0:00

Look, I I think the most important thing...

0:02

thing beyond his if you're trying

0:04

to make great products, you need

0:06

to have products, you need You need

0:08

people who are making to so

0:10

deeply understand deeply the experience

0:12

of a building, the pain that

0:14

people are going through, are going through. you know,

0:16

what they're actually, you know, the customer

0:18

at the end is actually doing

0:20

there to run your business that you

0:22

understand it not just decently, but

0:24

in some cases but in some cases the customer

0:26

has. has. think only then can you

0:28

actually build products that are so

0:30

well that they can they can actually the the

0:32

task, they can do it more efficiently. I And I

0:35

think part of why you know, as we're you know,

0:37

as we're releasing products, engineers are

0:39

on the call with customers when we

0:41

ship, they're accountable for metrics and how

0:43

it ultimately performs. And what I use you I

0:45

use, you really won't find people

0:47

at who haven't talked to customers with

0:49

any level of recency. And I think

0:51

that's just a core part of makes

0:54

makes great product cultures. Joining

1:11

us today is my friend Eric Gleiman, -founder

1:14

and CEO of RAMP, which offers companies

1:16

AI -powered financial tools to manage their

1:18

spending and expense processes. processes.

1:20

Many companies rushed to add add after

1:22

the the chat gPT moment. But Eric has

1:24

a different take. He He doesn't think that chatbots

1:26

are the right form factor for everything. everything. I

1:28

I mean, who wants to chat with their with their expense report?

1:30

He thinks zero -touch automation that works invisibly

1:33

in the background can be much more valuable

1:35

in many cases. more In fact, he thinks a

1:37

better analogy is to self -driving cars, a or

1:39

in this case, self -driving money. We'll

1:41

talk about in this to building an AI and

1:43

how Eric sees the space evolving. ramp's approach

1:45

to the show. an AI and how Eric sees the space

1:48

evolving. we are so happy that

1:50

here. so happy that we, We, you

1:52

know, the original invitation for for this podcast, as

1:54

you know, happened over dinner and we

1:56

were so happy that you decided to do

1:58

it with us. So, do it us. have tons

2:00

of questions for you about the future

2:02

of AI and how it's going to

2:05

impact finance and businesses. But for the

2:07

very few people that don't know, can

2:09

you mind? Do you mind just saying

2:11

a few sentences on what ramp is

2:13

and what you all do? For sure.

2:15

I mean, look, we're dedicated to making

2:18

companies more profitable and operate more smoothly.

2:20

The way you can think about ramp

2:22

is, you know, we built a command

2:24

and control system for company finances. So

2:26

from one place you can issue cards.

2:28

manage approvals, make payments of all kinds,

2:31

and even automate closing your books. And

2:33

so for your finance teams, it means

2:35

that your operations are simpler, you can

2:37

automate a lot of business processes, and

2:39

it surfaces up data and intelligence on

2:41

how your company can spend less. And

2:44

so the upshot is the average company

2:46

using RAMP is able to save about

2:48

5% per year. on their expenses, which

2:50

is pretty material. You know, over the

2:52

past five years we've been in business,

2:54

this is added into billions of dollars

2:57

of savings and the equivalent of thousands

2:59

of the years of labor that's been

3:01

saved. And so, whether it's large publicly

3:03

traded companies like Shopify, Virgin Voyages, Boys

3:05

and Girls Club in America, to 25,000

3:07

other businesses in spring and benefit today,

3:10

and a lot of what we're trying

3:12

to do is really answer the question

3:14

of how do you make people more

3:16

productive with their time and with their

3:18

money. I love it. Well, one thing

3:20

that, you know, I was surprised by

3:23

before we get into, you know, how

3:25

ramp is using AI, you and Karim,

3:27

you know, I saw you guys posted

3:29

this the other day, you built an

3:31

AI agent in 2015 at Parabas? I

3:33

mean, this is way before it was

3:36

cool. What did you build back then?

3:38

And maybe tell us about that, and

3:40

then we want to hear about what

3:42

you've built now and how it's changed

3:44

over the last nine years. Sure. Yeah.

3:46

So, um, Parabas was a much, um,

3:49

it was a very simple tool and

3:51

I think it is fair to call

3:53

it an agent. Basically, um, uh, go

3:55

back to a decade ago. If you

3:57

were buying something online, whether it's at

3:59

Amazon, Best Buy, All these stores would

4:02

guarantee that if you bought some TV

4:04

for a thousand bucks, the next day

4:06

it went on sale for $900, you

4:08

could get the difference back if you

4:10

asked them. These were price adjustment policies.

