June 2, 2025 • 61 mins
In this episode of The Blunt Dollar, I sit down with Manuel Grenacher, a serial entrepreneur and founder of Unique, a company revolutionizing financial services with agentic AI.
Manuel shares his journey from building startups in Switzerland to leading one of the most exciting AI-driven companies in finance today.
We dive deep into how AI is reshaping the industry, where the biggest disruption opportunities lie, and why empathy remains irreplaceable by machines.
Key topics we cover:
🔹 What is agentic AI and how it goes beyond ChatGPT
🔹 How Unique is helping financial institutions scale AI adoption
🔹 The future of relationship managers in an AI-driven world
🔹 Trust, hallucination scores, and responsible AI use in finance
🔹 Why mid-sized banks could outperform the giants with AI
🔹 The real-world use cases driving the fastest ROI today (KYC, investment insights, and due diligence)
🔹 Manuel’s personal journey: from corporate comfort at SAP back to the startup grind
🔹 How to stay relevant and thrive in a world where AI becomes a superpower
If you like conversations about AI innovation, entrepreneurship, and the future of finance, this one's for you.
Oh, and if you haven't already... subscribe to The Blunt Dollar for more raw and honest finance conversations.
New episodes drop every other week! Available on Spotify, Apple Podcasts, and wherever you get your podcasts.
And last, but not least, don't forget to follow me on LinkedIn: https://www.linkedin.com/in/ignacio-ramirez-moreno-cfa/
Enjoy the episode!
Disclaimer: This podcast is for informational and educational purposes only. It does not constitute financial, investment, or legal advice. Listeners should consult a qualified financial professional before making any financial decisions.
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
You might remember back in times where cloud was
coming or iPhone got launched.
I guess they saw these bigchanges we saw over the last 20
years.
There is always a big hype forit and it's as larger as better,
right.
So because then you have a nicemainstream.
Speaker 2 (00:13):
Welcome to a new episode of the Blunt Dollar.
Today's guest is ManuelGrenaher, someone who knows a
thing or two about buildingcompanies from the ground up.
Manuel has gone from launchingstartups in Switzerland to
leading the charge in AI-drivenautomation for finance, and he
has a sharp perspective on wherethings are headed next, both
for AI and for business as awhole.
(00:35):
I want to ask you about data,because AI obviously is only as
good as the data it learns from.
Speaker 1 (00:40):
I think what makes the difference is what you have
in your unique data sets, inyour bank or in your assets.
I mean, there is always acertain bias, but that's good.
So, in a way, because that'salso the culture, potentially,
or the direction as your firmworks right.
I get often asked is my jobsafe?
And I say I think it doesn'tmatter if you're a relationship
(01:01):
manager, a designer, a cod coderand ai can give you all the
information to maybe get to 90,but the last 10 you have still a
human advice.
You could google up what you,what you eat, for medicine, but
you still love to go to a doctor.
Just get a double check on it,you know and the same will be in
(01:36):
finance.
Speaker 2 (01:37):
this is the blonde Dollar with Ignacio Ramirez
legal advice.
This content is forinformational and educational
purposes only and should not berelied upon as a substitute for
professional advice.
Always do your own research andconsult a qualified advisor
before making any financialdecisions.
All investments involve risk,including the potential loss of
capital.
And now let's get started withthe episode.
(01:59):
Hello everyone, and welcome to anew episode of the Blunt Dollar
.
Today's guest is ManuelGrenaher, someone who knows a
thing or two about buildingcompanies from the ground up.
Manuel is a serial entrepreneur, tech visionary and the founder
of Unique, a company that'sbringing AI into the financial
services space in a way that'sactually very, very useful.
(02:23):
Before Unique, manuel builtCore Systems, a service
management platform that gotacquired by SAP.
He also founded Mila, acrowdsourced workforce company
that was later snapped up bySwisscom.
Basically, he has a habit ofbuilding things that big players
want to buy.
Manuel has gone from launchingstartups in Switzerland to
(02:43):
leading the charge in AI-drivenautomation for finance, and he
has a sharp perspective on wherethings are headed next, both
for AI and for business as awhole.
I'm very, very excited abouttoday's conversation with Manuel
.
We are going to be talking, ofcourse, a lot about AI, but not
only Manuel.
Welcome to the show.
Speaker 1 (03:05):
Nacho, thank you very much for having me in.
I'm super excited about thisshow.
So, to start things off, canyou tell us a little bit of what
you're doing today with Unique?
So today we really shaped thefinancial service industry,
bringing authentic AI to thefinancial service industry.
Help them to transform and helpthem to you know, better serve
(03:25):
their clients, reduce costs byimplementing generative AI, but
also agentic AI at scale.
Speaker 2 (03:31):
So let me sorry, let me just stop you there for a
second.
For those listeners that haveno clue about what agentic AI is
, can you maybe define it in acouple of sentences?
Speaker 1 (03:42):
Absolutely.
Think of chat.
Gpt today brings youinformation much faster and more
summarized than maybe Google,where you have to go through
various pages.
So authentic AI is not onlybringing you information more
efficiently, it also do work foryou.
So it makes tasks.
Think of your search on ChatGP.
(04:03):
It may be a good place to visit.
Agentic AI would book theflights, the hotels, negotiate
the prices for you and get thebest offer out of it.
So that's the agentic.
Both uses AI to be moreefficient, but agentic are kind
of agents, human AIs that do thework for you.
Speaker 2 (04:24):
So it goes beyond just providing information and
it actually does things.
Speaker 1 (04:27):
in short, Exactly, and that is the only way how you
can become more efficient.
Let's say, when you have toprepare for your client meeting
and, of course, chat.
Gpt-like application canalready help you to get the
information more summarized andmore to the point, but agentic
AI would also build thepresentation and the offer
(04:49):
letter for your client.
Speaker 2 (04:50):
So is that what you're doing with Unique
building agentic AI for finance?
Speaker 1 (04:55):
Absolutely in various ways.
So on one side, we help bankersto do better investment
insights and propose betteroffers to their client, have
more time for their clients byreducing admin effort internally
, but we also work on agents forKYC to make KYC more efficient
(05:15):
so that, again, bankers canspend more time with their
clients.
And also we work heavily in thedue diligence of potential
finance investments or privateequity investments.
So we have over 40, more than40 clients now on our platform
and over 30,000 financialprofessionals that use our
platform every day.
Speaker 2 (05:34):
Wow, that's absolutely fantastic and
congrats on that milestone.
On such a short period of time,very, very excited about what
you guys are building.
Period of time Very, veryexcited about what you guys are
building.
So obviously you might be alittle bit biased here, but tell
me, is AI really the next bigrevolution in finance or do you
(05:55):
think there's a lot of hypeinvolved in the way we're
talking about it at the moment?
Speaker 1 (06:01):
I mean, hype is always needed, you know, to do a
big transformation.
Uh, you might remember back inthe times where cloud was coming
or a mobile where iphone got,you know, launched.
I guess they saw these bigchanges.
We saw the last 20 years andthere is always a big hype curve
and this is as larger as better, right so, because then you
have a nice mainstream.
(06:22):
But, but what I see is we arejust in the very early
beginnings of AI.
I mean, I guess in our clientsthey all rolled out AI to their
people, but we are in apercentage.
There is still a long way to go, because what is now important
is we have great models outthere.
I mean they are fantastic, butnow it's important to find the
(06:43):
big problems to solve and makesure that those models can help
to really solve these bigproblems.
And there is still a lot to doand players like unique has a
huge market there to help largeorganizations on mid-size
organizations to use these largelanguage models in a way they
really solve real businessproblems and not just, you know,
(07:03):
and use chat shippity in ourprivate life.
Speaker 2 (07:05):
And how does Unique work?
Do you have your own languagemodels or you go out there, you
see everything that is availableand you pick up those that you
think are the best ones for thespecific tasks at hand.
Speaker 1 (07:18):
Absolutely.
We not build our own modelModels.
To build model is an own way.
You know, either you are veryvertical and you solve real
business problems or you'rehorizontal and build models like
open, ai and tropic and andalso, to be honest, we would not
have enough funding here ineurope to build such models.
When you look now, you knowwhat they raise to build models.
(07:40):
It's so much compute.
This is either america or china.
So what we do is we look to theprocesses on, like financial
institutions, where we see whatwe really could do better and
disrupt by using the.
You know the new way how youwork with an application, like
you chat the gans and you don'tuse these complicated screens
(08:03):
anymore, and then we look whichmodel performs best to make this
task.
This could be either an open AImodel or an open source model.
When we need to deploy themodel, even on a very secure way
to keep privacy and datasecurity high, because often our
clients they really want tohave everything in their private
(08:23):
cloud or even on-prem becauseit's very sensitive data.
They don't want that.
The model providers can learnfrom the data and potentially
train the next generation ofmodels, so that's why privacy is
a super important thing for usand depends on the requirements
we choose the right model.
And to make an own model makesno sense.
Yeah, and to make an own modelmakes no sense.
Speaker 2 (08:44):
Yeah, you were talking about disruption and
focusing on areas where you candisrupt.
What are the most disruptibleparts of the finance industry
that you feel AI can disrupt atthe moment?
Speaker 1 (09:05):
Yeah, you can go crazy, of course, and think you
build the next banker as an aiagent, but I believe still that
bankers or relationship managers, humans bring the empathy and
that you cannot replace by ai.
But what is our job is to makesure that these relationship
managers have much more time fortheir clients and we reduce
everything around, or let's say,supercharge everything around,
(09:27):
so that the bankers can do amuch better service to their
clients, because often arelationship manager spend too
much time to gather information,need to talk to various people
in the firm and it slows downthe customer service experience,
right?
So think of you have these allon fingertip and you can really
provide it to your clients inreal time.
So what we, our mission, isreally to make sure our clients,
(09:50):
the relationship managers, getmore client time and less admin
time.
That's the first phase.
The second phase, of course, isthen helping also to build
better products.
Help to, you know, in the end,drive alpha by choosing the
better products in the situationthe client is or the client has
needs.
So helping to build a better,let's say, offer for the client.
Speaker 2 (10:10):
But currently, I would say the next 12, 24 months
, it's all about efficiency,efficiency, efficiency, so that
relationship managers can domore sales and spend more time
with the clients I tend to agreewith the statement that RM's
jobs are going to be harder toreplace and everything related
to empathy is going to be moreshielded from AI disruption, but
(10:34):
I also think we're a little bitbiased by the state of AI today
, and we almost cannot imaginehow AI is going to look like 20,
30 years down the road, and youknow better than me that it's
moving crazy, crazy fast.
My question for you is why iseverything related to empathy so
(10:57):
hard for AI to touch as oftoday?
Speaker 1 (11:01):
Because it's human, so simple, you know.
I mean, if you spend money, youwant to talk to someone.
Even you might already haveyour decision.
You want to see someone, feelsomeone to give you feedback Is
this a good or bad decision?
And AI can give you all theinformation to maybe get to 90%,
but the last 10% you have stilla human advice in everything in
(11:25):
your life.
It's even with you could googleup what you, what you eat for
medicine, but you still love togo to a doctor to just get a
double check on it.
You know, and the same will bein finance.
That's why a relationship, apotentially the next generation
or so, has a more relation to anai agent.
But I believe this will not be,you know, something that get
(11:49):
disrupt.
What we see definitely is thatwhat happens.
Is that the biggest challengefor banks is that not that they
are slow in adapting, you know,against their competition, it's,
it's more that their endclients, especially the next
generation, let's say, wealthyclients that comes next are
using these tools, maybe evenmore efficient than the bank
(12:13):
itself.
So they are getting the clients, the end clients, getting much
more equipped by doing research.
You know, for a $200subscription, with perplexity.
You can already have a miniresearcher for you doing
investment research, and whatcould happen is that this
research is even better thansome banks can provide, you know
.
So that's why banks andadvisors need to really skill up
(12:37):
that you are outperform whatthe clients can do themselves
with using, you know, this kindof commodity tools out there,
and I guess that's what I see asa bigger threat for
relationship managers that theyget under pressure that their
clients becoming, let's say,more AI smart themselves.
Speaker 2 (12:58):
Which is probably a good thing, right?
