0:00

You're listening to start up for

0:02

the rest of us. I'm Rob

0:04

Walling. I'm your host this week

0:06

and every week. And in this

0:08

week's show, I talk with Jordan

0:10

Gahl, the founder of Rosie at

0:12

heyrosie.com. Jordan and his team have

0:14

been building this AI answering service

0:16

for your business calls and he

0:18

has pivoted the company from what

0:20

was an e-commerce headless checkout. And

0:22

this is all after raising quite

0:24

a bit of money. in venture

0:26

capital. And so you will hear

0:28

Jordan in this interview mention that

0:30

they had a few million dollars

0:32

left in the bank. And he

0:34

kind of flippantly says it. And

0:36

I think that's funny because when

0:38

you're in that game, you know,

0:40

he meant I only have a

0:42

few million left in the bank.

0:44

But obviously having that much money

0:46

in your bank account gives you

0:48

some optionality. And that's what Jordan

0:50

takes advantage of as you'll hear

0:52

in this interview. As you listen

0:54

to it, what I don't want

0:56

you to think is making in

0:58

the bank. Because money doesn't solve

1:00

your problems. What solves your problems

1:02

is trial and error, gut feel,

1:04

execution, finding that gap in the

1:06

market, being committed, doing a lot

1:08

of things quickly, and having most

1:10

of them work out. And that's

1:12

what Jordan Gall has done, both

1:14

with this company and his prior

1:16

effort, Cart Hook. You'll hear that

1:18

in the first three or four

1:20

minutes, the story of growing that

1:22

to millions in ARR, and then

1:24

getting squashed by Shopify. It's a

1:26

great conversation. We cover all manner

1:28

of topics ranging from getting started

1:30

to, you know, what it feels

1:32

like to build, and I'll put

1:34

in quotes, an AI wrapper moving

1:36

very quickly and what it feels

1:39

like to have product market fit.

1:41

Jordan has signed up to be

1:43

one of our founding coaches of

1:45

the SAS Institute. That's at SAS

1:47

institute.com, and it's for B2B SAS

1:49

founders doing a million or more

1:51

in ARR. He's a phenomenal founder

1:53

and we are lucky to have

1:55

him as one of our coaches.

1:57

If you're interested in potentially being

1:59

part of a mastermind, getting one

2:01

on one coaching having an amazing

2:03

community head to sass institute.com to

2:05

check it out it is a

2:07

premium paid coaching program that we

2:09

are rolling out through tiny seed

2:11

and lastly before we get going

2:13

applications for microcomf

2:16

mastermind matching are open right

2:18

now and those are open for

2:20

all levels of revenue not just

2:22

a million and up applications close

2:25

on March 31st head to microcom

2:27

master. And with that, let's

2:29

dive into my conversation with

2:31

Jordan. Jordan, Gal, welcome back to

2:34

the show. Rob, it's good to

2:36

be back with you. It's been

2:38

a few years? Not the first

2:40

time? Not the second time? Maybe

2:42

the third, fourth, I don't know.

2:45

But it's been a while, and

2:47

it's great for you back. Yeah,

2:49

you've been on at least three,

2:51

maybe four episodes, so really appreciate

2:54

it. Today we're here to tell the story

2:56

of Rosie. It has been a journey, sir.

2:58

You are telling me offline. You are four

3:01

years between product market fits with two different

3:03

apps. So you want to give like summarized

3:05

cart hook for folks and then kind of

3:07

catch them up where there's a rally and

3:10

then there is Rosie. And then we'll talk

3:12

about, I want to find out where Rosie

3:14

is today, where it stands, team size, you

3:16

know, if you don't have to share revenue,

3:19

but just anything that gives people an idea

3:21

of what's going on. And then we'll dive

3:23

into other stuff. All right, cool. So

3:25

so the reason you and I like

3:28

almost like giggle a little as we

3:30

talk about the dramas because you followed

3:32

along You invested in card hook, and

3:35

so you've been getting investor updates for

3:37

years for me Not every month as

3:39

any of my investors know, but but

3:42

regular Consistent anyway, so yeah, card hook

3:44

had product market fit card was a

3:46

checkout for Shopify stores before that was

3:49

the thing we were one of the

3:51

first and it flew So there was

3:53

a point in time where it was

3:55

adding 30, 40, 50K and MRR every

3:58

month. No advertising, word of mouth. just

4:00

flying. It went from a million they

4:02

are to two-ish and then six. So,

4:04

you know, very quick growth, bootstrap-ish. We

4:06

took a few bucks from friends and

4:08

family, voters effectively, bootstrapped in the way

4:10

it conducted itself, and then it ran

4:13

into a wall named Chopify. So they

4:15

did not want checkouts to take payment

4:17

processing off of their platform, and they

4:19

ended car hooks run. So I had

4:21

the taste of product market fit. It

4:23

felt really good. And then I started

4:25

a company called Rally, which was an

4:27

e-commerce checkup for everyone outside of Shopify.

4:29

And that was intended for what was

4:31

called the headless ecosystem. That didn't quite

4:34

work. We went toward Enterprise. That kind

4:36

of worked. We got to a few

4:38

hundred K&A R. It kind of wasn't

4:40

going fast enough. And then we pivoted

4:42

to Rosie an AI voice. agent that

4:44

answers the phone for small businesses and

4:46

It might be a bit premature to

4:48

call it product market fit But it

4:50

feels familiar Got it. So it's growing

4:53

fast. And Rosie, when you say small

4:55

businesses, do you mean like lawn care,

4:57

maybe electricians, cleaners, where they're like out

4:59

doing stuff and their cell phones ringing

5:01

all day, right? While they're trying to

5:03

do stuff. And so this Rosie acts

5:05

as an AI agent. I ask this

5:07

knowing the answer because I've called your

5:09

Rosie number and asked the questions, but

5:12

I want the listeners to understand you

5:14

know what's going on. Yes. There are

5:16

a lot of businesses in the economy

5:18

that rely on the phone. I'm a

5:20

painter and I'm on the job and

5:22

I can't answer the phone. Two, I'm

5:24

that same painter and it's 10 a.m.