4:12

And so we built it was an

4:15

app that integrated with your email software,

4:17

your Gmail, Yahoo, whatever you used, scanned

4:19

your inbox for receipts, tracked the prices

4:21

of what you bought. And if there

4:23

was a price drop, you were eligible

4:25

for something different, it generated an email

4:28

to sound like you. sent it to

4:30

the store, the store customer service responded,

4:32

and you know, you wake up to

4:34

$100 back the next day. And so

4:36

that was it. It was this agent

4:38

that lived in your inbox and help

4:41

you help you save money. And well,

4:43

I guess I would say like a

4:45

couple of things. I mean, first. I

4:47

think if many people are thinking about

4:49

agentic AI and how is it going

4:51

to change tools in this new use

4:54

case, I would argue it's actually been

4:56

around for a long time. This is

4:58

a decade ago. Millions of people use

5:00

this in order to get price drop

5:02

refunds from stores and retailers, but it

5:04

was a very narrow use case. When

5:07

I think about AI today in 2024,

5:09

I think the sense of use cases

5:11

where computers can be doing work on

5:13

your behalf is far easier. It's not

5:15

just going to be these narrow surfaces

5:17

like price adjustment guarantees. It's really intensive

5:20

and hard to do that, but more

5:22

generalized reasoning, helping your finance seem run

5:24

more efficiently. Maybe not just asking for

5:26

price refunds, but maybe even helping you

5:28

negotiate. and are running more complex analyses,

5:30

but I think this primitive of software

5:33

that does things on your behalf has

5:35

been here for a while, and I

5:37

think is going to be growing pretty

5:39

quickly. That's a really good segue into

5:41

what you're building now at ramp. I'd

5:43

love to learn more about your vision

5:46

for a ramp by AI. Yeah, for

5:48

sure. I mean, first, like, you know,

5:50

forget even just just this ramp AI,

5:52

like we're we're dedicated to this mission

5:54

of helping people spend less money and

5:56

spend less time. I think we're very

5:59

excited about AI because it's a new.

6:01

of tools that enables us to do

6:03

this, you know, but ultimately comes down

6:05

to like where where's the pain point?

6:07

Are there processes in your business that

6:09

results in your company spending more time

6:12

than you need to? Or are there

6:14

areas in your business where you're spending

6:16

too much money? And so, you know,

6:18

when we think about where things are

6:20

going to go over time, you know,

6:22

so much of running a finance team

6:25

and Rubby, you know this from your

6:27

time as being CFO to Insta Cart,

6:29

there's a lot of tedium and monotony.

6:31

You know, you're trying to grow your

6:33

business and yet when you look deep

6:35

in your finance team, you'll find people

6:38

who are auditing expense reports by hand.

