0:06

to Startupread.io your podcast

0:08

and YouTube blog covering

0:10

the German startup scene

0:12

with news interviews and

0:15

live events. Hello and

0:17

welcome everybody this is

0:19

Joe From Celebrate.0 your

0:21

starter podcast YouTube blog and

0:23

internet radio station from Germany.

0:26

Today I bring you the

0:28

second interview with Kasten the

0:31

current business angel of the

0:33

year for Germany. Yes, I'm

0:35

fine here in Zanzibar at

0:37

the moment. I'm on holiday

0:39

still. Yes, we do have

0:42

a little game going on

0:44

if you break into sweats

0:46

before the end of the

0:48

interview because you had to

0:50

turn off the air condition

0:53

for this interview. So it'll

0:55

be interesting. We may tell

0:57

our audience that you're the

0:59

business angel of the year

1:01

selected by band, the German

1:04

Business Angel Association. And they

1:06

usually elected, I do believe,

1:08

in October and November, and

1:10

the special thing about

1:13

it is that they are

1:15

awarded on the

1:17

recommendation of the startups

1:19

you invest in. Yes,

1:21

and also the Minister

1:24

of Economy of Germany

1:26

is head of the jury.

1:28

Ah, I see. Which will

1:30

be subject to change soon?

1:33

Yes. Okay. I see. We

1:35

already talked about in the

1:37

other interview that your serial

1:39

entrepreneur yourself, that you currently

1:41

run Casablanca.a.i, which everybody who's

1:44

watching this can see how

1:46

this is working here. And

1:48

I want to get straight

1:50

into your investments philosophy and

1:52

start-up selection because you told

1:55

me when we talked about

1:57

this, you get up to

1:59

10 pictures a week. How

2:01

do you cut through the

2:03

noise and identify opportunities? Yeah,

2:05

that's a very good question

2:07

and it's not so easy

2:09

because of course you always

2:11

fear missing the one hidden

2:13

gem that becomes the next

2:16

Facebook or Microsoft or open

2:18

AI or whatever. And so

2:20

what do we do? We

2:22

have own AI that filters

2:24

through the pitch decks. takes

2:26

out some red flags. So

2:28

I at least annotate some

2:30

red flags so that we

2:32

easily identify this without going

2:34

through the 45 pages of

2:36

some pitch decks. So there

2:38

are some things in there

2:41

like nobody has had any

2:43

experience in sales before or

2:45

they are emphasizing technology but

2:47

they don't have patterns and

2:49

It seems that it's not

2:51

patentable. So there are some

2:53

things where we say it's

2:55

probably not interesting to take

2:57

any deeper look at the

2:59

startup. And so this is

3:01

just the first filter. It

3:03

tells about half of the

3:05

teams that probably they're not

3:08

so interesting. So depending on

3:10

the time my people have,

3:12

so I have one and

3:14

a half people on the

3:16

investment side of Sun and

3:18

Kistof. And so they take

3:20

a deeper look only at

3:22

those which were not red

3:24

flagged by the AI. So

3:26

the AI just principally takes

3:28

the pitch deck and goes

3:30

through some analysis, also puts

3:32

a lot of information into

3:35

a kind of a database

3:37

so that we can later

3:39

on analyze all the startups.

3:41

And so if we miss

3:43

something, which turns out to

3:45

be very good. we learn

3:47

for the next time. So

3:49

after two years or so

3:51

when this data turns out

3:53

to be a high flyer.

3:55

I see. I'm wondering since

3:57

when do you start utilizing

4:00

AI in this way? Summer

4:02

last year I think. So

4:04

we started programming something in

4:06

I think April or so

4:08

and work then in summer

4:10

and we want we want

4:12

to do more but right

4:14

now I have more other

4:16

projects so I didn't program

4:18

it but I gave the

4:20

the idea what to do

4:22

to somebody in my in

4:24

my team and not in

4:27

the investment team somewhere else

4:29

in my company somewhere else

4:31

in my company. in my

4:33

company group and they just

4:35

program this. I see. It's

4:37

mainly an LEM that does

4:39

the analysis but you have

4:41

to do some things around

4:43

so to convert the PDF

4:45

to set up the database

4:47

and those things. You focus

4:49

more on a startups technology

4:51

than its founding team. Why?

4:54

Have you ever been wrong

4:56

with this approach? I personally

4:58

believe a lot of people

5:00

are saying they just focusing

5:02

on the team because they

5:04

want to see consultancy A,

5:06

B or C or university

5:08

C, D and E. That's

5:10

basically when they get interested

5:12

in invest or worked at

5:14

a startup A to set.

5:16

Right. So when they focus

5:19

on technology, it normally means

5:21

that they look if one

5:23

of the founders was working

5:25

at one of the... Batman

5:27

is I think the new

5:29

thing, so the magnificent age.

5:31

And if one of them

5:33

work as a software developer

5:35

there, then probably the problem

5:37

is technically good, or the

5:39

start of the startup is

5:41

technically well set up. That's

5:43

what most investors estimate. The

5:46

thing is that most investors

5:48

come from a financial background

5:50

and not from a tech

5:52

background. So it's much easier

5:54

for them to analyze. financials

5:56

or teams and take a

5:58

look at the founders rather

6:00

than really telling if this

6:02

technology is again changer is

6:04

really something that is really

6:06

ahead of the competition or

6:08

if it's just another startup

6:10

doing something with a very

6:13

good idea but not with

6:15

great tech in the background

6:17

and I'm one of the

6:19

few ones who are coming

6:21

from tech and can thus

6:23

tell something about this part

6:25

so I think this is

6:27

probably one of my my

6:29

USPs here that I can

6:31

take a deeper look into

6:33

technology. Of course my team

6:35

first analyzes just these standard

6:38

questions like, is the team

6:40

good, is the plan good,

6:42

is the market big enough?