Because it will push everyoneto move faster and talking about
more.
Speaker 1 (13:05):
You potentially get not replaced by AI, but you get
potentially replaced by anotherrelationship manager who uses AI
much more efficiently.
That could be definitely thebigger threat.
Because I get often asked is myjob safe?
And I say I think it doesn'tmatter if you're a relationship
manager, a designer, a coder,it's in all the professions out
(13:26):
there, in my opinion, you haveto skill up using this AI.
Let's say we said before hi,but I would say even using these
AI tools and skill up as fastas possible.
Speaker 2 (13:38):
Yeah, that sentence I mean.
All of you are going to gettired of hearing about it, but
it's definitely going to comeback over and over again over
the coming years which is, ai isnot going to replace you, but
someone using AI probably will,and hence why it's so important
to stay updated, educateyourself, learn how to use these
tools and become part of thechange, rather than staying on
(13:59):
the sidelines and watch how yourjob disappears or is taken over
by someone else or the machines.
But, yeah, we were talking aboutthe speed, the pace at which
this is happening and the factthat, yeah, eventually, clients
and retail investors couldbecome more familiar or have a
(14:23):
higher degree of expertise usingthese tools than some of the
advisors and relationshipmanagers and banks overall.
But I must say, in defense ofthe financial industry, that I
was really surprised by thespeed at which they have
embraced AI, because I remembera few years ago, everyone was
talking about compliance risksand how this would never happen
(14:46):
just because it was so dangerous, and what I've seen is that AI
adoption in finance is movingway faster than what any people
thought.
So what's actually drivingfinancial institutions to
embrace AI faster than whatpeople initially thought?
Speaker 1 (15:02):
Yeah, faster than what people initially thought,
yeah, yeah, and I can also tellyou why.
Because when you look tosoftware, software is
complicated, you know.
When you look to your corebanking systems or your CRM
systems, they are all difficultto learn, difficult to use,
because you have so many screens, fields, functions, buttons.
(15:22):
You have to learn to usesoftware, right.
But ChatGPT and we have to saythank you to them, kind of
disrupted is bringing a chatthat we all know from WhatsApp
in combination with these modelsand give you a new interface so
you speak with your languageand explain the problem to solve
and the solution is getting youknow more or less back in
(15:44):
normal language to you.
You know, and I think thatdisrupted the entire software
industry a little bit.
You know, I mean today I see iton my kids and so on.
I mean they even don't usegoogle anymore because you have
to to search too many screensright and searches.
So they use chat, gpt and justtalk.
I mean this is completely kindof flipped how we use software.
(16:08):
Even chat is not new Chat weknow and use all the time.
But the combination of a chatand large-language intelligent
model so let's say a chat andintelligence together was in the
end the disruptive point.
And now I believe, at least,that that's the reason why also
(16:28):
our clients.
You know that the adoption ofthe solution is so fast because
you don't need to train theusers, they understand it
immediately, because they canuse their language, they can
explain the problem.
Maybe the answer is not thebest and maybe they work a
little bit around, explain itanother way and, and that's
fantastic, right, when you thinkwhat you can do now, um, with
(16:53):
such a solution, but you wouldneed to build software in the
old way, you know, withdifferent screens and so on and
so on, that you will be neverfinished.
Speaker 2 (17:01):
You know, and this, I guess, is that the main driver
for adoption it's crazy what youwere saying about google,
because I completely relate tothat.
I was using google all the timeand now it's not that I don't
use it anymore, but I probablyuse it like 10 of the time.
Um, because chat gpt.
I'm a big fan of chatGPT, so itgoes one step ahead and I can
(17:26):
tailor the content or the outputthat I'm seeing to the specific
use case that I'm researchingfor and that makes the whole
difference right.
Like, if I'm organizing a tripto Lisbon, I just don't want to
get a list of the 10 best hotels.
I want to see them and get aprogram that is suitable for my
(17:47):
profile and my interests andhopefully tomorrow, with the
arrival of Agent TKI that youwere alluding to, also make the
bookings and all the necessaryarrangements in a trustworthy
fashion, I mean as they useChatGPT, they call it the
operator, what you could now tryout in Europe.
Speaker 1 (18:09):
I mean, this is fantastic, fantastic.
This operator is exactly doingwhat we do for our clients, but
for private life.
It really already book yourflight and book your hotel,
completely automated, and uh and.
And this is what we do forfinancial service industries,
but of course, in a much higher,let's say, enterprise level and
in a much more secure way,because, again, a chat GPT-like
application they also learn fromevery prompt from you, you know
(18:33):
, to then learn and train thenext generations of models, and
that's okay when you assign yourprivate data to it, but as a
financial service provider, ofcourse, you cannot.
Speaker 2 (18:44):
I have a question regarding the power struggle
between big financialinstitutions and fintechs,
because when we talk about AIand the future of the finance
industry and particularly, forexample, wealth management,
where I work some people thinkthat the big banks, the big
names, will dominate, becauseyou need resources to scale AI
(19:05):
across.
You know organizations, butothers believe that fintechs,
given that they're more agileand maybe more willing to take
risk and pushing boundaries,will adapt faster than what
banks could possibly do.
So who do you think is going towin this race?
(19:26):
Do you see traditional bankscoming out as the big winners,
or will fintechs grab some oftheir market share thanks to
their agility?
Speaker 1 (19:40):
Yeah, I personally believe that the mid-sized banks
have a right to win herebecause they can adapt the AI
much faster than maybe a largeorganization like UBS or
something like that, like aretail bank.
Because, again, it's not onlythe technology, it's how you do
(20:03):
the transformation that the userusing AI, that the employee is
using AI.
You know what I mean.
So, and as large yourorganization is, when you have
like a hundred thousandemployees and more, you know as
long as you take.
So that's why I believe themidsize banks that we also, by
the way, focus on on our clientbase, like we have a LGT, picte,
(20:26):
julius Baer and so on they, inmy opinion, can outperform large
organizations now.
But we also would see becausewhy Again I explained before you
can grow faster with hiringless people by implementing AI,
but a large organization needsto first bring costs down in a
(20:46):
larger amount, you know to tokind of keep the same cost
income ratio.
And of course, we also seefintech.
But I see more.
I believe more in the retail,um, you know, like insurance, uh
, like revolut, as an examplewhere you do you might saw the
announcement from RobinhoodPrivate Bank, yeah, but this is
(21:07):
still more in the retail, youknow, for smaller, for smaller,
let's say, private bankers butor private clients Out of
Involve Management, assetManagement in the private equity
.
I still believe it's this thesemid-sized companies will
potentially have the biggestopportunity I had Because, again
(21:29):
, you don't need any morethousands of people.
You need an awesome platform,like Unique AI, of course, and
then you need a small team thatreally builds these agents at
scale and you don't need anymore so much coder, because the
models build the code for you.
Right?
It's a complete new direction.
(21:50):
Some of our clients they havemaybe 10 people and build
awesome AI agents.
I mean, when you look to theoutput in, let's say, the
traditional way, you need maybea few hundred developers to get
the same improvements.
You know what I mean.
So it completely disrupts thething.
But, yeah, maybe we see alsosome new fintechs or new
(22:13):
companies that outperform bankswith having only a few hundred
or maybe a few hundred people.
But the only issue I see thereis what we discussed before why
our clients are so successfulbecause they have a trusted
brand, you know and trust.
You cannot accelerate with ai.
That's not possible.
You can be more efficient.
(22:34):
You can be more cheap, more,more efficient, but but you
cannot build trust faster withai.
Speaker 2 (22:39):
That doesn't work it's almost like sky is the
limit and the only limit reallyis your imagination.
Um, with all these tools thatyou know, give you kind of
superpowers and enhance yourproductivity.
I'm that's why I I love ai somuch.
It's, uh, it kind of levels upthe playing field and and you
(22:59):
can pretty much do almosteverything you want, which I
think is really, really cool.
I want to ask you about data,because AI obviously is only as
good as the data it learns fromand, as we all know it, big data
sets, particularly in thefinancial industry, are full of
biases.
How do we prevent AI fromunintentionally reinforcing
(23:23):
existing biases in financialdecision making?
And here's my promise If you do, I'll keep bringing you honest
(23:44):
conversations, freshperspectives and the kind of
finance talk that's engaging,insightful and worth your time.
Thanks so much for being hereand let's keep these great
finance conversations going.
Speaker 1 (23:57):
Good, you said it right.
I mean, the AI is only as goodyour data.
Of course, you can use somepublic information, but everyone
can use public information.
I think what makes thedifference is what you have,
let's say, in a way becausethat's also the culture,
potentially, or the direction asyour firm works right ways, how
(24:44):
you can ingest this private,unique data you have on hand in
a in a in, a, first of all,secure way, but also in a way
that most people in theorganization can benefit from.
Yeah, that's still a bigchallenge.
It's still a big challenge toget the enterprise, let's say,
set up that it's everything issecure and locked, and so on and
so on.
But also how you build usecases that they only use the
data they need and not have toomuch data or too less data,
(25:06):
because if it's too much data,then you have potential
hallucinations.
If it's not enough data, youdon't have enough accuracy on
the results.
And I always say this issomething every, every financial
institution need to try now andlearn uh, you know, and build
use cases and and, and you know,improve them and measure the
(25:30):
success.
And if the success is not there,stop the use case, make the
next use case.
There is no magic, you know.
So you really have to do yourhomework.
You have to figure out whichuse cases are, you know needs,
which data, and I think what wediscussed before most relevant
thing is that you not just havea better search of the
(25:51):
information, because that bringsyou certain efficiency, but it
will not change the game.
It's more like how you canreally build workflows or
replace, you know, human, boringwork by having workflows, using
these various data sets or datapoints to do a preparation of
(26:12):
your pitch deck or whatever.
I guess that's then somethingyou measure in hours of hours of
work improvements.
Right, and this ingestion ofthe data is still a challenge I
had in all our organization,because often the data is not up
to date, there is old, and youneed to tell the model hey, look
(26:34):
, if the data is not up to dateanymore, then you have to do
this and that you know that youthen have good results.
Speaker 2 (26:44):
Follow-up question.
Maybe the second derivativerelated to the biases in data
and the quality of data, whichis linked to responsibility.
So let's say, an AI modelmisjudges risk, approves a loan
(27:09):
to a wrong customer or justsomething that goes wrong on a
model.
So who's responsible?
The firm that deployed it?
The regulators who failed tooversee this kind of scenario?
The coders who build thelanguage model?
Who are we to blame ifsomething goes wrong with these
AI systems that we're deploying?
Speaker 1 (27:35):
Yeah, of course that's a difficult question
because in the end it's what isin the contracts between the
different parties you justmentioned.
But what I can just share, thatour clients always have a human
in the loop concept.
So that means, whatever oursolution produce, a human is in
charge.
The human take it either, sendit to a client you know, review
it, approve it.
(27:55):
We build tools around it tohelp the human to, let's say,
judge the hallucination scoreand this kind of stuff, but
still the human is in charge.
So we cannot, in financialservice, blame ai for wrong
decisions.
It's still the humans and Ithink that will keep over the
next, let's say, years.
Of course, you become moreconvinced that the results are
(28:17):
good.
You get a feeling.
You may not double checkeverything, but you always have
an eye on it.
And as more you use a feelingyou may not double check
everything, but you always havean eye on it and as more you use
these tools, you also know whatcould go wrong, where you might
need to double check.
But the responsibility is still, in my opinion, with the human,
like it was with all the toolsyou use in your daily life.
Right, um, if you go?
Because we at Unique, as anexample, we don't do solutions
(28:42):
where end clients directly.
You know, like use an AIadvisor as an example, we found
a banker between, because weknow it's very tough.
Still, hallucination can happen.
You know, as we discussedbefore, maybe your data sets are
not up to date.
Discussed before, maybe yourdata sets are not up to date,
(29:04):
you know.
So that's why a you, arelationship manager that
understands the business,understands the context, should
still look over the producedcontent or produced output that
you provide to your clients.
But then it's in the end, inyour policies and in your
governance who is responsible.
But I think it's hard to blamea model.
(29:25):
You know what I mean.
That will take a long way thatyou can blame a model.