5:26

on Saturday and the phone rings and

5:28

I'm with my family having breakfast and

5:31

I don't know if I should answer

5:33

the phone, but it might be a

5:35

$10,000 job. All the way too, hey,

5:37

it's after hours and would are a

5:39

water remediation company, we got to pick

5:41

up for emergencies and it's 11 o'clock

5:43

at night and the only option I've

5:45

had as an answering service. But any

5:47

businesses, it tends to be local businesses,

5:49

the type that you search for online,

5:52

and then you call on the phone

5:54

to interact with. So that's who we're

5:56

helping. It's like a very, very gnarly

5:58

problem, which I can talk about how

6:00

painful the problem is, has led to

6:02

challenging my assumptions around churn. And can

6:04

you give us an idea of where

6:06

the business stands today, team size, growth,

6:08

whatever, whatever you want to disclose? Somewhere

6:11

last year, I came to terms with,

6:13

I guess it was toward the beginning

6:15

of the year, 24, I came to

6:17

terms with, hey, this rally thing, it

6:19

might just not work. And in that

6:21

case, what should we do? Should we

6:23

just, we had a few million bucks

6:25

in the bank, should we continue to

6:27

push on, should we close the company,

6:30

should we pivot and keep the cap

6:32

table the same? So I had one

6:34

of those like corporate situations. By coincidence,

6:36

I was in London for a conference,

6:38

and I ran into a friend of

6:40

ours, Justin McGill. Justin McGill. Justin McGill.

6:42

Justin McGill. And I'm sitting around the

6:44

table with my co-founder and few colleagues

6:46

and we're talking to Justin about his

6:49

experience in AI. And I have to

6:51

give some credit to that conversation. It

6:53

just planted a seed in the back

6:55

of our minds because what we were

6:57

doing is we were talking to someone

6:59

who is experiencing product market fit in

7:01

all the glory and challenges of it.

7:03

And it reminded us of what it's

7:05

supposed to feel like. And I guess

7:07

I went home thinking, it's not supposed

7:10

to be this hard to grow. And

7:12

I think that was the moment where

7:14

I looked around and I really thought,

7:16

okay, I think it's time to make

7:18

some changes. So we took the team

7:20

from about 20 people down to six,

7:22

we all held hands virtually, and we

7:24

said, do you want to keep doing

7:26

this? If you want to keep doing

7:29

this, I'm going to come up with

7:31

another product idea, and we're going to

7:33

keep the team small, and we're going

7:35

to build something in AI. And if

7:37

you're in, cool, if you're not totally

7:39

understandable understandable, if you're not totally understandable.

7:41

did my search around for ideas and

7:43

landed on this voice agent thing and

7:45

we started. to build maybe May or

7:48

June and then we had something functional

7:50

to look at around July and then

7:52

we started to bring on the first

7:54

customers in August and then the first

7:56

paying customers in September and now here

7:58

we are in January I think it's

8:00

really worth talking about that journey and

8:02

the feature set and the decisions we

8:04

made around that because in hindsight it

8:07

seems to have gone really well. And

8:09

now we have been doubling every month

8:11

for several months in a row and

8:13

it's starting to get a little serious

8:15

in terms of the revenue growth. Yeah,

8:17

so, you know, we'll blow past a

8:19

million ARR very soon. So that's, you

8:21

know, that's a few months worth of

8:23

growth. And yeah, I think, I think,

8:25

I think it can fly. It's a

8:28

funny thing. I think this thing has

8:30

wings is something I repeated several times

8:32

somewhere around like November, December, because it

8:34

just felt like, like, oh, This is

8:36

how it's supposed to feel. It's not

8:38

hard at all. Everyone just wants it

8:40

and they're pulling at you and demanding

8:42

and asking and begging and want it

8:44

to work. So like that feeling just

8:47

gave us all this confidence and energy.

8:49

I remember at a certain point in

8:51

the life. time of drip, I think

8:53

it was 2014, where all of our

8:55

graphs shifted, right? It's like churn went

8:57

down, trial to paid went up, like

8:59

all in the right direction. And I

9:01

remember pointing and telling, I think it

9:03

was Derek, I said, that's what product

9:06

market fit looks like. Like, that's it.