6:40

They're downloading spreadsheets in order to be

6:42

able to tag and classify vendors. They're

6:44

rerunning analysis time and time again. And

6:46

what we're trying to do at ramp

6:48

our vision is how do you take

6:51

these tedious and monotonous tasks? And either

6:53

through better design integrated tools automate this,

6:55

and so the simple way to understand

6:57

this is instead of eating two apps

6:59

to buy one thing, your American Express

7:01

and your concur, you know, where people

7:04

are going in and out adding a

7:06

receipt. your ramp card and through zero

7:08

touch where we will pull the receipt

7:10

from the merchant from your email, your

7:12

expense report is not just easier to

7:14

do, but it's done for you. So

7:17

there's a lot of cases of zero

7:19

touch expenses, better categorization. We sort of

7:21

apply large language models to the transaction

7:23

data itself as well as your general

7:25

ledger. So we can auto complete expenses

7:27

for you certainly faster and for the

7:30

vast majority of customers more accurately. So

7:32

your books are effectively doing themselves. And

7:34

over time, ramp should be able to

7:36

point to ways for your business to

7:38

operate vastly more efficiently. Same sets of

7:40

vendors, better prices, same business processes done

7:43

automatically for you. And so there's a

7:45

lot in just the manifestations of how

7:47

ramp helps your finest scene run more

7:49

efficiently. There's also ways we're trying to

7:51

experiment with even how should people interact

7:53

with. Should it be something where you're

7:56

prompting ramp to do these things or

7:58

should it be taking care of these

8:00

things for you and doing working to

8:02

happen? So we can go either way,

8:04

but that's that's a loose framework on

8:06

how we think about it. And how

8:09

do you decide what's in scope? Like

8:11

anything that makes the finance team more

8:13

efficient? Is that what's in scope for

8:15

you or what's in scope and what's

8:17

out of scope? Yeah,

8:19

definitely. I mean, I think, I think,

8:22

I think, that's probably the right way

8:24

to put it. You know, it really

8:26

is will, can we create a product

8:29

that should save time and money for

8:31

foreign customers? If so, I think it's

8:33

in scope. You know, today ramp is

8:35

largely used for payments out of a

8:38

business. So I think hard expenses, bill

8:40

payment expenses, business processes on top of

8:42

that. But when you think about You

8:45

know, over time, building towards self-driving money,

8:47

you know, finance department, that's improving itself,

8:49

I think, thinking about higher yields, more

8:52

efficient collections, to even more efficient record

8:54

keeping, I think, are all things that

8:56

we're thinking about over time. So I

8:59

would argue there's very little out of

9:01

scope and I would say in the

9:03

same way. You know, we think about

9:05

our souls more as a productivity company

9:08

than we do is just a money

9:10

company. And so anything that makes the

9:12

productivity of capital in your business go

9:15

higher are things that we should be

9:17

thinking about. Eric, what is the intuition

9:19

then on your side? You mentioned you're

9:22

thinking about. Eric, what is the intuition

9:24

then on your side? You've mentioned you're

9:26

thinking about how folks are actually going

9:28

to use AI within ramp, right? So,

9:31

I mean, maybe just like a couple

9:33

of notes stepping back. You think about

9:35

the first phase, you know, what happened

9:38

sort of after a chat GPT came

9:40

out. Suddenly there was a chat bot

9:42

slapped on everything, you know, and people

9:45

say, finally it's your app and you

9:47

can talk to it. And, you know,

9:49

I've personally never met anyone who says,

9:51

you know, I just wish I could

9:54

chat to bank account. I wish I

9:56

could chat with my expense report. You

9:58

know, and so I would say, like,

10:01

I totally get the skepticism of, like,

10:03

the first set of people who find

10:05

this really silly. And I think that

10:08

we've tried to classify internally, and when

10:10

we think about the sets of user

10:12

experiences, probably in the two buckets, I

10:14

mean, the first, we think a lot

10:17

about this paradigm of really zero touch

10:19

AI. And I would argue even Parabas

10:21

was an early example of this. The

10:24

interaction was actually no UX. You would

10:26

sign up, you'd link your email, and

10:28

that was it. And you'd wake up

10:31

the next day and it was done

10:33

for you. And there's a whole category

10:35

of experiences like this. You tap your

10:37

card, your expense report is done for

10:40

you. Your accounting is done on your

10:42

behalf. We're suggesting memos for you to

10:44

click 123. And so we're trying to

10:47

predict. and understand what are the user

10:49

interaction inputs we may need. Can we

10:51

get to a level of confidence where

10:54

we can do this all the way

10:56

and can we expand the surface area

10:58

of these? And so there's a whole

11:00

set of things that happen from the

11:03

administration to the analyses to the operational

11:05

closure of tasks that you can do.

11:07

And so I think there's a large

11:10

category there. We can unpack and talk

11:12

about the expansion surface area. And there's

11:14

this next surface area, we think, you

11:17

know, internally we think about it as

11:19

agentic AI, where you're going to want

11:21

to effectively prompt some kind of outcome,

11:23

maybe monitor what's being done along the

11:26

way, or once you, you know, get

11:28

a level of trust, have that done

11:30

for you. So it's been a use

11:33

case. A few weeks ago, we launched

11:35

right out there, the launch of GPT-40,

11:37

the new multimodal model from Open AI,

11:40

which had vision, audio, and text, understanding

11:42

capabilities. We created a way for customers

11:44

to just ask. what it wanted done,

11:47

and ramp would show you how to

11:49

do it and do that on your

11:51

behalf. So you could say ramp, you

11:53

know, I'd like to issue a card

11:56

with a budget of $50 that only

11:58

can be used at Starbucks. Go. And

12:00

then effectively, the large language model could

12:03

read everything that's on your screen, everything

12:05

you're seeing, and would guide you through

12:07

steps in order to how to do

12:10

this. We call this the tour guide,

12:12

so which you click here, enter in

12:14

this text, and you could monitor this.

12:16

And you can see with extreme accuracy.

12:19

Again, we didn't build into the interface,

12:21

click here, this wasn't a managed demo.

12:23

The reasoning of large models are now

12:26

capable to go and do that for

12:28

you. And I think there's a whole

12:30

class of services and I would argue

12:33

the majority of business is actually done

12:35

this way. The interface that most people

12:37

think about this today is, I think

12:39

as you've joked once before, Ruby, you

12:42

know, you hire an analyst and you

12:44

say, hey, I want this thing done,

12:46

go figure this out for me. And,

12:49

you know, that's your interface. And there

12:51

have been forms of AI that have

12:53

been around a long time, as I

12:56

mentioned, too. Yeah. Exactly. You know, you've

12:58

had a large language model before. It's

13:00

an anal spin. No, but I think

13:02

that's actually a good way of thinking

13:05

about these things, right? I'd actually say

13:07

it's kind of strange that the status

13:09

quo today is, you know, people are

13:12

trained to learn how to use your

13:14

app. Instead. Like, I think it makes

13:16

more sense to me that people should

13:19

be thinking about how to run their

13:21

business, how to sell the customers, how

13:23

to, you know, find ROI versus how

13:25

to manage the intricacies of how your

13:28

app is designed. And I think, done

13:30

right, you should be able to say,

13:32

this is what I'm trying to accomplish,

13:35

go get it done in the early

13:37

era. You may monitor that, but over

13:39

time as you build trust in that

13:42

process and the reasoning capabilities increase, you'll

13:44

see it go. And I might draw

13:46

an analogy even to you two, the

13:48

progression in self-driving cars. You know, when

13:51

you think a decade ago around the

13:53

hype of, you know, self-driving. would have

13:55

people sitting in the driver's seat ready

13:58

to take over, you know, with a

14:00

steering wheel, and now there are wamos

14:02

all around the Bay Area. And it's

14:05

strange when there's someone in the seat,

14:07

because the capabilities are actually more accurate.

