fooding our own product and kind of dog fooding like these pioneering

56:33

practices but I also think like every business that figures figures this out and they could be in any industry right any category is going to have like not just a 2x but like quite literally an increasing up to like 10 to 100x advantage against their their competition in terms of the efficiency, the speed, the growth that they can achieve, right? Um, and sometimes in

56:54

Um, and sometimes in like what would have been ordinarily like a lower growth, lower margin type of business, right?

57:00

You think about like the landscaping business or insurance brokerages, etc.

57:04

Like, you know, these are not like super high leverage, you know, kind of pure software businesses.

57:10

And yet, I think if you do it right, if they if they figure out how to get agents uh in a very leveraged way deployed across the company, they actually might grow at these like kind of crazy uh and very profitable rates. >> You're here, Howie. Thank you. Appreciate it. >> Of course.

57:27

Granola is the best AI notepad I've tried.

57:29

It works everywhere on a video or phone call, in person, or an Apple Watch. Try it now at granola.

57:34

ai/sources AI/sources and use the promo code sources at checkout for 3 months off.

57:40

Jira by Atlassian gives teams the context, visibility, and control to build with clarity and ship with confidence. Try it free at jira. com. That's ji. com.

57:52

Banking should feel like modern software.

57:53

Get everything you need in one place. Visit mercury.

57:55

com to learn more and apply online in minutes.

57:57

Mercury is a fintech, not a bank.

58:00

Check the show notes for details.