9:08

And you just said a sentence that

9:10

I think was similar. You said, you

9:12

kind of said, this is what product

9:14

market fit feels like. Right? I mean,

9:16

that's, that's, it sounds like this poll,

9:18

this market poll. I mean, it does

9:20

make the question, I asked this periodically

9:22

on this podcast, because question I'm going

9:25

to ask in a second, because it

9:27

seems like everyone has a different definition

9:29

of product market failure. We kind of

9:31

know, we'll know when we see it,

9:33

is what we'll know when we see

9:35

it, is what we hear often, but

9:37

like, when we see it, is what

9:39

we're seeing a one or a one

9:41

or a zero, one or a zero,

9:43

we're a one or a zero, one

9:46

or a zero, one or a one

9:48

or a zero, one or a zero,

9:50

one or a zero, one or a

9:52

zero, one or a zero, one or

9:54

a zero, one, one or a zero,

9:56

one or a zero, one, one or

9:58

a zero, one, one, one, one, one,

10:00

one or a one, one or a

10:02

one or a one or a one,

10:05

one, one, one, one, one, one or

10:07

a one It was qualitative at first

10:09

and then quantitative after. So the qualitative

10:11

was just the responses from people, the

10:13

support chat, literally the vibe in the

10:15

chat. Like you just hear people, oh

10:17

my God, I can't believe this is

10:19

as good as it is. How do

10:21

I get this here? Like can you

10:24

jump on the phone right now to

10:26

help me? Like this, you know, just

10:28

very, very strong desire for it to

10:30

work and to be part of their

10:32

business. And then later it turned more

10:34

quantitative where I was taking screenshots of

10:36

like profitable graphs. and posting it to

10:38

like Rock and Jazz being like, this

10:40

is a joke. What is this? And

10:43

then you start to, you almost start

10:45

to learn more about your own product

10:47

as you go. So at first, when

10:49

I did my projections, I put churn,

10:51

I think I put it, turn it

10:53

15 or 20% monthly. Because I'm thinking,

10:55

this is an AI product for non-technical

10:57

people. And it's not actually that good,

10:59

which is one of the things that

11:01

attracted me to it, because I think

11:04

everything in the AI will be very

11:06

good. And so the things that aren't

11:08

that good right now are actually the

11:10

best things to build in, because everyone

11:12

just assumes, oh, is it a voice?

11:14

Is it really going to replace a

11:16

person? Yes, actually it is. But if

11:18

you start now, people laugh at you,

11:20

and then you kind of get in

11:23

the right spot for a year from

11:25

now. And the churn is like a

11:27

fractional. So it just has to beat

11:29

voicemail and you get ROI. And so

11:31

all these things start to reveal themselves,

11:33

like a few weeks into having paying

11:35

customers, a few months later you start

11:37

to basically understand your own product. Yeah,

11:39

this is where you had you had

11:42

a thesis that this was a need

11:44

and it would replace a human and

11:46

you get like you said you get

11:48

three weeks in you're like Oh, it

11:50

doesn't even need to do that. This

11:52

is the learning. So this is what

11:54

I see separating like really successful founders

11:56

from those that struggle is they can

11:58

both have a thesis or hypothesis and

12:01

they both build something and put it

12:03

into the market and the best founders

12:05

like bunch of noise comes in There's

12:07

a lot of noise. You're presenting it

12:09

like it was super clear. I know

12:11

that it wasn't, but you figured out,

12:13

the good founders figure out, they sift

12:15

the noise away and there's a bit

12:17

of gut feel. there, there's a bit

12:19

of conversation with team, you ask people

12:22

you trust, and the founders that I

12:24

see, I'm not trying to make a

12:26

blanket statement, but like founders who struggle,

12:28

they launch that and then they like

12:30

either don't listen at all to the

12:32

feedback and they're like no my initial

12:34

thesis holds, or they don't, my initial

12:36

thesis holds, or they don't, they can't,

12:38

they can't, they can't, they can't, they're,

12:41

you know what I mean? Sure, I

12:43

think what it helps, it helps with

12:45

is confidence. And confidence actually allows you

12:47

to say, I don't know what I'm

12:49

talking about, I'm just wrong, okay, so

12:51

I learned this and move on, much

12:53

more so than I am definitely right,

12:55

and I'm making all the right decisions

12:57

because of my confidence and my experience.

13:00

It's more like, when I'm on a

13:02

podcast six months from today, it's going

13:04

to sound like I'm smart, and I

13:06

did this on purpose, but in reality,

13:08

I'm just kind of going with what

13:10

the good. transition to talk about the

13:12

feature set, which was one of the

13:14

more interesting parts of the experience. Let's

13:16

do it. Okay. So I think of

13:19

this as like an accordion, like out

13:21

in and in. So what we did

13:23

is we built a very bad version

13:25

of all the features. We looked at

13:27

what we thought people needed. for an

13:29

AI voice agent, like a receptionist that

13:31

did all this stuff, that transferred calls

13:33

and set appointments and connected to your

13:35

CRM, answered questions, gave directions, all these

13:37

different things. And we put that out

13:40

there, then we put some energy and

13:42

money behind cold email because we just

13:44

wanted to talk to as many people

13:46

as possible immediately. So we spent money

13:48

for two months to just send out,

13:50

you know, 500 emails a day. And

13:52

that just gave us people to talk

13:54

to. Then we got them into the

13:56

product, somewhere in August. Immediately, like no

13:59

self-serve, no sign-up, no credit card, just

14:01

like ugly. What happened was that we

14:03

took the big feature set and went

14:05

to market. with it with our initial

14:07

cohort and then we immediately start to

14:09

see what people actually cared about. Out

14:11

of our call it eight features, what

14:13

do people care about? It became very

14:15

very obvious that there were two or

14:18

three, that everyone really needed and the

14:20

other ones were nice to have. And

14:22

what that allowed us to do is

14:24

to then shrink back down and we

14:26

removed the features. We removed them from

14:28

the admin entirely, right? They were in

14:30

code, but you couldn't see them. And

14:32

what that allowed us to do was

14:34

basically identify... what the features of the

14:37

base plan should be. So a whole

14:39

bunch of features, everyone cares about these,

14:41

remove all the other ones, those are

14:43

the nice to have, the ones that

14:45

people really, really need, that's our base

14:47

plan. So when we started, we didn't

14:49

have many features, so we only started

14:51

with one plan. We only started with

14:53

one plan. We removed the other ones,

14:55

we just had a $49-a-month plan. And

14:58

what we came to realize was that

15:00

what that is, is an answering service

15:02

toward a receptionist. So answering service basically

15:04

just takes a message. So answering questions

15:06

and taking a message, that's our base

15:08

plan. That's 49 bucks. And then as

15:10

we started to grow with just the

15:12

base plan, we then took those other

15:14

features that were nice to have and

15:17

we added them back into the higher

15:19

tiers. So it's like we had the

15:21

feature set kind of right with some

15:23

changes. But we just removed all the

15:25

ones that people weren't using right away.