6:44

So generally we say we

6:46

invest in a total addressable

6:48

market over one billion, so

6:50

if it's only a hundred

6:52

million, normally we just don't

6:54

take a closer look. That's

6:56

one of the things that

6:58

our AI also tries to

7:00

find out. And that's something

7:02

that the AI and the

7:05

team already challenge. And then

7:07

when I talk with the

7:09

founders, I'm also talking about

7:11

technology. Of course, if the

7:13

founding team would be like

7:15

not somebody I would like

7:17

to work with, then we

7:19

would also not invest. And

7:21

of course, technology is not

7:23

everything, but then it's not

7:25

that you have to get

7:27

every good deal. just you

7:30

have to make sure that

7:32

many of the deals you

7:34

get are good. But yeah,

7:36

so you will always miss

7:38

out on some deals. So

7:40

I had an opportunity to

7:42

invest very very early in

7:44

a startup started by two

7:46

pupils. They were below 18

7:48

years old and they were

7:50

starting the company when they

7:52

became 18, but they were

7:54

preparing for it. And I

7:57

said you need two pivots,

7:59

one technically. And the idea

8:01

is great, but technically and

8:03

also from the business model,

8:05

both will not work. How

8:07

did they react given that

8:09

they're not very old? No,

8:11

the thing is I said

8:13

I'm still investing. So, but

8:15

on the, that was, they

8:17

offered me a low variation

8:19

basis and I said, I'm

8:21

going through this with you

8:24

and I think you will

8:26

do fine because the technical

8:28

guy was very open to

8:30

learning and he was already.

8:32

quite great but just some

8:34

things you wanted to do

8:36

would not work and from

8:38

the business model it would

8:40

also not work I calculated

8:42

this in Excel but the

8:44

CEO also had a very

8:46

he was very far for

8:49

a 70 year old so

8:51

I thought I invest and

8:53

then the problem was that

8:55

we had an appointment the

8:57

notary and that was in

8:59

Stuttgart and I had an

9:01

appointment in the evening with

9:03

a politician in Berlin, a

9:05

member of the parliament. And

9:07

the notary called and said,

9:09

we need to shift this

9:11

by one hour, because my

9:13

previous case takes longer. And

9:16

I said, then I cannot

9:18

get my flight and I

9:20

cannot go on the meeting

9:22

with a politician. I will

9:24

have to cancel today, let's

9:26

find another apartment next week

9:28

and we didn't find it

9:30

next week, we didn't find

9:32

it the week after, and

9:34

after about four weeks they

9:36

came back to me, we

9:38

have a new, we had

9:40

an appointment in four weeks

9:43

or so, but they said

9:45

now our valuation is more

9:47

than twice what we had

9:49

before, because we have found

9:51

some other investors who want

9:53

to invest. And then I

9:55

said for two pivots, I'm

9:57

not going to do it

9:59

at that valuation. And we

10:01

still discussed a bit but

10:03

did not agree. And what

10:05

happened is they started this

10:08

and they did the first

10:10

period very soon. And then

10:12

they got, I think, 2

10:14

million on a 10 million

10:16

basis. after less than six

10:18

months. And that was quite

10:20

a lot already because we

10:22

were far lower in our

10:24

valuation before. And I'm not

10:26

so sure how much their

10:28

finance, I think they have

10:30

taken up something like 20

10:32

million already also and are

10:35

doing very well. So I

10:37

did wrong there because I

10:39

thought with the two pivots

10:41

necessary, the new valuation is

10:43

not appropriate because the risk.

10:45

is still high, too pivot

10:47

necessary, and the valuation is

10:49

not ultra-low as it was

10:51

before, just not very high

10:53

yet. And I was definitely

10:55

wrong there. The other thing

10:57

is, yes, I sometimes make

10:59

mistakes in investing, of course,

11:02

and that happens to everyone,

11:04

and there can be two

11:06

sides of mistakes. One is

11:08

you have missed the chance

11:10

and the other thing is

11:12

you invest in a company

11:14

where it doesn't pay off

11:16

and the company does not

11:18

do well. And I also

11:20

had this in the past.

11:22

I invested in a company

11:24

that just didn't want to

11:27

grow. They were happy after

11:29

they were slightly profitable and

11:31

still at a very low

11:33

level. And yes, we can

11:35

run the company now. We

11:37

don't need any more investors.

11:39

So what's the problem? And

11:41

that... that's also not so

11:43

good. So even if the

11:45

company doesn't go broke, it

11:47

might still not by far

11:49

not pay off for the

11:51

investor. Talking about investor approach

11:54

here, what is for you

11:56

a hard no or a

11:58

rack flag when you evaluating

12:00

the start-up pitch? So one

12:02

is if I can't work

12:04

with the company, that's a

12:06

hard no. Also if... Sometimes

12:08

they think they have invented

12:10

something totally new and I

12:12

find out that that what

12:14

the company invented is already

12:16

there or very easy to

12:19

reproduce and you can't get

12:21

a patent for it. So

12:23

that's normally a red flag

12:25

for me. So for example

12:27

I had somebody with an

12:29

invention in chemistry and they

12:31

would not tell me what

12:33

it actually was so actually

12:35

they had a discovery in

12:37

chemistry that that's what they

12:39

told me. And when they

12:41

finally told me after I

12:43

signed a hard NDA and

12:46

lots of things I found

12:48

out in less than five

12:50

minutes that the effect was

12:52

already known, but under a

12:54

different name with a slightly

12:56

different approach, but it didn't

12:58

make any difference. So it

13:00

was definitely existing. So that's

13:02

very hard no. So what

13:04

else is a hard no?