Speaker 2 (29:32):
So I want to transition now to maybe the
second block in thisconversation, which is the part
focused on agentic AI that youbriefly touched upon at the
beginning of the conversation,topic you and I are very
passionate about.
So my first question is how faraway we are from a world where
(29:55):
AI is fully and autonomouslyrunning portfolios, making
lending decisions, setting riskstrategies and so on and so
forth, without human approval.
Speaker 1 (30:10):
Yeah, I mean the human approval, as I said before
, will be I'm still many, manyyears, but just on the results
but that agents can worktogether to at least optimize,
as you said before, portfoliosand strategies that that we are
very close to in in a very,let's say, easy way.
It may not that supercomplicated stuff but, um, I
(30:34):
would say in the next two, threeyears you could, you could
really see that these agentswork more automated together,
maybe not completely autonomous,but automated together.
I give you an example what wesee already today, what what
happens, um, in on our clientside.
So what, over the last 10 yearsor so, you know, all the
(30:57):
companies try to build huge datalakes of data and then try to
clean the data and so on and soon and in the end often not get
really good success because it'sso difficult to do.
But what happens now is youbuild agent that has just a
subset of this data that youpotentially can clean up, just
(31:19):
make ready faster because youcan.
It's maybe just let's call it asimplified like.
You have a relationship manageragent that have access to CRM
data and portfolio informationdata and then you have a KYC
agent that has access to KYCinformation.
So you don't need to harmonizea data set between the two
agents, because every agent hastheir own little data set.
(31:42):
And now what happens is that thetwo agents can speak to each
other.
So that means when therelationship manager needs to
know something about the client,let's say, their companies that
are involved the client couldask the KVC agent hey, I'm a
relationship manager X, I wantto know about the client epsilon
(32:03):
the following and that agentknows then what data first of
all, compliance from acompliance point of view you can
pass over to the relationshipmanager agent so we can solve
all this compliance stuff andwith that, you don't need to
build any more such huge datalakes.
Right, and you will be more andmore of these agents that they
(32:23):
can smartly talk to each other,make sure they are compliant to
each other, make sure they passon information that are approved
, and so on and so on.
And this, in my opinion, is thehuge, huge opportunity in the
future, when you have a lot ofdata in your house.
So these agents can worksmartly together.
They can have even contractwith each other what they can do
(32:44):
and whatnot.
Speaker 2 (32:46):
So let's assume you know those agents start
proliferating, Everyone isadopting them.
Do you think we'll reach apoint where those agents are so
widespread that they startcompeting against each other
very aggressively, to a pointwhere, for example, in financial
markets, it could create somesort of AI-driven volatility and
(33:08):
market chaos?
Speaker 1 (33:11):
It's a good.
I mean, as you know, I wouldsay never know and I'm sure we
saw it in other situations,situations even with uh simple
online platforms, how fast youcan uh drive such situations.
That's why I would say,potentially, absolutely, yes, um
, but again, that's not mybusiness.
(33:33):
That's not my business, that'sone for the dystopic writers?
I guess exactly but when I onlylook about my own experience I
had when I we just recentlyraised 30 million in the series
a run and I got so many fishingso professionally, you know,
(33:53):
with fully ai driven, that Ijust was like, wow, there we are
.
You know, know what I mean.
It's just crazy how AI alsogets used to, in the end, to
criminal influence, right, and Ican only measure that on the
phishing things that arehappening at our company right
now and it's really crazy whatyou see.
Speaker 2 (34:14):
Congrats on raising those funds, by the way, which
I'm sure they're going to helpyou loads to keep expanding and,
yeah, offering your servicesacross more financial
institutions.
And talking about that, can yougive us a little bit of color,
a bit of juicy details of whatare the big banks doing at the
(34:38):
moment when it comes toimplementing agentic ai?
What is the low hanging fruitthat, uh, if you had to say most
, uh, most of the big guys aretrying to tackle or doing with
with these tools?
Speaker 1 (34:50):
so I guess more or less every financial
institutions has now a so-calledprivate chat gpt or like
application that you can uselike you use chat GPT at home,
right?
So upload a few files,summarize, translate and so on.
But again, most of the clientsfigured out that that maybe give
you a few percentage efficiency.
What we see now where ourclients really double down when
(35:14):
they have this private chat GPTkind of an internal knowledge
thing sorted out, they go intoreal business problems and I see
three at the moment.
That is involving one isinvestment insight assistance
right to combine your houseviews with external data
(35:36):
providers to really help you todo simple I would call it simple
investment research basedagainst your house views and
your you know other opinions.
You have combined with dataproviders from external and
maybe earning calls and so on.
That's a heavy thing for uh,for wealth managers, asset
(35:57):
managers and so on.
Then, um, kyc we discussedbefore.
So kyc is a huge pain.
It's a huge, huge pain.
It's a huge cost driver.
Um, so we see clients startingusing a unique ai to really help
bankers to do maybe not replacethe entire kyc process, but you
(36:17):
know how it is.
You get these massive checklists.
You should get all thisinformation from a client.
You know how it is.
You get these massivechecklists.
You should get all thisinformation from a client.
You get all the files from aclient just to check.
Do I have everything?
Can I answer all the questions?
What is missing?
Please build me a list of themissing documents, missing
information, so really make theonboarding of a client more
efficient.
But also then the liabilityreports and everything you need
(36:41):
later on to keep koc on, on, onon the right level.
And the last thing is everythingaround due diligence.
Due diligence on massive datarecords, right, so you have your
q a list or your due diligencequestionnaire of hundreds of
questions for maybe your next icmeeting or your next proposal
(37:03):
discussion, and you have a dataroom of the documents and you
use also in external informationof of a target fund or of a
target company.
I mean, that's insane how gooddue diligence work today, so we
could really save this kind ofconsultancies that do simple
(37:23):
research on those targetcompanies.
Today you really can do thiswith AI and then you have a due
diligence report that is as goodthat you would hire 10 people
to do basic research for yousomewhere in in a remote
location out of those three qic,due diligence and investing
(37:45):
where do you think the biggestbang for the buck is?
Speaker 2 (37:47):
where do you estimate , according to your yeah, to
your studies, uh, that thebiggest productivity gains can
be achieved over over the short?
Speaker 1 (37:57):
term.
So short term is definitely inthe.
I mean, when you look to costsavings, it's KYC, because
that's painful, it costs a lotmore and more.
Client time also goes awaybecause bankers has to do a lot
in this process, right, um, soit's not only a cost, is also
(38:18):
missing revenue opportunitiesbecause people are busy doing
kvc instead of selling toclients or, you know, engaging
with clients.
Then revenue wise is in ininvestment insights, because you
can build more offers, morefaster, you can bring faster
opinion.
You know various offers to yourclients, you have more talk
time with your clients, you know, as more you do, as more you
(38:40):
sell, simple, right, so, um, buttoday I think again we need to
free up this time for bankers byusing ai to to make faster,
better insights and betterresearch and due diligence, I
would say is another game on duediligence is less efficiency,
but more.
You get more deals into thepipeline or through the pipeline
(39:01):
because you can look into moretarget clients, you know.
So you get potentially a muchbetter deal flow and potentially
you can increase your assetunder management because you get
more deals, you get betterdeals faster through.
So all of those have a littledifferent angles how to drive
(39:22):
efficiency.
Speaker 2 (39:24):
And talk about the investing vertical.
I tend to agree.
I think you know having AI isgonna help a lot when it comes
to financial advice, but I havea question about trust.
What will it take for peopleand clients to trust AI driven
(39:45):
financial advice the same waythey trust a human advisor?
Hey there, quick ad break.
Do you work in the financeindustry and have a genuinely
interesting story to share?
I'm always on the hunt forgreat guests who bring raw,
unfiltered insights to the table, or maybe you know someone with
a story worth telling.
Please put us in touch.
(40:05):
You can reach out to medirectly via LinkedIn.
I'd love to hear from you.
And now back to the show.
Speaker 1 (40:17):
As I said before, when the human is still in the
loop for the client, it doesn'tchange anything for the end
client right.
So it makes you as arelationship manager more
efficient and maybe you'll findbetter opportunities based on
the needs.
You can go more into customizedoffers than just selling
standard products.
(40:37):
So it just makes your, let'ssay, portfolio that you could
offer to your client morecustomized faster.
So I guess that is the maindriver.
But the trust from your clientsis still the same because, as
we discussed before, when wehave the human in the loop,
nothing changes in the end.
Speaker 2 (40:58):
And do you think that there's like some sort of moral
obligation from people using AIto tell or to share with the
end user that AI has beeninvolved in the process when
coming with a specific outcome,like a presentation, a trade or
you name it?
Speaker 1 (41:16):
I think there will be soon more disclaimers, as you
have disclaimers anywhere,everywhere on slides and
proposals that you potentiallysee a certain content is
generated by AI.
Definitely, we'll see that, butit doesn't really change the
game because in the end, theclient invests into the
(41:37):
instruments or into the productsyou know, and even banks using
AI since ever, you know, maybenot generative AI, but AI for,
let's say, machine learning,driven funds and so on.
I mean, this is there alreadysince 1020 or even longer, you
know.
So I guess that will not be abig difference, but you will see
(41:58):
more disclaimers.
And what we have to do more andmore now is, you know, every
company implementing AIgovernance, and often what we
have to do with clients is thatusers need to understand more
what the AI is doing.
You know, as an example, whatare hallucinations?
(42:21):
Is there a potentialhallucination in that answer?
Right, so we need to do trafficlight systems.
Hey, this is green because wedon't believe there is
hallucination.
Yellow because we think not allthe directives got used, you
know.
Or even red because someresources that are provided the
(42:41):
answer not fits to theseresource files, and then this
means we need to make awarenessto the users where we
potentially, from a technologypoint of view, see hallucination
, so the hallucination is thebiggest, or a low accuracy
because we not found enoughinformation.
As an example, we only foundtwo sources and not 10, and so
(43:01):
on and so on, and I think thatwill be the biggest thing over
the next few years how you canbuild trust to the users that
the ai system provide you goodinformation and that we, at the
moment that unique ai, we dothis with hallucination scoring,
so we really tell the users um,look, we predict your
(43:25):
hallucination or not in thetraffic light system and then if
it's orange or red, then theuser spend more time on that
content to figure out.
Hey, okay, now I double checkand see what could be wrong.
Speaker 2 (43:37):
Maybe it's wrong, but at least I got alerted that I
have to look here.
Yeah, so you are the get yourtake on the future, what's ahead
?
I think AI can be used as aforce for good, but I'm curious
(43:58):
to hear from your mouth what canI do today to make sure I stay
relevant tomorrow in anAI-driven world and financial
industry an AI-driven?
Speaker 1 (44:11):
world and financial industry.
Yeah, so, as we discussedbefore, really spend time using
this tool in business but alsoprivate right.
Learn how to prompt, and Ialways say the easiest way to
learn prompt is asking ChatGPTto write me a prompt based on
your problem because peopledon't know that.
So you can ask ChatGPT to writeyou a prompt based on your
(44:32):
problem because people don'tknow that.
So you can ask chat gpt towrite you a prompt template to
solve that problem and youexplain the problem.
Right, let's say you need to doan end year closing reports,
what these are my materials,what you would write for a
prompt, and then you learn, youknow how to to to prompt the
right way, what you give forinstruction, because, again,
(44:54):
prompting means you need to tellsomeone, like a human, what to
do and see it like the model islike a virtual human and you
need to give instruction, youneed to give some background
information.
You need to maybe do that andthen check with me again before
you go into the next step.
(45:14):
And I think that's all for free,accessible out there right by
just using chat gpt.
Or you watch some youtubevideos?
Um, this is a, let's say, themost easiest way to start, just
use the, the.
The free stuff is already outthere.
And then, of course, you can gointo more learning programs,
and there are thousands ofcourses out there to learn, but
(45:36):
you have to be active yeah,rolling up your sleeves and and
getting your hands dirty, as wesay, but honestly, like it's
super fun.
Speaker 2 (45:45):
Uh, there's no better way of learning than just
trying and iterating.
And uh, you'll notice in notime that you start becoming
better and better if you make itpart of your daily routines.