15:27

That was our base plan, and then

15:29

we added the other ones back in,

15:31

in a fuller sense, with better understanding

15:33

of how people actually wanted call transfers

15:36

to work, or emergency things to work,

15:38

or sending a text message. You hear

15:40

people talk about MVPs, right? It's a

15:42

minimal viable product. Usually, it's either limited

15:44

functionality or it's just not very pretty.

15:46

It's not very well built, but if

15:48

it solves a pain point, people use

15:50

it. Then you hear some folks today

15:52

saying, well, you can't build an MVP

15:55

anymore because everyone's taste is too high.

15:57

Or, I don't know, there's different times

15:59

change, right? MVP started in almost 20

16:01

years ago now. And so it can't

16:03

be this. same thesis they had then,

16:05

because everything's changed. But it sounds like

16:07

you, would you even, would you consider

16:09

that you launched an MVP, and was

16:11

the accordion approach you talked about intentional,

16:13

or were you kind of like, well,

16:16

we don't really know. So just launch

16:18

a bunch of stuff and see what

16:20

sticks. It was intentional-ish. The direction of

16:22

it was, let's put out a bunch

16:24

of features. We don't actually need them

16:26

all. We can always remove them. So

16:28

that was generally, but you know, the

16:30

actual path it took. We raise money.

16:32

And what that allows us to do

16:35

was basically spend on things that you

16:37

normally wouldn't spend in an MVP context.

16:39

So when I called it ugly, it

16:41

was functionally ugly. Visually, it was not

16:43

ugly. So we worked with a designer

16:45

called Francois from clearly design, highly recommend.

16:47

So all of our stuff was actually

16:49

pretty. Now, the idea of an MVP,

16:51

very, very different in an AI context.

16:54

Another one of the things that I've

16:56

learned is the differences between a traditional

16:58

SAS experience and mindset and approach compared

17:00

to AI. And the reason MVPs in

17:02

AI and SAS are like very different

17:04

things is because in AI you don't

17:06

build nearly as much as you do

17:08

in SAS. So if you want to

17:10

build a CRM, you kind of can't

17:13

launch an MVP. If you want to

17:15

email, for example, you can't compete with

17:17

an MVP, you're not really going to

17:19

get anywhere. But AI, you're building at

17:21

the very, very top layer, the interaction

17:23

layer, the app layer of whatever you

17:25

want to call it. And then underneath,

17:27

you're really leveraging much more infrastructure than

17:29

you normally do. So yes, none of

17:31

us build hosting and servers. Great. So

17:34

everyone leverages that. Or maybe you have

17:36

a framework like Laravel. And you're using

17:38

Tail. the rapper context, you're really leaning

17:40

a lot on other services. So we

17:42

don't do the transcribing, we don't do

17:44

the voice, we don't do the LLM.

17:46

In some ways, like, what are we

17:48

actually doing? It's just that the UI

17:50

layer that allows a three-person painting company

17:53

to have an AI phone receptionist up

17:55

and running in 10 minutes. That's actually

17:57

what we do. And in that context,

17:59

an MVP, much easier to build. That

18:01

makes sense. That's how you move so

18:03

fast because I was going to ask

18:05

you, someone listening to this who says,

18:07

came up with a day in June,

18:09

had something in production in July, had

18:12

paying customers by or something. It's like,

18:14

yeah, wow, that's really fast. I was

18:16

going to ask if you moved fast

18:18

because you have money, because you raised

18:20

money, as you said. But it sounds

18:22

like it's just a simpler product to

18:24

build. If you're people, going from 20

18:26

people to six people, really changed things

18:28

for us, and we demanded things for

18:31

us, and we demanded things for us.

18:33

development process. And then when we went

18:35

to Rosie, we had to say, we

18:37

were super, super strict, because you can't

18:39

mess up. You can't mess anything up

18:41

in production, or you cost your customers

18:43

money and lose their trust. So we

18:45

had a specifically built, very tedious, careful

18:47

development process and deployment process. And then

18:49

when we went to Rosie, we had

18:52

to say, we are going to like

18:54

90% of our process. We have to

18:56

intentionally set it on fire. It's a

18:58

real challenge for our product leader, Jessica,

19:00

but she's done extremely well with it

19:02

because she wasn't ideological about the process.

19:04

She was just like, well, that product

19:06

needed X, this product needs Y, let's

19:08

change. So a lot of the speed

19:11

is attributable to that. And you've kind

19:13

of mentioned or hinted that AI wrappers.

19:15

Rapping AI, I don't know, AI rappers

19:17

is like a bad term now. It's

19:19

like, oh, if you build an AI

19:21

rapper, you have no defensible mode or

19:23

whatever. Like, do you think of what

19:25

you're building? Like, would you call it

19:27

an AI rapper? Is it something different?

19:30

And how do you think about that

19:32

mode? Could me and a team of

19:34

people compete with Deep Seek? Because it

19:36

poked a whole... in the potential value,

19:38

right, the flip side of this, the

19:40

commoditization of the LLLMs. And so it

19:42

put a lot more focus and attention

19:44

on the application layer, and I think

19:46

it evened out the debate a little

19:48

bit. It's not one-sided. You can make

19:51

money or get crushed all up and

19:53

down the stack. So I don't necessarily

19:55

think of us as an AI wrapper,

19:57

and there will be a lot of

19:59

competition, for sure. But there's still... a

20:01

very honest dynamic between service and customer.