13:06

So I'm not so much

13:08

on this single founder doesn't

13:10

work. So a single founder

13:13

can work well. I don't

13:15

have a big problem with

13:17

that. I know it for

13:19

myself, so I started. multiple

13:21

companies normally as a single

13:23

founder in the beginning and

13:25

then took up some people

13:27

later but a fact find

13:29

that I actually grew to

13:31

a hundred people without any

13:33

co-founder. So and then I

13:35

hired a CEO in order

13:38

to just get me a

13:40

bit away from management and

13:42

let me go back to

13:44

innovation again. Because I don't

13:46

like management so and this

13:48

this this guy really love

13:50

management. He was in a

13:52

10 times larger company before

13:54

a C level and joined

13:56

us and just restructured, reorganized

13:58

how we worked in terms

14:00

of how we managed the

14:02

company. When we talked before

14:05

you told me you have

14:07

currently a portfolio of 15

14:09

investments. First, my personal understanding

14:11

is that this is usually

14:13

a big portfolio. for German

14:15

business angels, right? Yeah, for

14:17

business angels, yes. So there

14:19

are some people, I have

14:21

found it also. the frequent

14:23

angel circle which is a

14:25

sub circle or subgroup in

14:27

band in the German business

14:29

internet with Deutschland and we

14:32

accept people who have done

14:34

10 invests in their past

14:36

so you don't need to

14:38

hold 10 invests now but

14:40

you only if you have

14:42

only one or two or

14:44

so right now and you

14:46

have already exited or some

14:48

have... gone broke or so

14:50

and you have done 10

14:52

investments in the past you

14:54

can join. Very very few

14:57

members of fund have had

14:59

10 investments. So I don't

15:01

know how many there are

15:03

in the business and effort

15:05

Georgetown to could qualify but

15:07

it's probably 100 or not

15:09

so much more and 34

15:11

are in our frequent engine

15:13

circuit. Actually the one with

15:15

the most investments has done

15:17

more than 100 investments. We

15:19

already have one with 75

15:21

who was also nominated for

15:24

Business Engine of the Year.

15:26

And we have one more

15:28

with more than 50 investments.

15:30

And all of those investments,

15:32

so we're talking about investments

15:34

of 50K euros and up.

15:36

We're not talking about people

15:38

who are just buying a

15:40

business engine club or so

15:42

diversify their money and invest

15:44

10,000 euros each in 50

15:46

startups. So, and we have

15:48

one who has done also

15:51

about 20 investments, but his

15:53

typical investment is 1 million.

15:55

So, he doesn't start with

15:57

1 million, he normally starts

15:59

with 250, 250K, but over

16:01

time he grows in the

16:03

companies, and sometimes he also

16:05

starts with a million. So,

16:07

there are these people. So

16:09

in the freaking angel circle

16:11

I'm rather one of the

16:13

smaller investors. But as compared

16:16

to the typical business angel,

16:18

I would say yes this

16:20

is quite a portfolio. So

16:22

I've done 21 investments, 15

16:24

are still active. Still active,

16:26

exactly. And I was wondering,

16:28

which one are you most

16:30

excited about right now and

16:32

why? There's actually quite a

16:34

number of really interesting investments.

16:36

So I'm very positive for

16:38

Wayland. That's the London investment

16:40

I made. So outside Germany

16:43

in the UK. Wayland. They

16:45

detect social media attacks. So,

16:47

you know, sometimes the company

16:49

has negative social media and

16:51

that's okay if it's natural.

16:53

But sometimes a competitor actively

16:55

attacks a company on social

16:57

media. It also happens in

16:59

politics. And Bayland has a

17:01

number of customers who are

17:03

big and who are big

17:05

and who are... who have

17:08

been under, who were under

17:10

social media attacks from competitors

17:12

or political opposing people. So

17:14

that means normally there is

17:16

some truth in the story,

17:18

so something has gone wrong

17:20

with something, so like disappointed

17:22

one customer or so, and

17:24

an attack normally means that

17:26

somebody takes this story and...

17:28

makes it huge as opposed

17:30

to the positive stories. And

17:32

this happens. This is not

17:35

just something that could happen,

17:37

but it does happen. There

17:39

are malevolent people who actively

17:41

attack their competitors in social

17:43

media. And Bayland detects this

17:45

with AI, and they're very

17:47

strong AI team. I can

17:49

tell this. Actually, when one

17:51

startup starts to attack another

17:53

one on social media, I

17:55

would be wondering, don't you

17:57

have anything else to do?

18:00

Yeah, so there are a number

18:03

of countermeasures you can do. And

18:05

that's a project that Beylant then

18:07

can do. That cannot be done

18:10

automatically yet. So some things are

18:12

legal, so you can, if you

18:15

find out, the first thing is

18:17

you try to find out who

18:19

is behind it, but if they're

18:22

really good, they somehow hide it

18:24

behind a Russian group or something.

18:26

But sometimes they can find out

18:29

who is attacking or who is

18:31

behind it. And then there are

18:34

other measures like how do you

18:36

work with the press? Because normally

18:38

then the press takes up this

18:41

and you have to do some

18:43

countermeasures there. And how do you

18:45

deal with the social media and

18:48

so on? I'm not an expert

18:50

for that. So you have to

18:52

interview Valen the next time in

18:55

order to talk about this topic.

18:57

But I think it's a great

19:00

company and a former investigative BBC

19:02

journalist as part of the founding

19:04

team. But they're also really great

19:07

in the eye. So I admire

19:09

what they do. So that's one

19:11

of the startups. I can tell

19:14

about more if you want to

19:16

talk about more startups. I do

19:19

have still a lot of questions

19:21

for you and we are already

19:23

running 20 minutes of recording. Yeah,

19:26

and keep in mind that AC

19:28

is not running right now. I

19:30

was wondering what one startup investment.

19:33

Can you name one that completely

19:35

surprised you with how fast they

19:37

scaled? I have one where I

19:40

was wondering how I was astonished

19:42

how quickly everything went well. I

19:45

have one startup, I cannot name

19:47

it if I tell more about

19:49

it because we are under not

19:52

too hard but still under the

19:54

NDA with a buyer. So, but

19:56

I invested in a hardware company.

19:59

which had a lot of innovation

20:01

and I was fascinated by the

20:04

innovation. and the hardware company was,

20:06

they showed me their very first

20:08

prototype of one part of the

20:11

hardware they were building. And so

20:13

what happened was they took in

20:15

investors, they tried to... They tried

20:18

to bootstrap and then found out

20:20

that it will not be sufficient.

20:22

They had some money but not

20:25

enough. And then they took in

20:27

investors and I was one of

20:30

the people that they asked because

20:32

of AI because they thought AI

20:34

will play a big role also

20:37

in their future product. And so

20:39

they had the very first prototype

20:41

of something where you need hundreds

20:44

of thousands of big companies. and

20:46

it did not work standalone, you

20:49

needed some other components also and

20:51

they just had this one thing.

20:53

But I was fascinated by the

20:56

height of the innovation they built.