I mean, personally, I use AIall day long, both personally
and professionally.
I cannot live without itanymore.
(46:05):
So which, actually, I thinkit's the beauty and the curse of
AI.
On one side, it makes you lazyfor certain things because
you're like, ah, I don't want towrite this anymore.
I'm going to ask you know ChadGPD to write it for me.
And that's bad, I guess,because it makes you think less
sometimes about specific things.
But on the other hand, it opensup a world of opportunity, as
(46:29):
we were saying before, becauseyou can do so many more things.
There's not that technicalconstraint anymore that you used
to have, and so on.
An example I like to give alsois one day I was at work and I
was thinking could actuallysomeone create some code to do
(46:50):
something very simple?
It was like a macro on VBA, onExcel, to automate a report.
But I was thinking couldsomeone that has no clue about
coding, actually do this usingAI.
And I sat down in front of thecomputer saying, okay, I'm going
to write this as if I had noclue about coding and I'm going
to try to do the task.
And actually I managed.
You know, like, because the AIexperience you step by step Okay
(47:13):
, do this, do that.
This is how you write the codeeditor copy paste this code.
Then you had an error, you putit back and it would suggest
that a change, and so on and soforth.
So I was like, wow, this istruly mind blowing.
I mean, yeah, if you spend sometime, you can do almost
anything nowadays.
So it's pretty cool.
So let's pivot now to the lastpart of the conversation.
(47:35):
We started very wide, talkingabout AI in general.
We went a bit more specific,talking about agentic AI and the
wealth management industry, andnow I want to wrap it up
talking about you, because Ithink you have a very
interesting personal journey.
So, to start with, let's talkabout the fact that, yeah, you
(47:56):
built and sold companies beforewe talked about core systems
which SAP acquired, and I thinka lot of people would have
stayed in that corporate world,but you jumped back into the
startup, grind and built Unique.
But you jump back into thestartup grind and build unique.
So what?
Speaker 1 (48:20):
made you leave the corporate safety of SAP and go
back to the trenches of buildinga crazy startup.
Look, I'm a passionateentrepreneur and I always say
you are or not right, so I couldnever be employed.
I mean, if I have to because Ipromise to do a post-murture, of
course I do, I keep always mypromise.
But I have to because I promiseto do a post-merger, of course
I do, I keep always my promise.
But I want to build stuff, Iwant to invent.
You know the future, and it'seasier to do that outside of a
(48:44):
corporate.
I mean, look, I mean I love SAP, it's a great company.
But you can imagine in such alarge company if you have ideas
to get funding for these ideas.
But you can imagine in such alarge company if you have ideas
to get funding for these ideas,you have to be in the end an
asshole because you have to doother people.
You need to make other peoplebad because then you get their
budgets for you.
You know what I mean, becausethere's not more budget in the
company.
But if you do a startup, youcan raise money from investors
(49:10):
right, and go to the market,pitch your ideas.
You get funding.
You know you can make them toshareholders, so it's just more
fun and participate on yoursuccess.
And in large corporation thisis very tough to do.
It's doable I don't say it'snot doable but you have to be
political.
You know, very political activeto to get some other
departments budget for yourdepartment right, and I just
(49:33):
don't like to do that.
I love to build stuff and youknow, motivate people to follow
me and you know, help me to tobuild the next generation
software company.
And I never.
I mean why I did this, do thisthe third time?
Because you know I neverachieved an ipo and I learned
also in my previous company.
I was never confident enough tosay let's build a public
(49:56):
company.
You know that was also onelearning I had.
That's why I tell now toeveryone look unique will become
one day a public company.
Um, so, so a company thatpeople like to work for, people
like to partner with, but alsopeople like to invest.
Because then you know, youbuild a company that everyone
loves.
Right, and this is what isstill a dream, and keep me is a
(50:22):
little bit my North Star.
You know that, keep me alwaysgoing, going, going.
Maybe it works with Unique,maybe with the next one,
definitely not giving up.
Speaker 2 (50:31):
Nice.
I love that.
That's a very big, bold, butreally cool objective to have.
So you're selling AI tofinancial institutions for a
living, which is by no means aneasy feat.
So what's your secret?
How do you manage to breakthrough the barriers of all
those big bureaucraticcorporations that you were
(50:53):
learning to?
That would kill most fintechs.
Um, what's the secret sauce?
Speaker 1 (51:00):
look, it's very hard to say there is no, the secret
sauce it's.
It's in the end you need to bealways on and try and try.
You know you need to winclients by building relationship
, like you have to do as arelationship with clients.
You build trust to the firstproducts, but mainly to the team
(51:21):
you have.
So I would say, as a serialentrepreneur, I have a little
bit the luck.
I already worked with the teambefore in other startups so I
can bring with me a very strongteam, because usually the first
clients not look into only theproduct, they also look into the
team you know they work with.
(51:42):
And I think there we, we justhave a.
You know, one of the why uniqueis successful is we have the
best people and I think you know, especially the early clients,
they can really see that.
That's why they partner up withunique now, since unique is
already three, four years on themarket.
And we have, of course, nicebrands like pigtail and then lgt
and then julius bear, and youcan name them right uh, swiss
(52:06):
life and so on.
So it helps, of course, thatother brands trust because we
have leveled up uniqueside-by-side with these brands.
You have now this, let's saybrand trust.
But the first clients are superhard.
To win them as a success story,you have to be as I said before
.
You have to be always on withthese clients and get them
(52:29):
excited about your team and yourproduct.
It really helps them to buildand solve problems in the field.
Speaker 2 (52:38):
I love also that we have a Swiss success story
within the field of AI.
It's pretty cool, because we'reused to read about American
names and Chinese names and soon, but it's cool to have
someone close to us doing this.
So I feel like, of course, aistartups are popping up
(52:59):
everywhere and investors arethrowing money at anything that
says AI powered at the moment,the same way it was for, for
example, the Metaverse some timeago or esg before that.
Um, how do you make sure thatunique doesn't become just
another ai hype story?
Speaker 1 (53:21):
so what's the difference now between, let's
say, also blockchain and otherstuff not blockchain in the
currency world, but in the,let's say, more softer world is
that clients can really measurebenefits, you know, really
measure efficiency gains and theadoption in in a in a, you know
, and in a in a, let's say,efficient way, also to how much
(53:46):
they have to spend and investand and I think that's different
between other technology shifts.
We saw, you know, likeblockchain was cool from a
technical point of view but veryhard to implement and see a
real business benefit in shortterm.
Right, when you implement theGen AI solution, like your firm
running as an example, you seefirst time, when you use this
(54:10):
application, you really get like, wow, okay, already may save
one hour, two hour efficiency,but just maybe onboard yourself
into a company or whatever youdo right, you feel it very fast.
And I think that's differentthan all the technology hypes we
have seen in the past, becausein the end, you have to make a
business case.
You have to run a business case.
(54:30):
That you know we have to helpour clients to save 10, 50, 100
millions so that we get a fewmillions out of it.
You know so and I think that wealready were able to prove
already today and again, as Isaid before, we had just started
, we're just in the firstpercentage.
So that's why I believe ai has aright to stay and especially,
(54:54):
you said it right yes, we hear alot from us vendors or chinese
vendors, and that's good so,because us always were in that
position.
You know they always do greatfoundation stuff, like cloud
services, like azure or aws andso on, and what is our
opportunity here in switzerlandis building now vertical ai
(55:15):
solution on top and partner upwith these american models.
Right, because the models wecannot build.
It's like back to the microsofttimes.
You know you always used theirwindows or whatever, let's say,
as the operating system.
But look to SAP.
They built an awesome verticalsolution for ERP and I think
(55:35):
that that is a right to win herein Switzerland or in Europe, to
focus on the vertical, tosolving real client problems.
I would not invest into a more.
I personally would not investin a horizontal solution in
Europe because, again, as I saidbefore, it's very hard to
finance because the largeinvestors are still in USA and
(55:56):
they usually love to invest inUSA, and for a vertical solution
, you need less money becauseyou don't need the large
infrastructure to train themodel.
You need more client time andyou need to be sure that you
learn from your clients and thatis your mode against the
vertical players, the horizontalplayers that you know the
(56:19):
client problems better.
You know the client data betterthan they do.
And, yes, I agree, it should bea little bit more unique in
Switzerland.
But we have other AI players nowpopping up and I hope this is
now a momentum in Zurich, inLausanne, around EPFL ETH, that
we really take this opportunityand also have our own, let's say
(56:43):
, ai supply in a way for ourclients.
I would love a little bit moreour Swiss corporations to trust
Swiss AI startups that stillmost of these managers love to
run after Microsoft or AWSbecause you know you do nothing
wrong.
That is something what Isometimes miss a little bit how
(57:05):
complicated some corporationsare to trust a Swiss startup to
go in a larger contract.
If we could overcome that also,then I believe we have a really
huge opportunity on working onour own AI supply here in
Switzerland.
Speaker 2 (57:23):
You mentioned Zurich, you mentioned Lausanne, but of
course also Geneva, absolutely Ijust mentioned.
Speaker 1 (57:30):
Hopefully we see some interesting companies popping
up In Borden.
I mean it's entire Switzerland.
Speaker 2 (57:38):
Maybe our last question on you and your journey
Not an easy one, becauserunning a company isn't all wins
.
You are the CEO, and CEOs haveto make brutal calls, sometimes
about obviously firing aboutpivoting strategies, walking
(57:58):
away from a big deal, thingslike that.
What's been the toughestleadership decision you've ever
had to make and what did youlearn from it?
Speaker 1 (58:08):
Wow, I mean there's tons of right.
Wow, I mean there's tons ofright.
But, of course, what is alwaysvery hard for, I guess, every
founder out there when you startto realize you missed your ICP,
your ideal client profile, andyou just have to let people go
because you had a belief insomething and it doesn't work
out.
And I had that at least threeto four times where I needed to
(58:29):
let a bigger group of group ofpeople just say sorry, it didn't
work out, you have to let themgo.
And this is always veryemotional.
But what I always say there is,when you get to that point, what
most startup do wrong.
They start to do small steps,you know, say, oh, let three
people go, then four and five,and that with that usually you
damage even more.
You should go then and say,okay, now we, we, we should let
(58:51):
10, 20 people go.
So, because this gives you airto breathe, to find another
direction, right, and that's myadvice to founders when you are
in that situation, then youshould better let more people go
, because you need the cash.
The cash burn is so hardbecause, you see, always the
(59:12):
cash burn is burning, burning,burning, burning and you don't
have new revenue to pay thesalaries.
Yeah, then you do it like inslot in in slices, because that
kills the culture, that killsyour brain, your energy and and
these are very hard decisionsand I did this many times and
it's not nice but it's neededit's the only way how you steer
(59:34):
the boat in another direction.
Speaker 2 (59:36):
Well, thanks for sharing those valuable insights,
manuel.
It's been absolutely amazinghaving this conversation with
you.
Thank you for sharing all theseinsights For the listeners that
want to hear more about Uniqueand what you guys are up to.
Where can they find moreinformation?
Speaker 1 (59:54):
So the easiest way is you go to uniqueai or on our
LinkedIn page.
You get always updates.
Feel free to comment there.
And we also share much morelearnings now from our financial
service clients.
We also do more and more eventsnow where we bring the
community together.
We just had recently one at sixwith over 200 financial experts
(01:00:18):
that showed up, and you know wehad a panel discussion about
the Gentic AI with some thoughtleaders and it's you know.
I see this as our mission toreally help the entire industry
to grow up in the world ofauthentic AI.
Speaker 2 (01:00:35):
Well, that's a great mission to have.
I'm sure you're going toabsolutely crush it.
I'll be front row to watch allof those successes.
I'm very excited for you guys.
Manuel, thank you so much fortaking the time of talking to me
today, and I hope 2025 is theyear where that IPO, that famous
IPO you're targeting, gets doneor at least gets a little bit
(01:00:56):
closer.
Speaker 1 (01:00:58):
Yeah, definitely a little bit more closer done and
still a way to go.
But I will definitely tell youwhen it's the right moment.
Thank you, all right.
Thank you very much for thetime and a nice rest of the day.