20:03

If you provide value and you continue

20:05

providing value, people will stick with you.

20:07

So what if there's competition? So it's

20:10

a bit of a, I would almost

20:12

call it a land grab right now,

20:14

the more customers you get in a

20:16

short amount of time, they are unlikely

20:18

to switch, if what you have is

20:20

working, right? That's right. I forget who

20:22

said it, it was a great term,

20:24

I don't know, I can remember, the

20:26

term clock, but it was like, like,

20:29

like, like, like, like, you, you, you,

20:31

you, you, you, you, you, you, you,

20:33

you, you, you, you, you, you, commitment

20:35

or dedication or to where someone gets

20:37

used to using your app and especially

20:39

and not with non-technical folks like my

20:41

parents are older and they don't man

20:43

they learn exactly so they learn exactly

20:45

which button to click right so that's

20:48

what you're referring to is like hey

20:50

you're an electrician or a lawn care

20:52

person and and stereotypically they don't want

20:54

to learn a new you I right

20:56

I have some very funny conversations on

20:58

Twitter with technical people who are like

21:00

You know, this can be done so

21:02

much cheaper. I'm like, you, yes, you

21:04

keep having that conversation. Go for it.

21:06

Yeah. That is the same reason why

21:09

if you go to our site, we

21:11

have not, we have not taken on

21:13

a vertical. Normally in SAS, what would

21:15

you and I recommend to ourselves and

21:17

to other people? Don't try to be

21:19

broad for everybody, find a niche. The

21:21

niches are big enough. I kill it

21:23

for one type of customer, then expand.

21:25

In this I think it's the exact

21:28

opposite. So I see a lot of

21:30

competitors doing really well, very niched, right,

21:32

AI voice for veterinary clinics, AI voice

21:34

for doctor's offices, AI voice for restaurants,

21:36

and so on. and I think it's

21:38

just not the right time for that.

21:40

I think one of the more interesting

21:42

aspects of building an AI in general

21:44

is that new markets are forming. And

21:47

if you think about our market, the

21:49

phone, you know, gigantic problem, enormous problem,

21:51

answering the phone, dealing with it, not

21:53

missing phone calls, just gigantic. It's very

21:55

rare in our economy that you have

21:57

that size of a problem, and there

21:59

are no solutions. Right, that is just

22:01

rare and it only happens in this

22:03

context because that solution wasn't possible. So

22:06

before, you had voicemail and you had

22:08

answering services that people pay two bucks

22:10

a minute for and are generally not

22:12

that happy about. So now all of

22:14

a sudden this third option comes around,

22:16

that's going to get filled in by

22:18

competition. And I think that land grab

22:20

slash demand rush calls for being very

22:22

wide and horizontal and everyone's welcome and

22:24

this is easy to get started with.

22:27

And so it's like, you know, against

22:29

specialization in many ways specialization in many

22:31

ways. We'll see if in hindsight that's

22:33

a good call in about a year,

22:35

but we'll see. Yeah, and I want

22:37

to call out, you know, as you

22:39

said, it's rare that there are problems

22:41

that are huge that have no solution

22:43

because people see money, big companies are

22:46

smart, they're dumb and they're smart, but

22:48

they do see big money and they'll

22:50

move into a space and so everything's,

22:52

you know, everything's crowded in quotes, but

22:54

these technological shifts. Move all of it

22:56

and then make it possible right the

22:58

internet did this like the World Wide

23:00

Web when it came about it was

23:02

suddenly like Oh, no, there's there's this

23:05

can do all it and social media

23:07

was another one I think of even

23:09

like web to oh email I phone.

23:11

Yes. Remember the iPhone coming out? Suddenly

23:13

there was a huge rush even like

23:15

remember like Facebook apps That was a

23:17

big rush, fear, that didn't pan out

23:19

the way, you know, drones, AI, like,

23:21

crypto, and some of these panned out

23:24

and some didn't, but they at least

23:26

created this, this space that suddenly there

23:28

was a lot that could happen that

23:30

wasn't possible last year. Yes, and the

23:32

key is that there's demand for it.

23:34

Yeah. Because crypto feels like a huge

23:36

innovation, not much demand outside of, you

23:38

know, gambling. Right. But AI is different.

23:40

Yeah. So I do think about the

23:42

email context a lot. If email comes

23:45

out, are you building drip or are

23:47

you building constant contact? Right. If it

23:49

first comes out, you don't need to

23:51

go niche and specialty. You just go,

23:53

hey, we're a milk chip, man, is

23:55

everybody. And it's simple. Like, I think

23:57

constant contact might have been the first,

23:59

or it's kind of the first one

24:01

that's still around, and Aweber was the

24:04

first one to do like sequences like

24:06

sequences. You know, and they're built in

24:08

Pearl, and they're hosted on a server

24:10

rack in downtown Manhattan in a cage,

24:12

and you have to go that, you

24:14

know what I mean? It's that stuff.

24:16

Yeah, crazy. And you have to raise

24:18

half a million dollars just to get

24:20

the thing built and deployed, right? But...

24:23

It didn't need to be drip. It

24:25

didn't need to be complicated at all.