20:58

Then it took only 18 months

21:00

until they sold to a big

21:03

company, they were still like, I

21:05

don't know, minimum one and a

21:07

half years pre-seeds, appropriate revenue. So

21:10

definitely what they had built then

21:12

was a fully working prototype, but

21:15

in a very small scale. So

21:17

like when you say I'm automating,

21:19

let's say a car manufacturing factory,

21:22

and you just demonstrate one robot

21:24

which screws in the screws or

21:26

something at one place. So they

21:29

could show a smaller part of

21:31

everything. But you could see that

21:34

you could see a smaller part

21:36

of everything. But you could see

21:38

that you could see that this

21:41

is a going to change the

21:43

world in this special aspect. And

21:45

then they already sold. So that

21:48

went very fast and it was

21:50

still far beyond a bit before

21:52

they could sell the product to

21:55

any customer. And we thought that

21:57

we have in the current situation,

22:00

we thought that we have to

22:02

have the first customer and we

22:04

tried to get some some customers

22:07

who sign LOI or so, which

22:09

was also very difficult in the

22:11

current situation, but actually they had

22:14

some inquiries from big companies who

22:16

wanted to invest in them or

22:19

rather buy them. And we went

22:21

fast with one of them, and

22:23

so after only 18 months, I

22:26

could sell my shares again. and

22:28

was very happy with the with

22:30

the the mouse multiple I got.

22:33

You learned a lot when you

22:35

do so many startup investments. I

22:37

was wondering what was one of

22:40

the counterintuitive lessons you learned? You

22:42

said you're already you're still open

22:45

to invest in single founder companies

22:47

and not teams. What else did

22:49

you learn? Oh,

22:52

that's a difficult question. Maybe

22:54

you have to skip that

22:56

if something was counter-intuitive and

22:58

I still had a gut

23:00

feeling, you mean, kind of,

23:02

right? So, so it's not

23:05

like, we normally, normally you

23:07

do this, but in this

23:09

case, I'm going for it.

23:11

So, but nothing comes to

23:13

my mind at this moment,

23:15

sorry, so for counterintuitive treatment.

23:17

I want to draw the

23:19

investment. So I mean normally

23:22

what I said about this,

23:24

so what I said about

23:26

this hardware company right now

23:28

was also normally when you

23:30

invest in hardware you think

23:32

it takes 10 years until

23:34

we have an exit. And

23:36

in this case, it was

23:39

not necessarily because it was

23:41

so clear that this is

23:43

very disruptive for the industry

23:45

and they were so far

23:47

ahead of anybody in this

23:49

field. I see. When I

23:51

talk to many successful investors,

23:54

they do have personal investment

23:56

thesis or rules. Do you

23:58

follow some of them? you

24:00

have some and you follow

24:02

them before making a deal? So

24:04

we have we have a list

24:07

of a few things we check

24:09

but these are the things that

24:11

Roxanne checks when they when

24:13

she talks or some the

24:16

AI checks some Roxanne checks

24:18

and so the the general

24:20

thing is that we are

24:22

always ready to to learn

24:24

and maybe still invest if

24:26

something is not according

24:28

to to our list, but

24:30

if something is, some other

24:33

parts are overwhelmingly good. So,

24:35

but, so in general,

24:37

for example, we invest

24:39

precedes. So that means if

24:42

the company is already running

24:44

and just wants to grow,

24:46

normally we don't invest. Still,

24:49

I did this once. So

24:51

I invested in a company

24:53

called Hybrid Lydar. And

24:55

so they're not really... running

24:58

but they have first customers

25:00

who have done tests, runs

25:03

with them and paid for

25:05

that and so on. But

25:07

this is already quite a

25:09

big company. But I thought

25:12

this innovation is also very

25:14

high so I invested also.

25:16

I do a new type

25:18

of LIDAR of the distance

25:21

measuring which is used

25:23

for. self-driving cars except by

25:25

Tesla who is only using

25:28

cameras. But it's also a

25:30

great application for manufacturing

25:32

and robots and everything

25:34

and they have a much higher

25:37

resolution at a less costly lighter

25:39

device and they're currently

25:41

going to the first prototype.

25:43

So not prototype, the first

25:46

production run. So for the first

25:48

production run. So for the first

25:50

production run. So for the first

25:52

That's very important for that right

25:54

now. But they're already quite big,

25:56

so normally I don't invest on

25:58

valuations over 5 million. Every

26:00

investor is currently getting giddy

26:02

when they hear AI which

26:04

was in the past something

26:07

like blockchain quick delivery and

26:09

so on and forth. Except

26:11

for those current technologies, if

26:13

you could only invest in

26:15

one emerging technologies in the

26:18

next five years, what would

26:20

it be? I would have

26:22

said deep reinforcement learning, so

26:24

that's one part of AI,

26:26

but it's AI that was

26:28

underestimated for a long time.

26:31

In my prediction, in the

26:33

beginning of 24, I said

26:35

deep RL, deep reinforcement learning

26:37

will do a lot in

26:39

the coming year in 24,

26:41

and actually it only did

26:44

now with deep seek. Everybody

26:46

could see that deep reinforcement

26:48

learning is still a very

26:50

important factor because this was

26:52

basically the foundation of why

26:55

they could be much faster

26:57

and much better than other

26:59

approaches in the past. So

27:01

it's a combination of this

27:03

transformer technology of LMs with

27:05

this deep reinforcement learning. So

27:08

there was some deep RL

27:10

also in the previous LMs

27:12

but it had only a

27:14

minor role and now it

27:16

became part of the... core

27:18

of the innovation of the

27:21

of deep seek. So that's

27:23

what I have would have

27:25

said then and right now

27:27

I think it's what is

27:29

totally underestimated is robots even

27:31

though they all already have

27:34

some hype but I think

27:36

people cannot see how big

27:38

the disruption will be in

27:40

five years. I actually talked

27:42

to Jan Lippert another interview

27:45

that we just published. He

27:47

is Frankfurt-born even though he

27:49

was raised in the US.

27:51

He is working currently on

27:53

a company that combines robotics

27:55

with large language models which

27:58

is also pretty interesting and

28:00

it's something I have. heard

28:02

a lot about before. So

28:04

definitely you need to combine

28:06

them if you want to

28:08

have practical applications in smaller

28:11

scale companies. So in bigger

28:13

scale companies when you need

28:15

like 200 robots doing the

28:17

same thing in a factory

28:19

or so you can train

28:21

the robots by other measures.