Speaker 2 (01:01:14):
The Blunt Dollar is written, produced, hosted and
edited by me, ignacio Ramirez.
Everything you hear concept,script, sound design and
production comes straight frommy desk and, occasionally, my
kitchen table.
Thank you so much for listeningand join me in the next episode
of the Blunt Dollar for moreraw, honest finance
conversations.
You might remember back in times where cloud was
coming or iPhone got launched.
I guess they saw these bigchanges we saw over the last 20
years.
There is always a big hype forit and it's as larger as better,
right.
So because then you have a nicemainstream.
Speaker 2 (00:13):
Welcome to a new episode of the Blunt Dollar.
Today's guest is ManuelGrenaher, someone who knows a
thing or two about buildingcompanies from the ground up.
Manuel has gone from launchingstartups in Switzerland to
leading the charge in AI-drivenautomation for finance, and he
has a sharp perspective on wherethings are headed next, both
for AI and for business as awhole.
(00:35):
I want to ask you about data,because AI obviously is only as
good as the data it learns from.
Speaker 1 (00:40):
I think what makes the difference is what you have
in your unique data sets, inyour bank or in your assets.
I mean, there is always acertain bias, but that's good.
So, in a way, because that'salso the culture, potentially,
or the direction as your firmworks right.
I get often asked is my jobsafe?
And I say I think it doesn'tmatter if you're a relationship
(01:01):
manager, a designer, a cod coderand ai can give you all the
information to maybe get to 90,but the last 10 you have still a
human advice.
You could google up what you,what you eat, for medicine, but
you still love to go to a doctor.
Just get a double check on it,you know and the same will be in
(01:36):
finance.
Speaker 2 (01:37):
this is the blonde Dollar with Ignacio Ramirez
legal advice.
This content is forinformational and educational
purposes only and should not berelied upon as a substitute for
professional advice.
Always do your own research andconsult a qualified advisor
before making any financialdecisions.
All investments involve risk,including the potential loss of
capital.
And now let's get started withthe episode.
(01:59):
Hello everyone, and welcome to anew episode of the Blunt Dollar
.
Today's guest is ManuelGrenaher, someone who knows a
thing or two about buildingcompanies from the ground up.
Manuel is a serial entrepreneur, tech visionary and the founder
of Unique, a company that'sbringing AI into the financial
services space in a way that'sactually very, very useful.
(02:23):
Before Unique, manuel builtCore Systems, a service
management platform that gotacquired by SAP.
He also founded Mila, acrowdsourced workforce company
that was later snapped up bySwisscom.
Basically, he has a habit ofbuilding things that big players
want to buy.
Manuel has gone from launchingstartups in Switzerland to
(02:43):
leading the charge in AI-drivenautomation for finance, and he
has a sharp perspective on wherethings are headed next, both
for AI and for business as awhole.
I'm very, very excited abouttoday's conversation with Manuel
.
We are going to be talking, ofcourse, a lot about AI, but not
only Manuel.
Welcome to the show.
Speaker 1 (03:05):
Nacho, thank you very much for having me in.
I'm super excited about thisshow.
So, to start things off, canyou tell us a little bit of what
you're doing today with Unique?
So today we really shaped thefinancial service industry,
bringing authentic AI to thefinancial service industry.
Help them to transform and helpthem to you know, better serve
(03:25):
their clients, reduce costs byimplementing generative AI, but
also agentic AI at scale.
Speaker 2 (03:31):
So let me sorry, let me just stop you there for a
second.
For those listeners that haveno clue about what agentic AI is
, can you maybe define it in acouple of sentences?
Speaker 1 (03:42):
Absolutely.
Think of chat.
Gpt today brings youinformation much faster and more
summarized than maybe Google,where you have to go through
various pages.
So authentic AI is not onlybringing you information more
efficiently, it also do work foryou.
So it makes tasks.
Think of your search on ChatGP.
(04:03):
It may be a good place to visit.
Agentic AI would book theflights, the hotels, negotiate
the prices for you and get thebest offer out of it.
So that's the agentic.
Both uses AI to be moreefficient, but agentic are kind
of agents, human AIs that do thework for you.
Speaker 2 (04:24):
So it goes beyond just providing information and
it actually does things.
Speaker 1 (04:27):
in short, Exactly, and that is the only way how you
can become more efficient.
Let's say, when you have toprepare for your client meeting
and, of course, chat.
Gpt-like application canalready help you to get the
information more summarized andmore to the point, but agentic
AI would also build thepresentation and the offer
(04:49):
letter for your client.
Speaker 2 (04:50):
So is that what you're doing with Unique
building agentic AI for finance?
Speaker 1 (04:55):
Absolutely in various ways.
So on one side, we help bankersto do better investment
insights and propose betteroffers to their client, have
more time for their clients byreducing admin effort internally
, but we also work on agents forKYC to make KYC more efficient
(05:15):
so that, again, bankers canspend more time with their
clients.
And also we work heavily in thedue diligence of potential
finance investments or privateequity investments.
So we have over 40, more than40 clients now on our platform
and over 30,000 financialprofessionals that use our
platform every day.
Speaker 2 (05:34):
Wow, that's absolutely fantastic and
congrats on that milestone.
On such a short period of time,very, very excited about what
you guys are building.
Period of time Very, veryexcited about what you guys are
building.
So obviously you might be alittle bit biased here, but tell
me, is AI really the next bigrevolution in finance or do you
(05:55):
think there's a lot of hypeinvolved in the way we're
talking about it at the moment?
Speaker 1 (06:01):
I mean, hype is always needed, you know, to do a
big transformation.
Uh, you might remember back inthe times where cloud was coming
or a mobile where iphone got,you know, launched.
I guess they saw these bigchanges.
We saw the last 20 years andthere is always a big hype curve
and this is as larger as better, right so, because then you
have a nice mainstream.
(06:22):
But, but what I see is we arejust in the very early
beginnings of AI.
I mean, I guess in our clientsthey all rolled out AI to their
people, but we are in apercentage.
There is still a long way to go, because what is now important
is we have great models outthere.
I mean they are fantastic, butnow it's important to find the
(06:43):
big problems to solve and makesure that those models can help
to really solve these bigproblems.
And there is still a lot to doand players like unique has a
huge market there to help largeorganizations on mid-size
organizations to use these largelanguage models in a way they
really solve real businessproblems and not just, you know,
(07:03):
and use chat shippity in ourprivate life.
Speaker 2 (07:05):
And how does Unique work?
Do you have your own languagemodels or you go out there, you
see everything that is availableand you pick up those that you
think are the best ones for thespecific tasks at hand.
Speaker 1 (07:18):
Absolutely.
We not build our own modelModels.
To build model is an own way.
You know, either you are veryvertical and you solve real
business problems or you'rehorizontal and build models like
open, ai and tropic and andalso, to be honest, we would not
have enough funding here ineurope to build such models.
When you look now, you knowwhat they raise to build models.
(07:40):
It's so much compute.
This is either america or china.
So what we do is we look to theprocesses on, like financial
institutions, where we see whatwe really could do better and
disrupt by using the.
You know the new way how youwork with an application, like
you chat the gans and you don'tuse these complicated screens
(08:03):
anymore, and then we look whichmodel performs best to make this
task.
This could be either an open AImodel or an open source model.
When we need to deploy themodel, even on a very secure way
to keep privacy and datasecurity high, because often our
clients they really want tohave everything in their private
(08:23):
cloud or even on-prem becauseit's very sensitive data.
They don't want that.
The model providers can learnfrom the data and potentially
train the next generation ofmodels, so that's why privacy is
a super important thing for usand depends on the requirements
we choose the right model.
And to make an own model makesno sense.
Yeah, and to make an own modelmakes no sense.
Speaker 2 (08:44):
Yeah, you were talking about disruption and
focusing on areas where you candisrupt.
What are the most disruptibleparts of the finance industry
that you feel AI can disrupt atthe moment?
Speaker 1 (09:05):
Yeah, you can go crazy, of course, and think you
build the next banker as an aiagent, but I believe still that
bankers or relationship managers, humans bring the empathy and
that you cannot replace by ai.
But what is our job is to makesure that these relationship
managers have much more time fortheir clients and we reduce
everything around, or let's say,supercharge everything around,
(09:27):
so that the bankers can do amuch better service to their
clients, because often arelationship manager spend too
much time to gather information,need to talk to various people
in the firm and it slows downthe customer service experience,
right?
So think of you have these allon fingertip and you can really
provide it to your clients inreal time.
So what we, our mission, isreally to make sure our clients,
(09:50):
the relationship managers, getmore client time and less admin
time.
That's the first phase.
The second phase, of course, isthen helping also to build
better products.
Help to, you know, in the end,drive alpha by choosing the
better products in the situationthe client is or the client has
needs.
So helping to build a better,let's say, offer for the client.
Speaker 2 (10:10):
But currently, I would say the next 12, 24 months
, it's all about efficiency,efficiency, efficiency, so that
relationship managers can domore sales and spend more time
with the clients I tend to agreewith the statement that RM's
jobs are going to be harder toreplace and everything related
to empathy is going to be moreshielded from AI disruption, but
(10:34):
I also think we're a little bitbiased by the state of AI today
, and we almost cannot imaginehow AI is going to look like 20,
30 years down the road, and youknow better than me that it's
moving crazy, crazy fast.
My question for you is why iseverything related to empathy so
(10:57):
hard for AI to touch as oftoday?
Speaker 1 (11:01):
Because it's human, so simple, you know.
I mean, if you spend money, youwant to talk to someone.
Even you might already haveyour decision.
You want to see someone, feelsomeone to give you feedback Is
this a good or bad decision?
And AI can give you all theinformation to maybe get to 90%,
but the last 10% you have stilla human advice in everything in
(11:25):
your life.
It's even with you could googleup what you, what you eat for
medicine, but you still love togo to a doctor to just get a
double check on it.
You know, and the same will bein finance.
That's why a relationship, apotentially the next generation
or so, has a more relation to anai agent.
But I believe this will not be,you know, something that get
(11:49):
disrupt.
What we see definitely is thatwhat happens.
Is that the biggest challengefor banks is that not that they
are slow in adapting, you know,against their competition, it's,
it's more that their endclients, especially the next
generation, let's say, wealthyclients that comes next are
using these tools, maybe evenmore efficient than the bank
(12:13):
itself.
So they are getting the clients, the end clients, getting much
more equipped by doing research.
You know, for a $200subscription, with perplexity.
You can already have a miniresearcher for you doing
investment research, and whatcould happen is that this
research is even better thansome banks can provide, you know
.
So that's why banks andadvisors need to really skill up
(12:37):
that you are outperform whatthe clients can do themselves
with using, you know, this kindof commodity tools out there,
and I guess that's what I see asa bigger threat for
relationship managers that theyget under pressure that their
clients becoming, let's say,more AI smart themselves.
Speaker 2 (12:58):
Which is probably a good thing, right?
Because it will push everyoneto move faster and talking about
more.
Speaker 1 (13:05):
You potentially get not replaced by AI, but you get
potentially replaced by anotherrelationship manager who uses AI
much more efficiently.
That could be definitely thebigger threat.
Because I get often asked is myjob safe?
And I say I think it doesn'tmatter if you're a relationship
manager, a designer, a coder,it's in all the professions out
(13:26):
there, in my opinion, you haveto skill up using this AI.
Let's say we said before hi,but I would say even using these
AI tools and skill up as fastas possible.
Speaker 2 (13:38):
Yeah, that sentence I mean.
All of you are going to gettired of hearing about it, but
it's definitely going to comeback over and over again over
the coming years which is, ai isnot going to replace you, but
someone using AI probably will,and hence why it's so important
to stay updated, educateyourself, learn how to use these
tools and become part of thechange, rather than staying on
(13:59):
the sidelines and watch how yourjob disappears or is taken over
by someone else or the machines.
But, yeah, we were talking aboutthe speed, the pace at which
this is happening and the factthat, yeah, eventually, clients
and retail investors couldbecome more familiar or have a
(14:23):
higher degree of expertise usingthese tools than some of the
advisors and relationshipmanagers and banks overall.