24:27

There were no workflows because you didn't

24:29

need them. And if you could just

24:31

communicate with an audience, you could build

24:33

it. I want to circle back to

24:35

something you said earlier, because I know

24:37

someone listening to this, just thinking about

24:39

it, just thinking about it. You built

24:42

this, just thinking about it. You built

24:44

a bunch of, you built eight features,

24:46

you said, and then you just like,

24:48

boop. Five. In software. we have a

24:50

lot more flexibility than we think we

24:52

do. Whether it's raising prices, taking features

24:54

away, or just coming up with other

24:56

solutions, so what we did, let's say

24:58

for the call transfers, we just asked

25:00

people, do you want to keep using

25:03

it? We're about to take it away,

25:05

do you want to keep using it?

25:07

And they said, yes, we just flagged

25:09

it, okay, fine, you can still see

25:11

that link. That's it. Move on. They

25:13

just knew that we weren't going to

25:15

support it and it wasn't going to

25:17

be good, but they were like, I

25:19

love it so much, I want it.

25:22

And we said, cool, we'll come back

25:24

to you with a better version of

25:26

it in two months. Right now it's

25:28

going to go away for everyone else,

25:30

it's going to stay for you. And

25:32

so in our admin, individual features literally

25:34

have their own links. So we just

25:36

removed the link. Simple as that. That's

25:38

actually how we reintroduceded directly from the

25:41

URL. I like that idea of, because

25:43

we used to do it at all

25:45

the time, feature flags were such a

25:47

popular thing when we were trying to,

25:49

I say popular, we used them all

25:51

the time in drip to figure out.

25:53

For testing and for this type of

25:55

stuff, there were features we built, RSS

25:57

to email is, I hate that. feature

26:00

and we built it because there were

26:02

some early power users that were like

26:04

oh if you just had this I

26:06

would switch from XYZ and I'm like

26:08

I will never use it I don't

26:10

endorse that feature it's a pain in

26:12

the ass it's brittle as hell but

26:14

I'm gonna build it for you five

26:16

because I knew then they would then

26:18

talk about it so we have a

26:21

feature flag and I bet probably tens

26:23

of thousands of users of drip at

26:25

this point and I bet there's like

26:27

a hundred that have that enabled right

26:29

so it and that's a thing you're

26:31

going for tens of thousands of paying

26:33

customers. I mean, that's that's your goal,

26:35

right? Because your price points remind, you

26:37

know, tell people about your price points.

26:40

Sure, we're at 49, 99, and 199,

26:42

and we do have, we have some

26:44

weird issues around minute usage, because our

26:46

cogs are directly tied to minute usage.

26:48

Usually I don't even like to think

26:50

about cogs in the software context, because,

26:52

you know, normally card hooks like $5,000

26:54

bucks for $500,000 in revenue. This is

26:56

not that, it's more linear. And I

26:59

don't like it because people think of

27:01

us as minutes. Do my minutes roll

27:03

over? As soon as I heard that,

27:05

I was like, we got to get

27:07

away from this minutes thing. We want

27:09

to focus on value, not the number

27:11

of minutes. The pricing has been good

27:13

to us so far. So in December,

27:15

we launched our self-serve onboarding. And that

27:18

was the turning point. So up until

27:20

then, it was, do we have it

27:22

right? Do we have the feature set?

27:24

Do we have all this other stuff?

27:26

But it was it was ugly to

27:28

help to get to get people onboarded

27:30

and we purposely did not build the

27:32

onboarding because we assumed we don't know

27:34

what the onboarding should be. So once

27:36

we understood the base plan feature set,

27:39

we looked at it and said out

27:41

of these base plan features, what is

27:43

absolutely necessary to get any value at

27:45

all? And we built that into the

27:47

onboarding. What is the minimum number of

27:49

things you need to do to just

27:51

get value from this thing? And that's

27:53

what our onboarding is. The other part

27:55

of our onboarding, I learned this from

27:58

Justin. McGill, actually, was to take the

28:00

value that's in the admin and drag

28:02

it out into the marketing experience so

28:04

that it starts there. Instead of saying,

28:06

here's a wall that you need to

28:08

get over to get value, AI has

28:10

this opportunity. You see with ChatGPT, you

28:12

see it with mid-journey, you just come

28:14

in to this product, you type something,

28:17

boom, you have value. So we needed,

28:19

I felt we needed to... provide that

28:21

type of an experience. So if you

28:23

go to our site, the onboarding is

28:25

actually part of the sign-up process. So

28:27

it's not, if you want value, get

28:29

over this hurdle of creating an account.

28:31

Instead, it is, come in, give us

28:33

a bit of information about your business.

28:36

So Google Business Profile is our primary

28:38

path. Our secondary path is your, just,

28:40

you know, give us a business name.

28:42

But the primary path gets taken 80%

28:44

of the time. You put in your

28:46

Google Business profile, we pull your information,

28:48

it's very structured data in Google, and

28:50

then the next thing you see is

28:52

your AI voice agent, and you have

28:54

a few clips, and you can just

28:57

hit play, and it'll say, thanks for

28:59

calling PetBusters in Illinois, how can we

29:01

help you today? So you get a

29:03

little taste of the value up front,

29:05

and then we ask you to create

29:07

an account. And then when you get

29:09

to the onboarding, we've already ingested your

29:11

website, Google Business Profile, all your information,

29:13

your business hours, all the stuff, and

29:16

you're like, whoa, you know, I'm almost

29:18

done. So that onboarding is kind of

29:20

what changed the trajectory. And we just

29:22

put the other features at the third

29:24

step and just keep exploring. So you

29:26

get through the onboarding, you get the

29:28

value, you get a phone number, you

29:30

can call your agent, and you are

29:32

blown away in less than three minutes.