28:24

But if you want them

28:26

to just take over any

28:28

kind of small job that

28:30

helps you. you need to

28:32

also be able to give

28:35

them commands in that national

28:37

language like when you have

28:39

them let's say at the

28:41

building site and it's a

28:43

helper for the mason. So

28:45

then then you want to

28:48

tell them like just hand

28:50

me over five more of

28:52

those stones, hand me over

28:54

this tool and so on.

28:56

Can you just go downstairs

28:58

and carry up two more

29:01

sacks of the cement or

29:03

something? Yes. and you want

29:05

to command them in that

29:07

way, and you don't want

29:09

to show them every step,

29:11

which is the normal way

29:14

of learning for the new

29:16

robots, so they observe and

29:18

they copy, but also for

29:20

their internal reasoning, for understanding

29:22

what you're doing, I think

29:25

they're also using maybe not

29:27

large language models. But the

29:29

same kind of foundation models

29:31

with some transformers inside or

29:33

so do represent what the

29:35

person is doing. So that

29:38

can reason what is it

29:40

that is doing because you

29:42

don't go like, okay, now

29:44

the hand moves 20 centimeters

29:46

in that direction, then the

29:48

fingers grab. Rather, you want

29:51

to do something like he's

29:53

grabbing the tool, he's... taking

29:55

the tool, putting it to

29:57

the screw, turning the tool

29:59

so you wanted to understand

30:01

what is happening and for

30:04

that it already needs to

30:06

be some knowledge representation internally

30:08

and that's easier with language

30:10

as it seems than with

30:12

most other things. I was

30:15

smiling when he started to

30:17

kind of combine these robots

30:19

with learning with what came

30:21

to mind other apprenticeship models

30:23

because if you're an apprentice

30:25

in Germany they will play

30:28

some practical jokes on you

30:30

in the start. I also

30:32

was a victim of that

30:34

and I was wondering what

30:36

jokes the people will play

30:38

with robots. And robots, yes

30:41

certainly they will. People will

30:43

want to play tricks on

30:45

the robots and then love

30:47

them. Yeah. I see. I'm

30:49

getting a little bit more

30:51

on the birdside view. What

30:54

do you think is the

30:56

biggest advantage that European startup

30:58

founders have over Silicon Valley

31:00

entrepreneurs? So you mean the

31:02

founders in Germany or the

31:05

small investors in Germany? They're

31:07

disadvantages. I would rather go

31:09

for the founders because if

31:11

they do have advantages, also

31:13

the investors should invest in

31:15

them have advantages. So I

31:18

cannot tell exactly about the

31:20

US scenery, but what I

31:22

would say is that probably

31:24

the founders in the US

31:26

will get money more easily,

31:28

but then they have to

31:31

deliver more on KPIs more

31:33

quickly. And when you're in

31:35

Europe, probably founders who have

31:37

more of a longer time

31:39

until they will reach results,

31:41

probably have a bigger chance

31:44

of doing this with... German

31:46

investors and European investors, so

31:48

business engines especially, then in

31:50

the US I think, because

31:52

we are not so hard

31:55

results driven as most of

31:57

the US investors I've talked

31:59

to so far, but there

32:01

might be others also in

32:03

the US. So maybe some

32:05

more disruptive innovations which take

32:08

a bit longer can emerge

32:10

in Europe more easily because

32:12

The founders will not have

32:14

so much pressure to get

32:16

results quickly. I see. Before

32:18

we go into a very

32:21

small outbreak, I was wondering,

32:23

how do you manage to

32:25

run Casa Planka plus all

32:27

the companies, all the start

32:29

of investments you have? The

32:31

secret is people, so you

32:34

have to have the right

32:36

team, you have to have

32:38

the right people, and I'm

32:40

very happy about my PA

32:42

banker. Also, very happy about

32:45

the other people in the

32:47

team, who actually manage a

32:49

lot on their own, and

32:51

I don't have to say

32:53

okay to everything before they

32:55

do it. They just do

32:58

it and ask for forgiveness

33:00

rather than for permission before.

33:02

And normally everything's right. And

33:04

of course, in California, I

33:06

have a CEO who runs

33:08

the daily business and he's

33:11

also a... an experienced person

33:13

at my side and he's

33:15

running everything in sales. And

33:17

so it's Marcus Falma. He

33:19

also was a CEO of

33:21

some other German smaller companies

33:24

which were not startups. So

33:26

he knows how normal business

33:28

runs. And so I have

33:30

a very experienced person at

33:32

my side and he's running

33:35

everything in sales and normal

33:37

daily business in Canada. And

33:39

the same is true for

33:41

omicon data solutions. I have

33:43

Albert there, who is also

33:45

very experienced and a great

33:48

operator, Cetto. C.O. Okay. We'll

33:50

be back after a small

33:52

outbreak. 2024 current business angel

33:54

because the award is only

33:56

the The award ceremony takes

33:58

only place in late fall,

34:01

so he's the current business

34:03

angel. Talking about this winning

34:05

business angel of the year

34:07

2024 is a huge honor.