14:09

And I think there's going to be

14:12

sets of tasks over a time that

14:14

will get handed off. And I, you

14:16

know, you can have an agent, its

14:18

own driver driving parts of your organization

14:21

where there's high fidelity. So I'll pause

14:23

there. By the way, there's so many

14:25

jokes I could make about Ravi handing

14:28

off work to others, but I'll hold

14:30

myself back. You know, some of us

14:32

have been focused on ROI for longer

14:35

than you. I'm not going to apologize

14:37

for that. Eric, if

14:39

you had to guess, like a

14:41

lot of what you just described

14:43

was almost more conversational. The way

14:46

you kind of get an agent

14:48

to do something, like, is the

14:50

way you go, the way that

14:52

Ravi would go talk to, talk

14:54

to a teammate. Do you think

14:56

that means that kind of software

14:58

interfaces as we know them in

15:01

terms of, you know, buttons and

15:03

knobs, etc. kind of go away

15:05

and fade into the background? Or

15:07

how do you think that all

15:09

plays out and like, like, like,

15:11

like, like, AI first user experience,

15:13

First, yes, I do. I do

15:16

think so, very much so. And

15:18

I think that great design actually

15:20

is about understanding, you know, the

15:22

job to be done so well,

15:24

that you can reduce steps, you

15:26

can make things easier, more intuitive,

15:29

and effortless to get things done.

15:31

And, you know, I think balancing,

15:33

you know, not only really the

15:35

range of powerful tools that you

15:37

can do against and outcomes that

15:39

you want to drive to the

15:41

simplicity of interface is always going

15:44

to be attention. But I would

15:46

say we think about that quite

15:48

a bit, you know, internal air

15:50

ramp. And so what I would

15:52

say is I think you will

15:54

want to be able to audit

15:56

to understand functionally what are these

15:59

powerful tools doing for you, but

16:01

done right. Yes, I think it's

16:03

going to be more, in many

16:05

cases, just prompting, here's the outcome

16:07

I want to drive. And the

16:09

tool will go and drive it,

16:12

which just is like a brief

16:14

aside, like for us, you know,

16:16

I'll put its way. over 25,000

16:18

businesses using ramp. Some are run

16:20

by sophisticated finance teams. You know,

16:22

others are, they're running small businesses.

16:24

There's lots of complexity and baggage

16:27

that prevent them from focusing on

16:29

the areas that really generate a

16:31

lot of value from them, meeting

16:33

new clients, writing better pitches, investing

16:35

in the parts of their business

16:37

that generate return, building the next-grade

16:39

products. in so much time I

16:42

think is stolen by people having

16:44

to learn an amalgamation of tools

16:46

stitching together processes and done right

16:48

the world should feel more frictionless

16:50

which should feel more smooth. And

16:52

so I would say like I

16:54

think if we can't accomplish that

16:57

I think it's been a failure

16:59

as an industry to deliver that

17:01

for people. Well, one thing, Eric,

17:03

that I'm curious about is you

17:05

think about ramp has a lot

17:07

of pride in its culture, right?

17:10

And, you know, it's talked about

17:12

externally. It's something that you guys

17:14

are known for speed and quality

17:16

and real customer obsession. How

17:19

did you build the culture and

17:21

engineering in particular, you and Karim,

17:23

that sort of embraces wanting to

17:25

implement AI and wanting to do

17:27

it well? What did you guys

17:29

do there? Because I think that

17:31

there are places that are a

17:33

bit more resistant to change and,

17:35

you know, less embracing of a

17:37

change like this. Well, so a

17:39

couple of things, I mean, I

17:41

would go back to again, like,

17:43

what, what is ramp like? We're

17:45

not some AI driven finance tool

17:47

that we're going marketing. Do you

17:49

want, do you want to adopt

17:51

AI in your finance team? No,

17:53

it's, it's, we exist to save

17:55

you time and money. We lead

17:57

with the benefit. We talk about

17:59

the outcome that we're trying to

18:02

drive and we put these in

18:04

simple terms, but I would, I

18:06

would say it's, it's actually not

18:08

possible, it's not possible, it's, it's

18:10

actually not possible, it's, it's, it's,

18:12

it's, it's, it's, using AI. When

18:14

people say that it's so easy

18:16

to submit expense reports or that

18:18

suddenly my books are getting closed

18:20

faster, it's often because AI is

18:22

inserted at lots of different parts

18:24

throughout the process. And so what

18:26

I would say is. I think

18:28

one of the first things that

18:30

we did to try to get

18:32

at that was focus on what

18:34

is the problem. We are trying

18:36

to help businesses operate using less

18:38

time, fewer hours, and less capital.