29:35

You are a painter, somewhere in the

29:37

suburbs of Illinois, and you sign up

29:39

on your phone while you're in your

29:41

truck, and within three minutes, you're like,

29:43

holy cow, I have this thing. I

29:45

made this thing. It knows my business

29:47

and can answer my phone. And that

29:49

creates enough desire to get through the

29:51

last part of onboarding, which is actually

29:54

forwarding your phone calls to us. And

29:56

as someone, you who built cart hook

29:58

that Ryan rally... that required, the onboarding

30:00

was not three minutes to see value.

30:02

The onboarding was extensive and it was

30:04

often getting developers and right right right

30:06

so you've seen both sides of it

30:08

I see both sides of it with

30:10

tiny sea companies you are in an

30:12

incredibly luxurious position right now I mean

30:15

I'm blown away like three minute onboarding

30:17

that's amazing if you can do it

30:19

I do wake up every day and

30:21

pinch yourself and think this is it

30:23

this is great. So this is one

30:25

of the few things that I can

30:27

actually say was completely deliberate and it

30:29

was a reaction to pain Cardhook was

30:31

pain, rally in terms of onboarding, pain,

30:34

and one of the requirements in this

30:36

product was self-serf. And I think that

30:38

has served this well because voice and

30:40

AI is really, really powerful. And when

30:42

something's really powerful, you're tempted to bring

30:44

all that power to the user. And

30:46

that's what creates onboarding friction. So a

30:48

lot of our competitors are like, sign

30:50

up for a demo or build this

30:53

visual workflow and choose your AI model.

30:55

And we were like. Absolutely not. Under

30:57

five minutes to value and to onboarding.

30:59

So that was very deliberate and it

31:01

does feel incredible. Part of the product

31:03

market fit sense came after we launched

31:05

self-serve and we watched completely non-technical people

31:07

sign up, get their phone number, make

31:09

a phone call, and then put their

31:12

credit card in. So we put the

31:14

credit card on the other side of

31:16

testing your agent. One of the more

31:18

important experiences we had was around pricing.

31:20

So maybe it's worth kind of like

31:22

taking a little detournt to that because

31:24

that kind of surprised us in general.

31:26

When we first started we had a

31:28

seven-day trial with a credit card required

31:30

and for the first few weeks we

31:33

looked like geniuses because everyone was converting.

31:35

Seven days goes by, your credit cards

31:37

on file, you convert and it felt

31:39

good and then we looked under the

31:41

hood at the usage and I was

31:43

like we are building our castle on

31:45

quixant, not even sand like wet quixant.

31:47

And it's because people weren't using it.

31:49

So what we did is we switched

31:52

from a seven-day trial to a 25-minute

31:54

trial. And what that did is it

31:56

went. from time to usage-based. So now

31:58

everyone that converts is an activated user

32:00

that's already converted their behavior, which is

32:02

partly why our churn is so, so

32:04

low, because no one's converting that isn't

32:06

using it. That's the thing. Usually with

32:08

a lot of SAS credit card up

32:11

front, the first 30 or 60 days,

32:13

the churn is a lot higher, and

32:15

then it drops way down, because people

32:17

are using it as an extended trial.

32:19

Yes, they want to try fine I'll

32:21

pay you the 50 bucks because I'm

32:23

so interested in this Let's see if

32:25

it works. We flipped that on its

32:27

head and that has made everything much

32:30

much healthier And so when we did

32:32

that we at we we took the

32:34

credit card requirement away from the on

32:36

boarding and we put it on the

32:38

other the end of the on boarding

32:40

as an optional step and so to

32:42

see someone come in self self-serve make

32:44

a phone call and then put the

32:46

credit card in means they were impressed

32:48

enough that they said, I don't want

32:51

this to go away, I'm going to

32:53

put my credit card in, even if

32:55

it's not required. And then 25 minutes

32:57

goes by and then they convert and

32:59

all of a sudden we have like

33:01

a real user as opposed to, hey,

33:03

seven days has gone by and now

33:05

we have more revenue. Right, and for

33:07

folks listening, 25 minutes is not a

33:10

linear time, it's minutes, it's minutes of

33:12

talk time, right? It's someone calling in

33:14

and it's 25 minutes of interacting with

33:16

interacting with Rosie. Reveal itself to effectively

33:18

be a six-day trial or whatever else

33:20

now Here's the interesting thing you said

33:22

building a healthier business by doing this

33:24

because you're making sure that only That

33:26

the people who convert are only those

33:29

who are actively using it But if

33:31

you didn't do that if you just

33:33

did the credit card in seven day

33:35

you would probably you'd have more revenue

33:37

right now You'd grow faster significantly as

33:39

a founder. How do you wrestle with

33:41

that? That's the right decision Or how

33:43

do you internally think, boy, let's say

33:45

you could be, you said you're doubling

33:48

every month, you could be quadrupling every

33:50

month, literally. And that is pretty tantalizing,

33:52

right? That's actually what we see with

33:54

some of the, you know, the Big

33:56

Silicon Valley companies, right? A bowl too.

33:58

the payment process are not bolt-a-eye, where

34:00

they raise all this money and they're

34:02

trying to, da-da-da-da, and then they just

34:04

crash and burn, right? So how do

34:06

you reconcile that as a founder who's

34:09

ambitious, you want to grow as fast

34:11

as possible, but you're like, well, I'm

34:13

gonna, I'm gonna not put the pedal

34:15

to the metal, because I think it's

34:17

an unhealthy business. You want to grow

34:19

as fast as possible, but you're like,

34:21

well, I'm gonna, I'm gonna, I'm gonna

34:24

not put the not put the pedal,

34:26

put the pedal, put the pedal, put

34:28

the pedal, put the pedal, I put

34:30

the pedal, and we're, and we're, So

34:33

that made the decision to go into

34:35

it lighter. The second part of that

34:37

is, I have seen what churn does.