34:09

What does this award mean

34:11

to you? I was quite

34:14

astonished when I finally got

34:16

it. So in the beginning

34:18

I thought yes, I'm doing

34:20

very well and probably I

34:22

have a good chance. So

34:25

like 30% or so. And

34:27

then I saw who got

34:29

into the short list. and

34:31

out of I think 83

34:33

they selected seven one I

34:35

was one of them and

34:38

the others were so great

34:40

so one of them had

34:42

75 investments and so on

34:44

and then I thought my

34:46

chances are like below 10%

34:48

to get elected and when

34:51

they finally I didn't hear

34:53

anything before it was on

34:55

the event and when they

34:57

finally called my name I

34:59

was I was really so

35:01

shocked and astonished and and

35:04

and so It somehow resonated

35:06

with me so deeply, wow,

35:08

I got this with these

35:10

great people. So, yeah, so

35:12

I felt very, very, very

35:15

happy. And now what does

35:17

it mean? So for one

35:19

thing, it means that we

35:21

get even more startup inquiries,

35:23

of course. And not all

35:25

of them. have really read

35:28

what I do, so I'm

35:30

focusing on AI, I'm focusing

35:32

on technology, and some people

35:34

just say I'm addressing the

35:36

business age of the year,

35:38

and I don't care what

35:41

he does, I just send

35:43

out my pitch date and

35:45

so on. So some of

35:47

them do their research quite

35:49

well before they address me,

35:52

but some don't. So that's

35:54

one thing. then there is

35:56

some there was some interest

35:58

from from formats press podcast

36:00

I think you also came

36:02

through through band network network

36:05

but actually because I'm a

36:07

well-known AI expert in Germany

36:09

already I was in the

36:11

press anyway, so it's nothing

36:13

new that suddenly erupted and

36:15

I was not prepared to

36:18

give interviews or so. So

36:20

I was into Dr. Tseidong,

36:22

if I had said many

36:24

other formats four times on

36:26

TV last year for AI

36:28

topics. And so in general

36:31

it was... Not a new

36:33

thing for me to get

36:35

pressing queries, but suddenly from

36:37

a new side because before

36:39

people did not interview me

36:42

on the topic of investment,

36:44

but rather on the topic

36:46

of what is AI going

36:48

to change in our world.

36:50

I've read through why, by

36:52

the way, we need to

36:55

point out again that you

36:57

get only into the group

36:59

to be nominated as business

37:01

agent of the year because

37:03

the startups you are investing

37:05

in recommend you. Yes. And

37:08

you got the award you

37:10

were recognized for, I quote,

37:12

building bridges between startups. What

37:14

is your approach of network

37:16

and collaboration there? Yes, so

37:18

I mean, as a business

37:21

manager, what can I do

37:23

for the startups? So I

37:25

can give them money, yes,

37:27

but money is not everything

37:29

a startup needs. Very often

37:32

they also need advice, but

37:34

also I think interchanging ideas

37:36

between startups and... good practices

37:38

between the startups and so

37:40

on. That's also very helpful.

37:42

And we do this actively.

37:45

So we meet with startups

37:47

online, we bring them together,

37:49

we also have done some

37:51

forms like do AI together,

37:53

exchange your knowledge on AI,

37:55

exchange your experience, maybe also

37:58

exchange some code and some

38:00

other things together. And also

38:02

we have a format called

38:04

the startup Future Days, which

38:06

takes once a year in

38:08

summer. and we invite all

38:11

starters so we pay the

38:13

the general hospitality except for

38:15

their own travel and so

38:17

so we hire the conference

38:19

room and pay all that

38:22

and invite some guests and

38:24

so we bring all the

38:26

starters together the sea level

38:28

and they do we start

38:30

with the bar camp format

38:32

so that means people can

38:35

spontaneously say I could tell

38:37

something on this best practice

38:39

I have or I want

38:41

to discuss with other people

38:43

on the topic of how

38:45

to get good talents or

38:48

on whatever, so many typical

38:50

startup things. And the second

38:52

day we have prepared workshops,

38:54

so somebody from the startups

38:56

prepares a workshop, which is

38:58

then going on for four

39:01

hours. And because that's prepared,

39:03

there's a lot of theory,

39:05

theoretical knowledge and so on,

39:07

they can be or they

39:09

have prepared. practice so you

39:12

have to do something some

39:14

tasks you have to work

39:16

on on your own to

39:18

do something and so on

39:20

and my startups actually value

39:22

this very very positively every

39:25

time we do it and

39:27

I have not heard of

39:29

any other business agencies who

39:31

does this it's quite an

39:33

investment also in time not

39:35

only money and but I

39:38

think This is something that

39:40

really brings the startups forward.

39:42

Talking about bringing the startups

39:44

forward, you often help startups

39:46

to optimize their AI models.

39:48

I was wondering, how hands

39:51

on are you? Are you

39:53

sitting in front of the

39:55

desk together with one of

39:57

the senior developers? I say,

39:59

do it like this, do

40:02

it like this, like on

40:04

the vertical screens? Black background,

40:06

that's what I had in

40:08

mind there. So the last

40:10

time I programmed myself is

40:12

about five years ago before

40:15

the pandemic. So I'm not...

40:17

as practical as you now

40:19

described, but last year I

40:21

actually went into some details

40:23

and we built a new

40:25

kind of loss function for

40:28

one of the models. And

40:30

I was deeply involved in

40:32

that. And I think, so

40:34

I mean, this sounds like

40:36

a bit bragging, but I'm

40:38

quite sure it would not

40:41

have worked by far as

40:43

well as when I joined.

40:45

and help them prove that.

40:47

So sometimes I do deep?

40:49

Sometimes I do go deep

40:52

and normally I have not

40:54

coded for a few years.

40:56

It ain't bragging if it's

40:58

true. You are an AI

41:00

specialist. So I was curious,

41:02

how will AI impact startup

41:05

investing in the next five

41:07

to 10 years? Will AI

41:09

start picking winning startups? And

41:11

will this be a reinforcing

41:13

circle? Yes, it definitely will.

41:15

So already right now, a

41:18

lot of professional investors, so

41:20

funds who invest with other

41:22

people's money, are already using

41:24

AI tools to scan the

41:26

market to discover potential startups

41:28

for their deal flow so

41:31

that they don't have to

41:33

wait for those dollars applying,

41:35

but rather can go actively.

41:37

and tell them, oh, we're

41:39

interested in investing, so please

41:42

tell us more. And this

41:44

is already happening right now.

41:46

Also for selecting out of

41:48

the deal flow coming in,

41:50

who is interesting, and who

41:52

is not so interesting, I'm

41:55

not the only one who

41:57

has built own AI for

41:59

that. As far as I

42:01

know, there are no tools

42:03

who do everything yet. There

42:05

are some tools who we

42:08

have taken a closer look

42:10

at the frequent angels. in

42:12

band. We've taken a closer

42:14

look at some tools who

42:16

promise to help you invest

42:18

better and select the good

42:21

ones from the bad ones,

42:23

but the tools were actually

42:25

producing a lot of paper,

42:27

non-physical paper, long PDFs with

42:29

so many research items and

42:32

did not really help invest

42:34

rather help. justify investments once

42:36

you had decided because if

42:38

some investor and LP in

42:40

a fund asks you why

42:42

did you invest there then

42:45

you can send them the

42:47

75 page rationale in the

42:49

PDF and the the LP

42:51

will probably not go through

42:53

all the details. I'm very

42:55

confident he will have an

42:58

AI to give him five

43:00

bullet points out of that.