But I must say, in defense ofthe financial industry, that I
was really surprised by thespeed at which they have
embraced AI, because I remembera few years ago, everyone was
talking about compliance risksand how this would never happen
(14:46):
just because it was so dangerous, and what I've seen is that AI
adoption in finance is movingway faster than what any people
thought.
So what's actually drivingfinancial institutions to
embrace AI faster than whatpeople initially thought?
Speaker 1 (15:02):
Yeah, faster than what people initially thought,
yeah, yeah, and I can also tellyou why.
Because when you look tosoftware, software is
complicated, you know.
When you look to your corebanking systems or your CRM
systems, they are all difficultto learn, difficult to use,
because you have so many screens, fields, functions, buttons.
(15:22):
You have to learn to usesoftware, right.
But ChatGPT and we have to saythank you to them, kind of
disrupted is bringing a chatthat we all know from WhatsApp
in combination with these modelsand give you a new interface so
you speak with your languageand explain the problem to solve
and the solution is getting youknow more or less back in
(15:44):
normal language to you.
You know, and I think thatdisrupted the entire software
industry a little bit.
You know, I mean today I see iton my kids and so on.
I mean they even don't usegoogle anymore because you have
to to search too many screensright and searches.
So they use chat, gpt and justtalk.
I mean this is completely kindof flipped how we use software.
(16:08):
Even chat is not new Chat weknow and use all the time.
But the combination of a chatand large-language intelligent
model so let's say a chat andintelligence together was in the
end the disruptive point.
And now I believe, at least,that that's the reason why also
(16:28):
our clients.
You know that the adoption ofthe solution is so fast because
you don't need to train theusers, they understand it
immediately, because they canuse their language, they can
explain the problem.
Maybe the answer is not thebest and maybe they work a
little bit around, explain itanother way and, and that's
fantastic, right, when you thinkwhat you can do now, um, with
(16:53):
such a solution, but you wouldneed to build software in the
old way, you know, withdifferent screens and so on and
so on, that you will be neverfinished.
Speaker 2 (17:01):
You know, and this, I guess, is that the main driver
for adoption it's crazy what youwere saying about google,
because I completely relate tothat.
I was using google all the timeand now it's not that I don't
use it anymore, but I probablyuse it like 10 of the time.
Um, because chat gpt.
I'm a big fan of chatGPT, so itgoes one step ahead and I can
(17:26):
tailor the content or the outputthat I'm seeing to the specific
use case that I'm researchingfor and that makes the whole
difference right.
Like, if I'm organizing a tripto Lisbon, I just don't want to
get a list of the 10 best hotels.
I want to see them and get aprogram that is suitable for my
(17:47):
profile and my interests andhopefully tomorrow, with the
arrival of Agent TKI that youwere alluding to, also make the
bookings and all the necessaryarrangements in a trustworthy
fashion, I mean as they useChatGPT, they call it the
operator, what you could now tryout in Europe.
Speaker 1 (18:09):
I mean, this is fantastic, fantastic.
This operator is exactly doingwhat we do for our clients, but
for private life.
It really already book yourflight and book your hotel,
completely automated, and uh and.
And this is what we do forfinancial service industries,
but of course, in a much higher,let's say, enterprise level and
in a much more secure way,because, again, a chat GPT-like
application they also learn fromevery prompt from you, you know
(18:33):
, to then learn and train thenext generations of models, and
that's okay when you assign yourprivate data to it, but as a
financial service provider, ofcourse, you cannot.
Speaker 2 (18:44):
I have a question regarding the power struggle
between big financialinstitutions and fintechs,
because when we talk about AIand the future of the finance
industry and particularly, forexample, wealth management,
where I work some people thinkthat the big banks, the big
names, will dominate, becauseyou need resources to scale AI
(19:05):
across.
You know organizations, butothers believe that fintechs,
given that they're more agileand maybe more willing to take
risk and pushing boundaries,will adapt faster than what
banks could possibly do.
So who do you think is going towin this race?
(19:26):
Do you see traditional bankscoming out as the big winners,
or will fintechs grab some oftheir market share thanks to
their agility?
Speaker 1 (19:40):
Yeah, I personally believe that the mid-sized banks
have a right to win herebecause they can adapt the AI
much faster than maybe a largeorganization like UBS or
something like that, like aretail bank.
Because, again, it's not onlythe technology, it's how you do
(20:03):
the transformation that the userusing AI, that the employee is
using AI.
You know what I mean.
So, and as large yourorganization is, when you have
like a hundred thousandemployees and more, you know as
long as you take.
So that's why I believe themidsize banks that we also, by
the way, focus on on our clientbase, like we have a LGT, picte,
(20:26):
julius Baer and so on they, inmy opinion, can outperform large
organizations now.
But we also would see becausewhy Again I explained before you
can grow faster with hiringless people by implementing AI,
but a large organization needsto first bring costs down in a
(20:46):
larger amount, you know to tokind of keep the same cost
income ratio.
And of course, we also seefintech.
But I see more.
I believe more in the retail,um, you know, like insurance, uh
, like revolut, as an examplewhere you do you might saw the
announcement from RobinhoodPrivate Bank, yeah, but this is
(21:07):
still more in the retail, youknow, for smaller, for smaller,
let's say, private bankers butor private clients Out of
Involve Management, assetManagement in the private equity
.
I still believe it's this thesemid-sized companies will
potentially have the biggestopportunity I had Because, again
(21:29):
, you don't need any morethousands of people.
You need an awesome platform,like Unique AI, of course, and
then you need a small team thatreally builds these agents at
scale and you don't need anymore so much coder, because the
models build the code for you.
Right?
It's a complete new direction.
(21:50):
Some of our clients they havemaybe 10 people and build
awesome AI agents.
I mean, when you look to theoutput in, let's say, the
traditional way, you need maybea few hundred developers to get
the same improvements.
You know what I mean.
So it completely disrupts thething.
But, yeah, maybe we see alsosome new fintechs or new
(22:13):
companies that outperform bankswith having only a few hundred
or maybe a few hundred people.
But the only issue I see thereis what we discussed before why
our clients are so successfulbecause they have a trusted
brand, you know and trust.
You cannot accelerate with ai.
That's not possible.
You can be more efficient.
(22:34):
You can be more cheap, more,more efficient, but but you
cannot build trust faster withai.
Speaker 2 (22:39):
That doesn't work it's almost like sky is the
limit and the only limit reallyis your imagination.
Um, with all these tools thatyou know, give you kind of
superpowers and enhance yourproductivity.
I'm that's why I I love ai somuch.
It's, uh, it kind of levels upthe playing field and and you
(22:59):
can pretty much do almosteverything you want, which I
think is really, really cool.
I want to ask you about data,because AI obviously is only as
good as the data it learns fromand, as we all know it, big data
sets, particularly in thefinancial industry, are full of
biases.
How do we prevent AI fromunintentionally reinforcing
(23:23):
existing biases in financialdecision making?
And here's my promise If you do, I'll keep bringing you honest
(23:44):
conversations, freshperspectives and the kind of
finance talk that's engaging,insightful and worth your time.
Thanks so much for being hereand let's keep these great
finance conversations going.
Speaker 1 (23:57):
Good, you said it right.
I mean, the AI is only as goodyour data.
Of course, you can use somepublic information, but everyone
can use public information.
I think what makes thedifference is what you have,
let's say, in a way becausethat's also the culture,
potentially, or the direction asyour firm works right ways, how
(24:44):
you can ingest this private,unique data you have on hand in
a in a in, a, first of all,secure way, but also in a way
that most people in theorganization can benefit from.
Yeah, that's still a bigchallenge.
It's still a big challenge toget the enterprise, let's say,
set up that it's everything issecure and locked, and so on and
so on.
But also how you build usecases that they only use the
data they need and not have toomuch data or too less data,
(25:06):
because if it's too much data,then you have potential
hallucinations.
If it's not enough data, youdon't have enough accuracy on
the results.
And I always say this issomething every, every financial
institution need to try now andlearn uh, you know, and build
use cases and and, and you know,improve them and measure the
(25:30):
success.
And if the success is not there,stop the use case, make the
next use case.
There is no magic, you know.
So you really have to do yourhomework.
You have to figure out whichuse cases are, you know needs,
which data, and I think what wediscussed before most relevant
thing is that you not just havea better search of the
(25:51):
information, because that bringsyou certain efficiency, but it
will not change the game.
It's more like how you canreally build workflows or
replace, you know, human, boringwork by having workflows, using
these various data sets or datapoints to do a preparation of
(26:12):
your pitch deck or whatever.
I guess that's then somethingyou measure in hours of hours of
work improvements.
Right, and this ingestion ofthe data is still a challenge I
had in all our organization,because often the data is not up
to date, there is old, and youneed to tell the model hey, look
(26:34):
, if the data is not up to dateanymore, then you have to do
this and that you know that youthen have good results.
Speaker 2 (26:44):
Follow-up question.
Maybe the second derivativerelated to the biases in data
and the quality of data, whichis linked to responsibility.
So let's say, an AI modelmisjudges risk, approves a loan
(27:09):
to a wrong customer or justsomething that goes wrong on a
model.
So who's responsible?
The firm that deployed it?
The regulators who failed tooversee this kind of scenario?
The coders who build thelanguage model?
Who are we to blame ifsomething goes wrong with these
AI systems that we're deploying?
Speaker 1 (27:35):
Yeah, of course that's a difficult question
because in the end it's what isin the contracts between the
different parties you justmentioned.
But what I can just share, thatour clients always have a human
in the loop concept.
So that means, whatever oursolution produce, a human is in
charge.
The human take it either, sendit to a client you know, review
it, approve it.
(27:55):
We build tools around it tohelp the human to, let's say,
judge the hallucination scoreand this kind of stuff, but
still the human is in charge.
So we cannot, in financialservice, blame ai for wrong
decisions.
It's still the humans and Ithink that will keep over the
next, let's say, years.
Of course, you become moreconvinced that the results are
(28:17):
good.
You get a feeling.
You may not double checkeverything, but you always have
an eye on it.
And as more you use a feelingyou may not double check
everything, but you always havean eye on it and as more you use
these tools, you also know whatcould go wrong, where you might
need to double check.
But the responsibility is still, in my opinion, with the human,
like it was with all the toolsyou use in your daily life.
Right, um, if you go?
Because we at Unique, as anexample, we don't do solutions
(28:42):
where end clients directly.
You know, like use an AIadvisor as an example, we found
a banker between, because weknow it's very tough.
Still, hallucination can happen.
You know, as we discussedbefore, maybe your data sets are
not up to date.
Discussed before, maybe yourdata sets are not up to date,
(29:04):
you know.
So that's why a you, arelationship manager that
understands the business,understands the context, should
still look over the producedcontent or produced output that
you provide to your clients.
But then it's in the end, inyour policies and in your
governance who is responsible.
But I think it's hard to blamea model.
(29:25):
You know what I mean.
That will take a long way thatyou can blame a model.
Speaker 2 (29:32):
So I want to transition now to maybe the
second block in thisconversation, which is the part
focused on agentic AI that youbriefly touched upon at the
beginning of the conversation,topic you and I are very
passionate about.
So my first question is how faraway we are from a world where
(29:55):
AI is fully and autonomouslyrunning portfolios, making
lending decisions, setting riskstrategies and so on and so
forth, without human approval.
Speaker 1 (30:10):
Yeah, I mean the human approval, as I said before
, will be I'm still many, manyyears, but just on the results
but that agents can worktogether to at least optimize,
as you said before, portfoliosand strategies that that we are
very close to in in a very,let's say, easy way.
It may not that supercomplicated stuff but, um, I
(30:34):
would say in the next two, threeyears you could, you could
really see that these agentswork more automated together,
maybe not completely autonomous,but automated together.
I give you an example what wesee already today, what what
happens, um, in on our clientside.
So what, over the last 10 yearsor so, you know, all the
(30:57):
companies try to build huge datalakes of data and then try to
clean the data and so on and soon and in the end often not get
really good success because it'sso difficult to do.
But what happens now is youbuild agent that has just a
subset of this data that youpotentially can clean up, just
(31:19):
make ready faster because youcan.
It's maybe just let's call it asimplified like.
You have a relationship manageragent that have access to CRM
data and portfolio informationdata and then you have a KYC
agent that has access to KYCinformation.