18:40

And AI was a means to

18:42

an end, but was not the

18:44

end. So we didn't want to

18:46

have technology in search of a

18:48

problem, but really focus on what

18:50

is the problem. Then once you

18:52

start to decompose the question of

18:54

like what are all the areas

18:56

we, you know, that are wasting

18:58

lots of time, it turns out

19:00

a lot of time, it's process

19:02

automation. And so I think it

19:04

just became like the right tool

19:06

for the job in so many

19:08

different cases. And I think rather

19:10

than trying to say, would you

19:12

like to, you know, do your

19:14

expense reports. Yeah, exactly. Exactly. And

19:16

I think that the other thing

19:18

too is in so many businesses,

19:20

I know there are a lot

19:22

of founders listening to this, I

19:24

think abstract makers away from the

19:26

problem directly, which is to say,

19:28

you can go to a small

19:30

startup with 10 people and Everyone's

19:32

talking to users. And then somehow

19:34

you go to a company with

19:36

500 to 1,000 people and you

19:38

try to figure out who's talked

19:40

to a customer over the last

19:42

week, hopefully all the sales people.

19:44

But you know, you start talking

19:46

in marketing and engineering and people

19:48

haven't done it. And what I

19:50

would argue, look, I think the

19:52

most important thing beyond is empathy.

19:54

If you're trying to make great

19:56

products, you need to have great

19:58

taste. You need people who are

20:00

making to so deeply understand really

20:02

the experience. building the pain that

20:04

people are going through, you know,

20:07

what they're actually, you know, the

20:09

customer theory is actually doing there

20:11

to run your business that you

20:13

understand it not just decently, but

20:15

in some cases better than the

20:17

customer has. And I think only

20:19

then can you actually build products

20:21

that are so well designed that

20:23

they can actually automate the task.

20:25

They can do it more efficiently.

20:27

And I think part of why.

20:29

as we're releasing products, engineers are

20:31

on the call with customers when

20:33

we ship. They're accountable for metrics

20:35

and how it ultimately performs. And

20:37

what I use, you really won't

20:39

find people at RAM who haven't

20:41

talked to customers with any level

20:43

of recency. And I think that's

20:45

just a core part of what

20:47

makes great product cultures. Eric, since

20:49

you've been building with AI for

20:51

so long, do you think that,

20:53

you know, the last year or

20:55

two with these foundation models has

20:57

been a kind of discontinuous step

20:59

change in terms of your ambitions

21:01

and what you're building with AI?

21:03

And in what way? Or do

21:05

you think it's kind of just,

21:07

you know, gradually compounding more and

21:09

more, more more AI magic in

21:11

the product? I think it has

21:13

all sorts of ramifications for builders,

21:15

right, and people building software. You

21:17

know, when I think about like

21:19

what are the sources of durability

21:21

and moats in a lot of

21:23

software businesses, sometimes it's, you know,

21:25

there's more features, there's more integrations.

21:27

There's lock-in. You've been using my

21:29

tool for 10 years and it's

21:31

really hard to take all your

21:33

data out. And I think is

21:35

now, Today functional you can have

21:37

human level reasoning and in some

21:39

cases superhuman level reasoning available through

21:41

an API. I think it has

21:43

profound ramifications for people building businesses.

21:45

I think it's expressed not just

21:47

in the ability to understand large

21:49

sets of data and act on

21:51

it, but it's a wider variety.

21:53

I think that there's, you know,

21:55

I can say. you know, beyond

21:57

even just the services we're providing

21:59

to RAM. Part of how we've

22:01

grown so quickly is we have

22:03

AI automation in outreach, or we

22:05

have SDRs that are multiple times

22:07

more productive than at competitors. It's

22:09

changed how we do customer service.

22:12

You know, it's changed how we

22:14

do copywriting. We can listen to

22:16

100,000 sales calls at once and

22:18

ask what did 100,000 people think.

22:20

It's just things that weren't possible.

22:22

you know, even just a few

22:24

years ago. And so I think

22:26

it's changed really rapidly. And I

22:28

don't think most people are really,

22:30

I think people are experimenting in

22:32

some cases with chat GPT, which

22:34

is great. But I think far

22:36

too few people have actually started

22:38

to incorporate into the crevices of

22:40

how they're actually working day to

22:42

day and have felt it. But

22:44

I think it should accelerate. When

22:46

you think about that, Eric, just

22:48

that you're talking about the rate

22:50

of change and how people are

22:52

kind of scratching the surface on

22:54

this, what do you think the

22:56

job of a forward-thinking, excellent finance

22:58

leader looks like five years from

23:00

now, pick the time frame, versus

23:02

today? How do they spend their

23:04

time today? How do you think

23:06

they will in the case that

23:08

they incorporate ramp, they lean into

23:10

AI, and they sort of maximize

23:12

what this can do? Well, I

23:14

mean, first, I think it's incumbent

23:16

on anyone, whether it's a finance

23:18

leader or someone building tools, an

23:20

engineer or a designer, like, I

23:22

think people should be thinking about,

23:24

you know, really automating all the

23:26

parts of the job. Maybe you

23:28

don't like or maybe that are

23:30

low value. Because I think there's

23:32

a whole class of problems in

23:34

acts in doing work that actually

23:36

can't. automated now. And I would,

23:38

I think this is something like

23:40

I've always internalized and appreciated from

23:42

our conversations ready of like I

23:44

think that great leaders are able

23:46

to not just like create focus

23:48

and urgency but identify like where

23:50

there's outsized returns to it to

23:52

our times. And I think what's

23:54

really true in a lot of

23:56

finance organizations is yes, maybe the

23:58

CFO has had the time and

24:00

focus on where is their value,

24:02

but you look at actually the

24:04

calendars and what's happening in a

24:06

lot of the rest of the

24:08

team. It's a lot of repetition.