43:02

So definitely we will do

43:04

it better. So with the

43:06

things we are doing, we

43:08

are already doing it better

43:11

because our AI helps us

43:13

decide. Some other investors have

43:15

also built their own tools

43:17

to help them decide rather

43:19

than produce a lot of

43:22

PDF paper. What is one

43:24

area of the German startup

43:26

ecosystem that needs some major

43:28

change to stay competitive globally?

43:30

So there are several things

43:32

which need to change in

43:35

Europe. So for one thing

43:37

the AI Act has to

43:39

be reduced and currently I'm

43:41

not so confident that this

43:43

is going to happen. Because

43:45

it's not because of the

43:48

startups, because startups can adapt

43:50

to bad conditions as well

43:52

as to good conditions. The

43:54

problem is that... If people

43:56

are afraid of using AI

43:59

because they might violate the

44:01

AI Act and might get

44:03

punished, and I see this

44:05

a lot in the mittlestand,

44:07

so the mid-sized German companies

44:09

which are very important in

44:12

Germany, that they are... very

44:14

insecure what they are allowed

44:16

to use and what they

44:18

are not allowed to use.

44:20

And so the market breaks

44:22

away. People are reluctant to

44:25

use AI. They are afraid

44:27

to use AI, partly because

44:29

they don't understand it. But

44:31

even if you get them

44:33

educated, the AI still is

44:35

not precise and very difficult

44:38

and is subject to be

44:40

extended. very easily and we

44:42

need to get rid of

44:44

this of this threat of

44:46

this we call it Democles,

44:49

Democles sword, so the sword

44:51

that's above the head in

44:53

order to to also get

44:55

the market better in Europe

44:57

for buying AI tools. We

44:59

are likely to see a

45:02

new government in pre-soon. What

45:04

role do government policies play

45:06

in making Germany more or

45:08

less attractive for startups? I

45:10

think it does play a

45:12

big role for several reasons,

45:15

actually. So I'm not so

45:17

much a friend of state

45:19

financing everything, but if you

45:21

hinder it by the state,

45:23

that also does not. stand

45:25

for a good ecosystem. So

45:28

in Germany we need to

45:30

get down with a lot

45:32

of bureaucracy. And when I

45:34

started my first company, I

45:36

think I remember that we

45:39

had to do the VAT

45:41

only once per year. And

45:43

now as a new founder

45:45

startup you have to do

45:47

it every month. You have

45:49

to do the VAT calculation.

45:52

and the VAT text review

45:54

and so on. So that

45:56

has had to be done

45:58

for some... fraud also

46:00

that had happened. But so

46:03

it would be much better

46:05

if like when you just

46:07

declare that you are still

46:10

negative, so you're not making

46:12

money or so, then you

46:14

maybe can just omit all

46:17

the tax things. So you

46:19

just say, I vow that

46:21

we have no profits or

46:24

even we have no revenues

46:26

or even we have no

46:28

revenues. Then you can just

46:31

omit everything in the tax

46:33

regulations. That would take a

46:35

lot of pressure from new

46:38

startups, so starting from the

46:40

beginning. Because in the first

46:42

time for Casablanca, it has

46:45

taken like four years until

46:47

we have the first revenues

46:49

and we're far away from

46:52

profits right now. So, of

46:54

course, this is an exponential

46:56

curve, so I hope that

46:59

soon we will be very

47:01

profitable, but for the first

47:03

years we have to develop,

47:06

develop, develop, develop, and then

47:08

we also have a lot

47:10

of paperwork to do, and

47:13

that's not so nice. You

47:15

want to focus on your

47:17

innovation, you don't want to

47:20

focus on paperwork. So that's,

47:22

and Germany is very difficult,

47:24

so when you start a

47:27

company it takes like six

47:29

weeks until you can start

47:31

the company, so a limited

47:34

company. And I think politics

47:36

can do a lot of

47:38

good for the start of

47:41

ecosystem by just reducing all

47:43

these obstacles. Hmm. They can.

47:45

I like to have you

47:48

as guess because you're known

47:50

for making bold predictions. What's

47:52

one thing about the startup

47:55

world in 2030 that only

47:57

few people are seeing coming?

47:59

Yes, the startup world in

48:02

2030. So I think what

48:04

few people see coming is

48:06

that probably the economy will

48:09

change a lot until 2030.

48:11

So I feel that the

48:13

German car industry will go

48:16

down by 50% until then.

48:18

So I hope that I'm

48:20

completely wrong, but it might

48:23

happen and the car industry

48:25

is so central for Germany.

48:27

So what will happen is

48:30

that we have. many people

48:32

who are unemployed at that

48:34

time. And if we continue

48:37

to just say, okay, if

48:39

you're unemployed, the state will

48:41

care for everything, then we

48:44

will not get power into

48:46

the starter ecosystem. But if

48:48

the state rules change, and

48:51

you really, for one thing,

48:53

empower the people to start

48:55

a company, on the other

48:58

hand, also drive them to

49:00

take up... either work or

49:02

start a new manufacturing company

49:05

for glasses. So he had

49:07

some idea of how to

49:09

improve that he was in

49:11

that industry before and he's

49:14

quite successful with that. And

49:16

also the starter I told

49:18

you about before where I

49:21

had the exit within 18

49:23

months was also founded by

49:25

two people who were at

49:28

the end of their 50s.

49:30

So now they're both over

49:32

60 and who had left

49:35

their traditional careers actually a

49:37

few years ago, built another

49:39

startup and then started this

49:42

one. But So it's not

49:44

that startups can only be

49:46

started by 18 year olds

49:49

or 25 year olds who

49:51

just left university or so.