So you don't need to harmonizea data set between the two
agents, because every agent hastheir own little data set.
(31:42):
And now what happens is that thetwo agents can speak to each
other.
So that means when therelationship manager needs to
know something about the client,let's say, their companies that
are involved the client couldask the KVC agent hey, I'm a
relationship manager X, I wantto know about the client epsilon
(32:03):
the following and that agentknows then what data first of
all, compliance from acompliance point of view you can
pass over to the relationshipmanager agent so we can solve
all this compliance stuff andwith that, you don't need to
build any more such huge datalakes.
Right, and you will be more andmore of these agents that they
(32:23):
can smartly talk to each other,make sure they are compliant to
each other, make sure they passon information that are approved
, and so on and so on.
And this, in my opinion, is thehuge, huge opportunity in the
future, when you have a lot ofdata in your house.
So these agents can worksmartly together.
They can have even contractwith each other what they can do
(32:44):
and whatnot.
Speaker 2 (32:46):
So let's assume you know those agents start
proliferating, Everyone isadopting them.
Do you think we'll reach apoint where those agents are so
widespread that they startcompeting against each other
very aggressively, to a pointwhere, for example, in financial
markets, it could create somesort of AI-driven volatility and
(33:08):
market chaos?
Speaker 1 (33:11):
It's a good.
I mean, as you know, I wouldsay never know and I'm sure we
saw it in other situations,situations even with uh simple
online platforms, how fast youcan uh drive such situations.
That's why I would say,potentially, absolutely, yes, um
, but again, that's not mybusiness.
(33:33):
That's not my business, that'sone for the dystopic writers?
I guess exactly but when I onlylook about my own experience I
had when I we just recentlyraised 30 million in the series
a run and I got so many fishingso professionally, you know,
(33:53):
with fully ai driven, that Ijust was like, wow, there we are
.
You know, know what I mean.
It's just crazy how AI alsogets used to, in the end, to
criminal influence, right, and Ican only measure that on the
phishing things that arehappening at our company right
now and it's really crazy whatyou see.
Speaker 2 (34:14):
Congrats on raising those funds, by the way, which
I'm sure they're going to helpyou loads to keep expanding and,
yeah, offering your servicesacross more financial
institutions.
And talking about that, can yougive us a little bit of color,
a bit of juicy details of whatare the big banks doing at the
(34:38):
moment when it comes toimplementing agentic ai?
What is the low hanging fruitthat, uh, if you had to say most
, uh, most of the big guys aretrying to tackle or doing with
with these tools?
Speaker 1 (34:50):
so I guess more or less every financial
institutions has now a so-calledprivate chat gpt or like
application that you can uselike you use chat GPT at home,
right?
So upload a few files,summarize, translate and so on.
But again, most of the clientsfigured out that that maybe give
you a few percentage efficiency.
What we see now where ourclients really double down when
(35:14):
they have this private chat GPTkind of an internal knowledge
thing sorted out, they go intoreal business problems and I see
three at the moment.
That is involving one isinvestment insight assistance
right to combine your houseviews with external data
(35:36):
providers to really help you todo simple I would call it simple
investment research basedagainst your house views and
your you know other opinions.
You have combined with dataproviders from external and
maybe earning calls and so on.
That's a heavy thing for uh,for wealth managers, asset
(35:57):
managers and so on.
Then, um, kyc we discussedbefore.
So kyc is a huge pain.
It's a huge, huge pain.
It's a huge cost driver.
Um, so we see clients startingusing a unique ai to really help
bankers to do maybe not replacethe entire kyc process, but you
(36:17):
know how it is.
You get these massive checklists.
You should get all thisinformation from a client.
You know how it is.
You get these massivechecklists.
You should get all thisinformation from a client.
You get all the files from aclient just to check.
Do I have everything?
Can I answer all the questions?
What is missing?
Please build me a list of themissing documents, missing
information, so really make theonboarding of a client more
efficient.
But also then the liabilityreports and everything you need
(36:41):
later on to keep koc on, on, onon the right level.
And the last thing is everythingaround due diligence.
Due diligence on massive datarecords, right, so you have your
q a list or your due diligencequestionnaire of hundreds of
questions for maybe your next icmeeting or your next proposal
(37:03):
discussion, and you have a dataroom of the documents and you
use also in external informationof of a target fund or of a
target company.
I mean, that's insane how gooddue diligence work today, so we
could really save this kind ofconsultancies that do simple
(37:23):
research on those targetcompanies.
Today you really can do thiswith AI and then you have a due
diligence report that is as goodthat you would hire 10 people
to do basic research for yousomewhere in in a remote
location out of those three qic,due diligence and investing
(37:45):
where do you think the biggestbang for the buck is?
Speaker 2 (37:47):
where do you estimate , according to your yeah, to
your studies, uh, that thebiggest productivity gains can
be achieved over over the short?
Speaker 1 (37:57):
term.
So short term is definitely inthe.
I mean, when you look to costsavings, it's KYC, because
that's painful, it costs a lotmore and more.
Client time also goes awaybecause bankers has to do a lot
in this process, right, um, soit's not only a cost, is also
(38:18):
missing revenue opportunitiesbecause people are busy doing
kvc instead of selling toclients or, you know, engaging
with clients.
Then revenue wise is in ininvestment insights, because you
can build more offers, morefaster, you can bring faster
opinion.
You know various offers to yourclients, you have more talk
time with your clients, you know, as more you do, as more you
(38:40):
sell, simple, right, so, um, buttoday I think again we need to
free up this time for bankers byusing ai to to make faster,
better insights and betterresearch and due diligence, I
would say is another game on duediligence is less efficiency,
but more.
You get more deals into thepipeline or through the pipeline
(39:01):
because you can look into moretarget clients, you know.
So you get potentially a muchbetter deal flow and potentially
you can increase your assetunder management because you get
more deals, you get betterdeals faster through.
So all of those have a littledifferent angles how to drive
(39:22):
efficiency.
Speaker 2 (39:24):
And talk about the investing vertical.
I tend to agree.
I think you know having AI isgonna help a lot when it comes
to financial advice, but I havea question about trust.
What will it take for peopleand clients to trust AI driven
(39:45):
financial advice the same waythey trust a human advisor?
Hey there, quick ad break.
Do you work in the financeindustry and have a genuinely
interesting story to share?
I'm always on the hunt forgreat guests who bring raw,
unfiltered insights to the table, or maybe you know someone with
a story worth telling.
Please put us in touch.
(40:05):
You can reach out to medirectly via LinkedIn.
I'd love to hear from you.
And now back to the show.
Speaker 1 (40:17):
As I said before, when the human is still in the
loop for the client, it doesn'tchange anything for the end
client right.
So it makes you as arelationship manager more
efficient and maybe you'll findbetter opportunities based on
the needs.
You can go more into customizedoffers than just selling
standard products.
(40:37):
So it just makes your, let'ssay, portfolio that you could
offer to your client morecustomized faster.
So I guess that is the maindriver.
But the trust from your clientsis still the same because, as
we discussed before, when wehave the human in the loop,
nothing changes in the end.
Speaker 2 (40:58):
And do you think that there's like some sort of moral
obligation from people using AIto tell or to share with the
end user that AI has beeninvolved in the process when
coming with a specific outcome,like a presentation, a trade or
you name it?
Speaker 1 (41:16):
I think there will be soon more disclaimers, as you
have disclaimers anywhere,everywhere on slides and
proposals that you potentiallysee a certain content is
generated by AI.
Definitely, we'll see that, butit doesn't really change the
game because in the end, theclient invests into the
(41:37):
instruments or into the productsyou know, and even banks using
AI since ever, you know, maybenot generative AI, but AI for,
let's say, machine learning,driven funds and so on.
I mean, this is there alreadysince 1020 or even longer, you
know.
So I guess that will not be abig difference, but you will see
(41:58):
more disclaimers.
And what we have to do more andmore now is, you know, every
company implementing AIgovernance, and often what we
have to do with clients is thatusers need to understand more
what the AI is doing.
You know, as an example, whatare hallucinations?
(42:21):
Is there a potentialhallucination in that answer?
Right, so we need to do trafficlight systems.
Hey, this is green because wedon't believe there is
hallucination.
Yellow because we think not allthe directives got used, you
know.
Or even red because someresources that are provided the
(42:41):
answer not fits to theseresource files, and then this
means we need to make awarenessto the users where we
potentially, from a technologypoint of view, see hallucination
, so the hallucination is thebiggest, or a low accuracy
because we not found enoughinformation.
As an example, we only foundtwo sources and not 10, and so
(43:01):
on and so on, and I think thatwill be the biggest thing over
the next few years how you canbuild trust to the users that
the ai system provide you goodinformation and that we, at the
moment that unique ai, we dothis with hallucination scoring,
so we really tell the users um,look, we predict your
(43:25):
hallucination or not in thetraffic light system and then if
it's orange or red, then theuser spend more time on that
content to figure out.
Hey, okay, now I double checkand see what could be wrong.
Speaker 2 (43:37):
Maybe it's wrong, but at least I got alerted that I
have to look here.
Yeah, so you are the get yourtake on the future, what's ahead
?
I think AI can be used as aforce for good, but I'm curious
(43:58):
to hear from your mouth what canI do today to make sure I stay
relevant tomorrow in anAI-driven world and financial
industry an AI-driven?
Speaker 1 (44:11):
world and financial industry.
Yeah, so, as we discussedbefore, really spend time using
this tool in business but alsoprivate right.
Learn how to prompt, and Ialways say the easiest way to
learn prompt is asking ChatGPTto write me a prompt based on
your problem because peopledon't know that.
So you can ask ChatGPT to writeyou a prompt based on your
(44:32):
problem because people don'tknow that.
So you can ask chat gpt towrite you a prompt template to
solve that problem and youexplain the problem.
Right, let's say you need to doan end year closing reports,
what these are my materials,what you would write for a
prompt, and then you learn, youknow how to to to prompt the
right way, what you give forinstruction, because, again,
(44:54):
prompting means you need to tellsomeone, like a human, what to
do and see it like the model islike a virtual human and you
need to give instruction, youneed to give some background
information.
You need to maybe do that andthen check with me again before
you go into the next step.
(45:14):
And I think that's all for free,accessible out there right by
just using chat gpt.
Or you watch some youtubevideos?
Um, this is a, let's say, themost easiest way to start, just
use the, the.
The free stuff is already outthere.
And then, of course, you can gointo more learning programs,
and there are thousands ofcourses out there to learn, but
(45:36):
you have to be active yeah,rolling up your sleeves and and
getting your hands dirty, as wesay, but honestly, like it's
super fun.
Speaker 2 (45:45):
Uh, there's no better way of learning than just
trying and iterating.
And uh, you'll notice in notime that you start becoming
better and better if you make itpart of your daily routines.
I mean, personally, I use AIall day long, both personally
and professionally.
I cannot live without itanymore.
(46:05):
So which, actually, I thinkit's the beauty and the curse of
AI.
On one side, it makes you lazyfor certain things because
you're like, ah, I don't want towrite this anymore.
I'm going to ask you know ChadGPD to write it for me.
And that's bad, I guess,because it makes you think less
sometimes about specific things.
But on the other hand, it opensup a world of opportunity, as
(46:29):
we were saying before, becauseyou can do so many more things.
There's not that technicalconstraint anymore that you used
to have, and so on.
An example I like to give alsois one day I was at work and I
was thinking could actuallysomeone create some code to do
(46:50):
something very simple?
It was like a macro on VBA, onExcel, to automate a report.
But I was thinking couldsomeone that has no clue about
coding, actually do this usingAI.
And I sat down in front of thecomputer saying, okay, I'm going
to write this as if I had noclue about coding and I'm going
to try to do the task.
And actually I managed.
You know, like, because the AIexperience you step by step Okay
(47:13):
, do this, do that.
This is how you write the codeeditor copy paste this code.
Then you had an error, you putit back and it would suggest
that a change, and so on and soforth.
So I was like, wow, this istruly mind blowing.
I mean, yeah, if you spend sometime, you can do almost
anything nowadays.
So it's pretty cool.