34:39

And I have googled maximum churn given

34:41

growth rate and... You know, it's not

34:44

good. Yeah. It's not good. If anyone

34:46

doesn't know what I'm talking about, there

34:48

is a formula for your maximum MRR

34:50

given... your growth and your churn. And

34:53

you do not care about it at

34:55

50K MRR. At 300K MRR you really

34:57

care that your maximum is 350. Because

34:59

here we are, the wall has arrived.

35:02

At Cardhook we found the wall. Because

35:04

we were churning at 15% a month.

35:06

So we were skyrocketing, but churning like

35:08

crazy. I called it a washing machine.

35:11

And if you recall, what I did

35:13

is I went to the support team

35:15

and I said what percentage. of your

35:17

interactions or with people that will not

35:19

become paying customers. And they were like

35:22

at least 50% at least. And I

35:24

was like, that's an absolute disaster, not

35:26

only for the support team, but also

35:28

because we're wasting time on people that

35:31

don't care as much and we're not

35:33

giving time to the best people. So

35:35

that really gnarly decision at Cardhook to

35:37

force a demo and only take on

35:40

people that we thought were right and

35:42

raise prices at the same time, that

35:44

taught me. you actually are better off

35:46

building a better business as opposed to

35:49

a faster business. So I was more

35:51

open to it than we saw it

35:53

as an experiment and it worked. And

35:55

then there's one final step in the

35:57

process that just came out I think

36:00

earlier this month in February that has

36:02

again changed. the trajectory again and that

36:04

was what we call premium UI. So

36:06

up until the beginning of this month

36:09

you looked at the feature set on

36:11

the marketing site and on the pricing

36:13

page and then you signed up and

36:15

you just didn't see the premium features.

36:18

You would basically have to ask us,

36:20

hey, that call transferring thing, or this

36:22

custom training, can I hear more about

36:24

that? And then we would say, yes,

36:26

that requires this higher tier. So we

36:29

had no mechanism for people to discover

36:31

the higher tiers. And instead of just

36:33

jamming it in there, we very deliberately

36:35

went with the designer and a product

36:38

team and said, how do we make

36:40

this not only incredibly attractive, but also

36:42

self-serve? And so now you go through

36:44

the onboarding and that third step, the

36:47

keep exploring, is now expanded. And as

36:49

you navigate the admin, there are individual

36:51

elements with a little yellow star. And

36:53

you click on those and we reveal

36:56

the feature and we clearly state this

36:58

is a premium feature that requires these

37:00

plans. Do you want to use it?

37:02

Here's what will happen. And building that

37:04

self-serve. Now we have a mechanism for...

37:07

increasing our poo. So now everyone comes

37:09

in at the at the base tier

37:11

at the $49 tier and now the

37:13

percentage we're almost at 50% of our

37:16

revenue is now at the higher tiers

37:18

right and then that starts to argue

37:20

for expanding the strategy around that lower

37:22

tier. Do we have a free tier?

37:25

Do we lower it to 29 like

37:27

now that we have more confidence around

37:29

a self-serve revenue expansion now it starts

37:31

a factor into our you know marketing

37:34

plans and our advertising and our positioning

37:36

and pricing and all that. Man, I'm

37:38

looking forward to following this story both

37:40

through investor updates and having you back

37:42

on the show in I don't know

37:45

six or 12 months to kind of

37:47

follow up because everything changes I mean,

37:49

it's if you look back at the

37:51

past five or six months It's changed

37:54

dramatically for you and I imagine building

37:56

in building in AI in that world,

37:58

things are going to change quick for

38:00

you again. Yes, I just sent a

38:03

tweet out yesterday that encapsulates this extreme

38:05

optimism and paranoia. I think that is

38:07

the appropriate stance for the CEO of

38:09

this company is we are going to

38:11

push so hard, we're going to crush

38:14

it, we're going to get to 10

38:16

million ARR in 18 months, and at

38:18

the same time at any minute. Someone

38:20

could release something, someone could do something

38:23

like, so we have to hurry. And

38:25

I think AI kind of, that's what

38:27

founders in this space right now are

38:29

feeling. No one should be overconfident about

38:32

anything unless you're lovable and you got

38:34

the $17 million ARR on three months

38:36

because then you find to be comfortable.

38:38

There you are. Yeah. So folks want

38:41

to read that tweet. They can head

38:43

to Jordan Gaul. That's GAL on Twitter.

38:45

And of course, if they want to

38:47

check out Rosie, what we've been talking

38:49

about for the past 30 minutes, the

38:52

AI answering service for your business for

38:54

your business calls, they can head to

38:56

Hay Rosie. Jordan, thanks so much for

38:58

joining me. Thank you very much. I

39:01

want to give thanks again to Jordan

39:03

for coming on the show and sharing

39:05

his wisdom. He is a founder that

39:07

executes and oftentimes when I'm saying on

39:10

this show, you know, the founders I

39:12

know who succeed do X, Y, and

39:14

Z, he's one of those founders that

39:16

I think about. Like he is a,

39:18

if you know, coding, he's a design

39:21

pattern for doing smart things executing well,

39:23

getting shit. done, moving fast, a little

39:25

bit of gut, a little bit of

39:27

data, and being super scrappy and making

39:30

it work. So it's always great to

39:32

have Jordan on the show, and I

39:34

hope you enjoyed The Conversation. This is

39:36

Rob Walling signing off from episode 764.