49:53

That's not necessary. Also, older

49:56

people can be very innovative

49:58

and very inventive and we

50:00

will get all this knowledge

50:03

from the engineers coming into

50:05

the startup world in a

50:07

few years. I've

50:10

once heard that statistically speaking,

50:12

the most successful entrepreneurs found

50:14

the company with 45. By

50:17

the way, what came to

50:19

me, most people listening to

50:21

this already speak German, they

50:24

speak three to four languages,

50:26

but for everybody who wants

50:28

to practice German, the monthly

50:31

VAT filing you were referring

50:33

to is called Unzachroya for

50:36

Anmeldung, one word. Yeah,

50:38

so the German words

50:40

get very bad. Oh,

50:43

don't worry. We only

50:45

have two more questions

50:47

left. So you are

50:49

a mentor and invest

50:52

in AI startups. If

50:54

you were to start a

50:56

fund today, what would

50:59

its focus be? If

51:01

I would start a

51:03

startup now. A fund.

51:05

A fund. A fund. There

51:07

were so many questions I

51:09

prepared some but most I

51:12

didn't. So anyway. So if I

51:14

were to start a startup fund, what

51:16

would I do? So I think what

51:18

I would do is I would try

51:20

to spot the, I would use, so

51:23

if I would start a startup fund

51:25

now, I would use a lot of

51:27

AI and I would use a lot

51:29

of data because in a fund you

51:31

always have to justify also

51:34

what you're doing. and because

51:36

you're working with other people's

51:38

money, currently I'm just investing

51:40

my own money. So gut-feeding

51:42

is not so good for

51:44

a startup fund. So what

51:46

would I do? I would

51:48

use a lot of AI.

51:50

I would try to get

51:52

hold of many start-up stories

51:55

and analyze them and see

51:57

how the market shifts and

51:59

changes. over time and how

52:01

this changes. And then I

52:04

would also try to identify

52:06

which business angels benefit the

52:08

companies and I would try

52:10

to get into startups they

52:13

invested in and then or

52:15

try to get them on

52:17

board also if I'm investing.

52:20

Because when I started my

52:22

first company, if we had

52:24

had a business angel, oh,

52:26

we would have avoided so

52:29

many mistakes. It's not just

52:31

because of money, but we

52:33

did so many things wrong,

52:35

like not having a proper

52:38

packaging for our first product,

52:40

for example, which was sold

52:42

also in stores and so

52:45

on. So there are so

52:47

many things which we can

52:49

learn from an experienced business

52:51

person. So I would probably

52:54

take the business ends with

52:56

more into account than I

52:58

have the impression the current

53:01

startup funds do. To close

53:03

this interview and by the

53:05

way, we will ask you

53:07

where all the startups can

53:10

pitch you. I was wondering,

53:12

startup founders often chase VC

53:14

funding without considering other options.

53:17

When should they not raise

53:19

VC money? So I think

53:21

that people go for money

53:23

because they think money is

53:26

the... most important part of

53:28

the startup but I think

53:30

it's if you get the

53:33

wrong investor in too soon

53:35

they can destroy your company

53:37

because they so I think

53:39

you have to be wise

53:42

in selecting the investor so

53:44

so you should not take

53:46

seed seed funding from somebody

53:48

who is not helping you

53:51

but just pushing you in

53:53

their direction and you don't

53:55

like this direction or you're

53:58

not convinced also you shouldn't

54:00

do that as a startup

54:02

founder. What else? So I

54:04

think you should do some

54:07

part of your innovation before

54:09

you go out for money.

54:11

There are some people who

54:14

seek money when they have

54:16

just an idea. And so

54:18

of course, if you have

54:20

worked for as a, whatever,

54:23

CEO or co-founder or something

54:25

in a well-running company before,

54:27

then you can probably get

54:30

a lot of money already,

54:32

very early. Also, when you

54:34

work for Google or something

54:36

before, then you probably also

54:39

will get some money early.

54:41

But if you try to

54:43

get money, whatever it takes,

54:46

then it takes a lot.

54:48

And you have to make

54:50

many concessions with some people

54:52

who are just... investing early

54:55

at their high risk actually

54:57

also and but then they

54:59

will take a lot of

55:01

a lot of shares from

55:04

you at a low valuation

55:06

and for some founders that's

55:08

maybe not so necessary to

55:11

secure this so early but

55:13

generally people rather take take

55:15

up an investor too late

55:17

than too early so in

55:20

fact finder I could strap

55:22

fact finder and In 2007,

55:24

2009 we had over 40%

55:27

growth and while we were

55:29

still 5% to 10% evid

55:31

positive, not evidid positive, so

55:33

we were making real money

55:36

and also cash flow positive.

55:38

And we were growing over

55:40

40% and investors were coming

55:43

and asking, so can we

55:45

invest? Can we put some

55:47

money on the table? And

55:49

I did not do it

55:52

because I thought. Well 42

55:54

to 47% growth rate is

55:56

already so huge if we

55:58

grow more. probably the company

56:01

will do wrong or bad

56:03

things because we cannot control

56:05

all the growth. But I

56:08

think I made a mistake

56:10

because if I had taken

56:12

some money at that time,

56:14

probably we would have done

56:17

the international expansion much faster

56:19

and maybe much better. And

56:21

in the end... So I'm

56:24

not allowed to tell the

56:26

exact amount we had for

56:28

fact final, what was, let's

56:30

say, a middle, double-digit minimum.

56:33

And DECA, our US competitors,

56:35

sold for one billion to

56:37

Oracle. And even though the

56:40

founder, Steve Papa, only had

56:42

a minority share at that

56:44

time, He still definitely had

56:46

more than 10% and that

56:49

means he definitely made more

56:51

than 100 million for that.

56:53

So probably it would have

56:56

been better to just do

56:58

the international expansion, especially to

57:00

the US with investors' money

57:02

and also maybe with some

57:05

investors' guidance. So choose the

57:07

investor well, but do not

57:09

say I don't need money.

57:11

I don't need investors. If

57:14

you actually should do need

57:16

them to do the next

57:18

step. to do the next

57:21

real close step. Thank you.

57:23

Thank you very much for

57:25

answering for two interviews, taking

57:27

up two evenings of your

57:30

vacation. Thank you very much.

57:32

Greatly appreciated. Hope to have

57:34

you back soon. Thank you

57:37

very much. That's

57:42

all folks. Find

57:45

more news, streams,

57:47

events, and interviews

57:50

at www. Startuprad.io.

57:52

Remember, sharing is

57:54

carrying. You