So let's pivot now to the lastpart of the conversation.
(47:35):
We started very wide, talkingabout AI in general.
We went a bit more specific,talking about agentic AI and the
wealth management industry, andnow I want to wrap it up
talking about you, because Ithink you have a very
interesting personal journey.
So, to start with, let's talkabout the fact that, yeah, you
(47:56):
built and sold companies beforewe talked about core systems
which SAP acquired, and I thinka lot of people would have
stayed in that corporate world,but you jumped back into the
startup, grind and built Unique.
But you jump back into thestartup grind and build unique.
So what?
Speaker 1 (48:20):
made you leave the corporate safety of SAP and go
back to the trenches of buildinga crazy startup.
Look, I'm a passionateentrepreneur and I always say
you are or not right, so I couldnever be employed.
I mean, if I have to because Ipromise to do a post-murture, of
course I do, I keep always mypromise.
But I have to because I promiseto do a post-merger, of course
I do, I keep always my promise.
But I want to build stuff, Iwant to invent.
You know the future, and it'seasier to do that outside of a
(48:44):
corporate.
I mean, look, I mean I love SAP, it's a great company.
But you can imagine in such alarge company if you have ideas
to get funding for these ideas.
But you can imagine in such alarge company if you have ideas
to get funding for these ideas,you have to be in the end an
asshole because you have to doother people.
You need to make other peoplebad because then you get their
budgets for you.
You know what I mean, becausethere's not more budget in the
company.
But if you do a startup, youcan raise money from investors
(49:10):
right, and go to the market,pitch your ideas.
You get funding.
You know you can make them toshareholders, so it's just more
fun and participate on yoursuccess.
And in large corporation thisis very tough to do.
It's doable I don't say it'snot doable but you have to be
political.
You know, very political activeto to get some other
departments budget for yourdepartment right, and I just
(49:33):
don't like to do that.
I love to build stuff and youknow, motivate people to follow
me and you know, help me to tobuild the next generation
software company.
And I never.
I mean why I did this, do thisthe third time?
Because you know I neverachieved an ipo and I learned
also in my previous company.
I was never confident enough tosay let's build a public
(49:56):
company.
You know that was also onelearning I had.
That's why I tell now toeveryone look unique will become
one day a public company.
Um, so, so a company thatpeople like to work for, people
like to partner with, but alsopeople like to invest.
Because then you know, youbuild a company that everyone
loves.
Right, and this is what isstill a dream, and keep me is a
(50:22):
little bit my North Star.
You know that, keep me alwaysgoing, going, going.
Maybe it works with Unique,maybe with the next one,
definitely not giving up.
Speaker 2 (50:31):
Nice.
I love that.
That's a very big, bold, butreally cool objective to have.
So you're selling AI tofinancial institutions for a
living, which is by no means aneasy feat.
So what's your secret?
How do you manage to breakthrough the barriers of all
those big bureaucraticcorporations that you were
(50:53):
learning to?
That would kill most fintechs.
Um, what's the secret sauce?
Speaker 1 (51:00):
look, it's very hard to say there is no, the secret
sauce it's.
It's in the end you need to bealways on and try and try.
You know you need to winclients by building relationship
, like you have to do as arelationship with clients.
You build trust to the firstproducts, but mainly to the team
(51:21):
you have.
So I would say, as a serialentrepreneur, I have a little
bit the luck.
I already worked with the teambefore in other startups so I
can bring with me a very strongteam, because usually the first
clients not look into only theproduct, they also look into the
team you know they work with.
(51:42):
And I think there we, we justhave a.
You know, one of the why uniqueis successful is we have the
best people and I think you know, especially the early clients,
they can really see that.
That's why they partner up withunique now, since unique is
already three, four years on themarket.
And we have, of course, nicebrands like pigtail and then lgt
and then julius bear, and youcan name them right uh, swiss
(52:06):
life and so on.
So it helps, of course, thatother brands trust because we
have leveled up uniqueside-by-side with these brands.
You have now this, let's saybrand trust.
But the first clients are superhard.
To win them as a success story,you have to be as I said before
.
You have to be always on withthese clients and get them
(52:29):
excited about your team and yourproduct.
It really helps them to buildand solve problems in the field.
Speaker 2 (52:38):
I love also that we have a Swiss success story
within the field of AI.
It's pretty cool, because we'reused to read about American
names and Chinese names and soon, but it's cool to have
someone close to us doing this.
So I feel like, of course, aistartups are popping up
(52:59):
everywhere and investors arethrowing money at anything that
says AI powered at the moment,the same way it was for, for
example, the Metaverse some timeago or esg before that.
Um, how do you make sure thatunique doesn't become just
another ai hype story?
Speaker 1 (53:21):
so what's the difference now between, let's
say, also blockchain and otherstuff not blockchain in the
currency world, but in the,let's say, more softer world is
that clients can really measurebenefits, you know, really
measure efficiency gains and theadoption in in a in a, you know
, and in a in a, let's say,efficient way, also to how much
(53:46):
they have to spend and investand and I think that's different
between other technology shifts.
We saw, you know, likeblockchain was cool from a
technical point of view but veryhard to implement and see a
real business benefit in shortterm.
Right, when you implement theGen AI solution, like your firm
running as an example, you seefirst time, when you use this
(54:10):
application, you really get like, wow, okay, already may save
one hour, two hour efficiency,but just maybe onboard yourself
into a company or whatever youdo right, you feel it very fast.
And I think that's differentthan all the technology hypes we
have seen in the past, becausein the end, you have to make a
business case.
You have to run a business case.
(54:30):
That you know we have to helpour clients to save 10, 50, 100
millions so that we get a fewmillions out of it.
You know so and I think that wealready were able to prove
already today and again, as Isaid before, we had just started
, we're just in the firstpercentage.
So that's why I believe ai has aright to stay and especially,
(54:54):
you said it right yes, we hear alot from us vendors or chinese
vendors, and that's good so,because us always were in that
position.
You know they always do greatfoundation stuff, like cloud
services, like azure or aws andso on, and what is our
opportunity here in switzerlandis building now vertical ai
(55:15):
solution on top and partner upwith these american models.
Right, because the models wecannot build.
It's like back to the microsofttimes.
You know you always used theirwindows or whatever, let's say,
as the operating system.
But look to SAP.
They built an awesome verticalsolution for ERP and I think
(55:35):
that that is a right to win herein Switzerland or in Europe, to
focus on the vertical, tosolving real client problems.
I would not invest into a more.
I personally would not investin a horizontal solution in
Europe because, again, as I saidbefore, it's very hard to
finance because the largeinvestors are still in USA and
(55:56):
they usually love to invest inUSA, and for a vertical solution
, you need less money becauseyou don't need the large
infrastructure to train themodel.
You need more client time andyou need to be sure that you
learn from your clients and thatis your mode against the
vertical players, the horizontalplayers that you know the
(56:19):
client problems better.
You know the client data betterthan they do.
And, yes, I agree, it should bea little bit more unique in
Switzerland.
But we have other AI players nowpopping up and I hope this is
now a momentum in Zurich, inLausanne, around EPFL ETH, that
we really take this opportunityand also have our own, let's say
(56:43):
, ai supply in a way for ourclients.
I would love a little bit moreour Swiss corporations to trust
Swiss AI startups that stillmost of these managers love to
run after Microsoft or AWSbecause you know you do nothing
wrong.
That is something what Isometimes miss a little bit how
(57:05):
complicated some corporationsare to trust a Swiss startup to
go in a larger contract.
If we could overcome that also,then I believe we have a really
huge opportunity on working onour own AI supply here in
Switzerland.
Speaker 2 (57:23):
You mentioned Zurich, you mentioned Lausanne, but of
course also Geneva, absolutely Ijust mentioned.
Speaker 1 (57:30):
Hopefully we see some interesting companies popping
up In Borden.
I mean it's entire Switzerland.
Speaker 2 (57:38):
Maybe our last question on you and your journey
Not an easy one, becauserunning a company isn't all wins
.
You are the CEO, and CEOs haveto make brutal calls, sometimes
about obviously firing aboutpivoting strategies, walking
(57:58):
away from a big deal, thingslike that.
What's been the toughestleadership decision you've ever
had to make and what did youlearn from it?
Speaker 1 (58:08):
Wow, I mean there's tons of right.
Wow, I mean there's tons ofright.
But, of course, what is alwaysvery hard for, I guess, every
founder out there when you startto realize you missed your ICP,
your ideal client profile, andyou just have to let people go
because you had a belief insomething and it doesn't work
out.
And I had that at least threeto four times where I needed to
(58:29):
let a bigger group of group ofpeople just say sorry, it didn't
work out, you have to let themgo.
And this is always veryemotional.
But what I always say there is,when you get to that point, what
most startup do wrong.
They start to do small steps,you know, say, oh, let three
people go, then four and five,and that with that usually you
damage even more.
You should go then and say,okay, now we, we, we should let
(58:51):
10, 20 people go.
So, because this gives you airto breathe, to find another
direction, right, and that's myadvice to founders when you are
in that situation, then youshould better let more people go
, because you need the cash.
The cash burn is so hardbecause, you see, always the
(59:12):
cash burn is burning, burning,burning, burning and you don't
have new revenue to pay thesalaries.
Yeah, then you do it like inslot in in slices, because that
kills the culture, that killsyour brain, your energy and and
these are very hard decisionsand I did this many times and
it's not nice but it's neededit's the only way how you steer
(59:34):
the boat in another direction.
Speaker 2 (59:36):
Well, thanks for sharing those valuable insights,
manuel.
It's been absolutely amazinghaving this conversation with
you.
Thank you for sharing all theseinsights For the listeners that
want to hear more about Uniqueand what you guys are up to.
Where can they find moreinformation?
Speaker 1 (59:54):
So the easiest way is you go to uniqueai or on our
LinkedIn page.
You get always updates.
Feel free to comment there.
And we also share much morelearnings now from our financial
service clients.
We also do more and more eventsnow where we bring the
community together.
We just had recently one at sixwith over 200 financial experts
(01:00:18):
that showed up, and you know wehad a panel discussion about
the Gentic AI with some thoughtleaders and it's you know.
I see this as our mission toreally help the entire industry
to grow up in the world ofauthentic AI.
Speaker 2 (01:00:35):
Well, that's a great mission to have.
I'm sure you're going toabsolutely crush it.
I'll be front row to watch allof those successes.
I'm very excited for you guys.
Manuel, thank you so much fortaking the time of talking to me
today, and I hope 2025 is theyear where that IPO, that famous
IPO you're targeting, gets doneor at least gets a little bit
(01:00:56):
closer.
Speaker 1 (01:00:58):
Yeah, definitely a little bit more closer done and
still a way to go.
But I will definitely tell youwhen it's the right moment.
Thank you, all right.
Thank you very much for thetime and a nice rest of the day.
Speaker 2 (01:01:14):
The Blunt Dollar is written, produced, hosted and
edited by me, ignacio Ramirez.
Everything you hear concept,script, sound design and
production comes straight frommy desk and, occasionally, my
kitchen table.
Thank you so much for listeningand join me in the next episode
of the Blunt Dollar for moreraw, honest finance
conversations.
Popular Podcasts
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.
Dateline NBC
Current and classic episodes, featuring compelling true-crime mysteries, powerful documentaries and in-depth investigations. Special Summer Offer: Exclusively on Apple Podcasts, try our Dateline Premium subscription completely free for one month! With Dateline Premium, you get every episode ad-free plus exclusive bonus content.
On Purpose with Jay Shetty
I’m Jay Shetty host of On Purpose the worlds #1 Mental Health podcast and I’m so grateful you found us. I started this podcast 5 years ago to invite you into conversations and workshops that are designed to help make you happier, healthier and more healed. I believe that when you (yes you) feel seen, heard and understood you’re able to deal with relationship struggles, work challenges and life’s ups and downs with more ease and grace. I interview experts, celebrities, thought leaders and athletes so that we can grow our mindset, build better habits and uncover a side of them we’ve never seen before. New episodes every Monday and Friday. Your support means the world to me and I don’t take it for granted — click the follow button and leave a review to help us spread the love with On Purpose. I can’t wait for you to listen to your first or 500th episode!