24:10

Okay, it's the month is closed.

24:12

We're going to spend the first

24:14

six days, you know, it's eight

24:17

in bad cases, 15, right, tagging

24:19

transactions, downloading spreadsheets, matching things into

24:21

it. And that closes it's done.

24:23

And finally, just in the last

24:25

five days of the month, you're

24:27

able to do the real work,

24:29

the reason you got into that.

24:31

I think people should be thinking

24:33

now about how do I really

24:35

take those tasks which are wrote

24:37

intensive and turn that process into

24:39

a more automated process that I'd

24:41

be thinking about. I think if

24:43

you're doing this right. I really

24:45

do believe, you know, in five

24:47

years from now, you know, these

24:49

work streams will tend to be

24:51

more strategic, more insightful, more around

24:53

where's their creation, you know, value

24:55

created in a business, and having

24:57

people really obsess over that even

24:59

more entirely. And I think to

25:01

get there, it's about how do

25:03

you automate the process is how

25:05

do you design a more efficient

25:07

system in the interim. Eric, what

25:09

do you think happens when we

25:11

have a more efficient system? Like,

25:13

is everyone just out on the

25:15

golf course or are we going

25:17

to find new ways to work

25:19

hard? It's, look, for me, I

25:21

happen to think a lot of

25:23

purpose in life is creation. I

25:25

think people build tools. I think,

25:27

you know, want to move things

25:29

forward. And so, look, don't get

25:31

me wrong. I'm sure people find

25:33

more time for leisure, but, you

25:35

know, I think of it. There

25:37

was a really interesting, there was

25:39

making the rounds a few months

25:41

ago. think it was a set

25:43

of statistics that the number of

25:45

bookkeepers, there was a crisis in

25:47

the US, how the number of

25:49

bookkeepers had dropped by, I think

25:51

over the last 10 to 20

25:53

years, like a million less bookkeepers

25:55

were employed. And she was saying,

25:57

what was happening to all the

25:59

bookkeepers? And it turned out if

26:01

you looked at job descriptions for

26:03

financial analysts, strategist CFOs that had

26:05

grown by almost a million. And

26:07

functionally, there really wasn't much of

26:09

a change, but people were doing

26:11

different things. Rather than tagging, tabulating,

26:13

doing low-value tasks, people I think

26:15

had moved to a higher level

26:17

abstraction in doing more valuable work

26:19

for a business. And I think,

26:21

I actually think a lot of

26:24

that will happen. I think there

26:26

are certain levels of work that

26:28

are uniquely human that are uniquely

26:30

high value. And frankly, too, I

26:32

think in many cases much more

26:34

fulfilling. And I think that for

26:36

those who are for leaning, I

26:38

think I think there's going to

26:40

be much more of that over

26:42

time. And so, you know, that's

26:44

how I think about things for

26:46

sure. I mean, I think there's

26:48

also going to be strange things

26:50

too, or there's certain creative work

26:52

that computers in some cases will

26:54

do better. But, you know, I

26:56

do think that part of what

26:58

gives people purpose, or at least

27:00

for me excitement, like, has to

27:02

do with, you know, creation. And

27:04

I think that's always going to

27:06

be a very human thing. I

27:08

mean, that's as good of a

27:10

segue as you can imagine for

27:12

my next question, which is, Eric,

27:14

you are a founder through and

27:16

through, and you are someone that

27:18

a lot of other founders probably

27:20

look up to. What advice do

27:22

you have for folks, for creators,

27:24

for makers, for builders, given the

27:26

moment we're in, you know, how

27:28

do you meet it, what should

27:30

you go and build, how do

27:32

you approach this if you're someone

27:34

who wants to be a founder

27:36

or a builder? Well, first, I

27:38

mean, like part of being a

27:40

founder and a builder, I think

27:42

is just about like running this

27:44

very long. and continuous race. I

27:46

think that great companies are built

27:48

over many years and decades. And

27:50

I hope Bram is the last

27:52

company that I ever work on.

27:54

I want to be working on

27:56

it for a long time. And

27:58

I think there's always these questions

28:00

of what's changing in the world

28:02

and how is that going to

28:04

reset certain industry. I think there's

28:06

a lot of opportunities and we

28:08

can talk about like the place

28:10

to be spending time. But I

28:12

actually think when you look at

28:14

ramp and part of what's made

28:16

it work is really been starting

28:18

of what are the timeless truths

28:20

that are not going to change,

28:22

whether it's now or 10 years

28:24

from now or 100 years from

28:26

now. I can't imagine, you know,

28:29

that 100 years from now people

28:31

would say, you know, like I

28:33

wish to paraphrase Jeff phrase this,

28:35

like I just wish you would

28:37

have raised prices on on on

28:39

us, Amazon. Or I wish you

28:41

would deliver these goods a little

28:43

bit more slowly. I think this

28:45

is very much the case for

28:47

ramp. I think people want to,

28:49

you know, no matter what they're

28:51

creating, if you can create great

28:53

work with less effort, less time,

28:55

fewer dollars I think that is

28:57

always going to be in style.

28:59

And so I would say, I

29:01

would start first with being curious

29:03

about people's problems in the timeless.

29:05

who are real customers that could

29:07

serve, what are real businesses, and

29:09

what are actual problems that they

29:11

have now, and what are these

29:13

problems that are not gonna go

29:15

away? And then I think you

29:17

start to discover and uncover new

29:19

technological shifts that can help you

29:21

solve this in a new and

29:23

unique, or in some cases, you

29:25

know, very disruptive way. And so

29:27

I would say, like, you know,

29:29

focus on the timelists would be,

29:31

you know, my top advice of

29:33

this. Yeah, a friend of mine

29:35

has this great quote, which is,

29:37

he's like, we try to be

29:39

timeless rather than timely. Because the

29:41

time Lee, it just, you know,

29:43

it evaporates and it's ephemeral, where,

29:45

and I think that the way

29:47

you all are building ramp, you

29:49

know, certainly fits with that. Yeah,

29:51

thank you. Okay, we're going to

29:53

close that with some rapid fire

29:55

questions. for starters, what is your

29:57

favorite AI app? Oh, can't say

29:59

ramp. Oh my gosh. Well, to

30:01

be honest with you, I've been

30:03

really interested, just from like a

30:05

UX perspective, and just like how

30:07

it's bent our thinking, cognition labs

30:09

with Devin. I mean, really what

30:11

they're trying to do is build

30:13

an AI engineer, and I know

30:15

they're working hard at that. But

30:17

they took this agentic use case

30:19

and had a few poor innovations.

30:21

They realized if you were going

30:23

to hire an AI engineer to

30:25

do work on your behalf, well,

30:27

you would want it to have

30:29

access to the tools that an

30:31

engineer would have, and you would

30:34

want to be able to understand

30:36

what it's doing. And so rather

30:38

than just as a prompt and

30:40

you see what it does, Devin

30:42

has a notebook, a planner, just

30:44

like any engineer and thinking through

30:46

what is it going to do,

30:48

it has a browser for things

30:50

or to check stack overflow when

30:52

it gets confused. It has shell

30:54

access, you know, and it's connected

30:56

to your tools. And so I

30:58

think what's most fascinating about it

31:00

is as it does things, you

31:02

can watch what Devin is doing,

31:04

what it gets right, and the

31:06

mistakes that it makes. And so

31:08

my favorite app right now is

31:10

that because it's taught me a

31:12

new way to look and think

31:14

about design and how these tools

31:16

may feel over the coming years.

31:18

All right, so over the next

31:20

next part of the lightning round,

31:22

over the next 5-10 years, other

31:24

than finance, what other industries do

31:26

you think are going to change

31:28

the most? And by the way,

31:30

I really like the self-driving money

31:32

or self-driving finance term that you

31:34

had earlier. That was, I hadn't

31:36

heard that before. It's a what

31:38

else is going to go down

31:40

driving. Yeah, I hope people make

31:42

this come to life. Look, for

31:44

me, like I really hope health

31:46

care. And I really believe it

31:48

will be. I mean, I think

31:50

it's already changing radically how diagnosis

31:52

is done. You know, I think

31:54

a lot of whether sickness or

31:56

wellness is taking large sets of

31:58

data. not just your annual checkup,

32:00

but if you can have continuous

32:02

measurement over many years and decades,

32:04

that goes a long way. I

32:06

hope it does too. I think,

32:08

you know, for many doctors, you

32:10

know, it's a glorified note-taking job.

32:12

There's very little in the way

32:14

of diagnosis. listening to patients and

32:16

having time to spend. And so

32:18

I actually think, you know, that's

32:20

one that can actually return some

32:22

of the humanity to the care

32:24

of it. So I'm very hopeful

32:26

there for sure. I think very

32:28

obviously design and creation. I think

32:30

it's no longer about what can

32:32

you make, and do you understand

32:34

the tools in order to create,

32:36

but can you make something that's

32:39

fundamentally interesting? That's intuitive. And so

32:41

I actually think that becomes very

32:43

interesting too. So those would be

32:45

kind of the health care and

32:47

broader design ecosystems would be probably

32:49

for me ones I think about.

32:51

Now on the health care side

32:53

I totally agree with you Eric

32:55

for what it's worth because I

32:57

think you think about how crazy

32:59

does that companies have you know

33:01

dashboards that we look at every

33:03

hour you know for leading indicators

33:05

on what's going to happen and

33:07

for our health we go to

33:09

an annual checkup that maybe we

33:11

don't even go to annually can

33:13

you imagine if like all you

33:15

looked at was your company's metrics

33:17

once a year just to see

33:19

well I don't know how to

33:21

go okay I guess everything's okay

33:23

I mean that we don't with

33:25

the idea of it being, you

33:27

know, know before you know, right,

33:29

the leading indicator side, is crazy.

33:31

And so I agree and share

33:33

your optimism and hopefully there's more

33:35

than just glucose monitors that we

33:37

can have. I hope so. I

33:39

mean, like if you just like

33:41

totally like if you could only

33:43

get a look at how your

33:45

business is doing, you know, every

33:47

year or every six months or

33:49

it's going really badly and we're

33:51

trying to figure out like it's,

33:53

it's, it's, it's a mess like

33:55

you want to find this stuff

33:57

early and it's, it's possible. consumer

33:59

and the B to B side,

34:01

who else in financial services is

34:03

doing interesting things with genitive AI?

34:05

Or like what are the big

34:07

ideas in like kind of the

34:09

intersection of finance and AI that

34:11

you wish you had time to

34:13

explore and build? So I think

34:15

there's a couple maybe meta points

34:17

that I would get at. I

34:19

mean first, One,

34:22

I think just financial service is

34:24

a long, long way to go,

34:26

right? Like, like I think about

34:28

our competitors, folks like American Express

34:30

and Chase and City and all

34:32

led by wonderful people, but their

34:34

founders wore top hats, right? Like

34:37

they've been around for a long

34:39

time and as the world went

34:41

from like no phones to flip

34:43

phones to iPhones, a lot of

34:45

these things never really changed. And

34:47

there was a class of businesses

34:49

that if you were a bank,

34:52

you were allowed to move money

34:54

and store money. And if you

34:56

weren't. Well, you're part of the

34:58

rest of the world. And I

35:00

think it's been a big part

35:02

of why it's moved so slowly.

35:04

And what I would say is,

35:07

I think it's really wonderful that

35:09

now it's not just banks that

35:11

can do this, because I think

35:13

you can finally introduce great technology

35:15

in companies like RAMP, or yes,

35:17

we're a FinTech, but I think

35:19

we're a productivity company. I think

35:22

we're a company that's interested in

35:24

how does time intersect with the

35:26

movement of mine. And so whether

35:28

it's directly in financial services or

35:30

if it's disruptors from adjacent who

35:32

are going to encroach. I think

35:34

it's about time for the whole

35:37

sector to get even optimized. And

35:39

so I know they wouldn't call

35:41

themselves this and they probably would

35:43

blush and say, no, we're not

35:45

as a FinTech company. But I

35:47

think Apple right now is truly

35:49

a FinTech innovator because they're connecting

35:52

the movement of money to. identification,

35:54

you know, of who people are,

35:56

you know, if they said he

35:58

in order to create less fraud

36:00

in the system, reducing friction dramatically,

36:02

whether it's offline or in person,

36:04

and I think are increasingly be

36:07

able to connect to outcomes of

36:09

times. And so there's a variety

36:11

of folks, but I actually think

36:13

a lot of disruptors won't be

36:15

from traditional finance companies would be

36:17

not my long-witted answer to you,

36:19

Sonia. Sorry, you're giving you too

36:22

much time. You're just letting me

36:24

run on. No, that's awesome. All

36:26

right. Last question, Eric. Who do

36:28

you admire most in AI? And

36:30

you're not allowed to say, Sonia.

36:32

It's Robbie, it's gotta be, it

36:34

has to be, it has to

36:37

be, you're Robbie. It's, of course,

36:39

it's, it's, it's been a great

36:41

podcast, we, thanks for your time.

36:43

It's, no, present company excluded, yeah.

36:45

It's, um. Look, I just think

36:47

what Satya Nadella has done for

36:49

the creation of the category and

36:52

how he's, you know, he's partnered

36:54

with, I think, incredible, the incredible

36:56

team at Open AI, you know,

36:58

to go from what was primarily

37:00

research to in production and in,

37:02

you know, I think in almost

37:04

every aspect of computing today, I

37:07

think has been incredible. You know,

37:09

he's been an incredible. mentor to

37:11

me and the ramp team and

37:13

thinking about how can a genetic

37:15

AI be designed over time, I

37:17

think is thinking, you know, in

37:19

much larger cycles, you know, not

37:22

just, you know, weeks and years,

37:24

but truly decades. And, you know,

37:26

last, I think he's, person who's

37:28

been through a lot, and I

37:30

think has a lot of sincerity

37:32

and kindness in how he operates.

37:34

You know, I haven't read it.

37:37

I think his book, Hif Refresh,

37:39

is excellent. I think the way

37:41

that he'd been able to partner

37:43

deeply with all kinds of companies

37:45

is pretty inspiring. And so, present

37:47

company excluded. I think Saginadella would be

37:50

my tech. That's a fantastic answer. And

37:52

I, he is a very admirable person.

37:54

Eric, thank you. This was awesome. We

37:56

had a great time. We learned a

37:59

lot. learn expression in

38:01

self -driving money, self

38:03

-driving finance, and I think I

38:05

all seriousness, this is the

38:07

combination the combination of innovation well

38:09

as applicability. is very

38:11

unique. very unique so we are very

38:13

very very happy that we get to

38:15

be partnered with you guys guys.