July 25, 2025 • 40 mins
In this episode, Chris sits down with Igor Nikitin, CEO and co-founder of Nice Technologies, to explore how AI and modern engineering practices are transforming the actuarial field and setting the stage for the future of actuarial modeling. We discuss the introduction of programming into insurance pricing workflows, and how their Python-based calc engine, AI copilots, and DevOps-inspired workflows are enabling actuaries to collaborate more effectively across teams while accelerating innovation.
Featuring:
Links:
Sponsors:
- Shopify – The commerce platform trusted by millions. From idea to checkout, Shopify gives you everything you need to launch and scale your business—no matter your level of experience. Build beautiful storefronts, market with built-in AI tools, and tap into the platform powering 10% of all U.S. eCommerce. Start your one-dollar trial at shopify.com/practicalai
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Jerod (00:04):
Welcome to the Practical AI podcast, where we break down
the real world applications ofartificial intelligence and how
it's shaping the way we live,work, and create. Our goal is to
help make AI technologypractical, productive, and
accessible to everyone. Whetheryou're a developer, business
leader, or just curious aboutthe tech behind the buzz, you're
(00:24):
in the right place. Be sure toconnect with us on LinkedIn, X,
or Blue Sky to stay up to datewith episode drops, behind the
scenes content, and AI insights.You can learn more at
practicalai.fm.
Now, onto the show.
Chris (00:48):
Welcome to another episode of the Practical AI
podcast. My name is ChrisBenson, your co host. And with
me today, we have Igor Nikotin,who is the CEO and co founder of
NICE Technologies. Welcome tothe show, Igor.
Igor (01:02):
Thank you, Chris.
Chris (01:04):
So one of the things that we love doing on the show is to
get into different industriesand explore how different
industries are implementing andusing AI in different ways. And
you are an actuary, or at leastthe CEO of an actuarial modeling
company. And that, we may havecertain folks in the audience
(01:25):
who are not intimately familiarwith your industry, and I was
wondering if we could startwith, kinda tell us a little bit
about what your industry is, aswell as kind of how you got into
it, and then kind of into whythe technology around it
specifically?
Igor (01:39):
Sure. So Actuary, not to be confused with ACTR, is a risk
management professional. We workat insurance companies, and we
are the people who figure outhow much insurance costs, what
are the terms of insurance, andvarious aspects of managing
insurance companies. So if yougo to any insurance company,
(02:01):
people there would tell you thatit is really run by actuaries in
in that company.
Chris (02:06):
You're the folks with the power in it there. You're you're
the ones that are actuallymaking the wheels turn.
Igor (02:11):
Well, us and sales. Because people don't wake up in
the morning thinking, Oh, youknow what, today I'm gonna buy
myself some life insurance.
Chris (02:21):
That's good. Given the fact that it's risk management
of this, could you talk a littlebit about what is involved? If
your job is to wake up, you'rean actuary, and you're going to
work at an insurance company,what does that look like? What
is it that you're doing for theinsurance company to do, like
how do you implement riskmanagement? What are the
(02:43):
traditional tools, aside fromsome of the technologies that
we're about to get into withyou?
What does that world look like?
Igor (02:50):
It usually looks like a lot of Excel spreadsheets. And
for various items, there aremore sophisticated models. So
you may see proprietary softwarebeing used for certain things
like projecting benefits. Youmay see some software
specialized on the projecting ofyour investments and on your And
(03:14):
in general, your day looks likedoing a lot of financial
modeling to answer questions orto satisfy various regulations
or to develop new products. Butin the end of the day, most of
what actuaries do are some formof modeling or regulation
(03:34):
research or compliance.
Chris (03:36):
Gotcha, so that is the wheels behind the scene in that
industry that makes it turn. SoI guess that when an insurance
company, for instance, is goingto price its products, know, the
insurance that a consumer mightbuy, I assume that all the
aspects that go into that arepart of what the actuary figures
out in terms of like, they can'tprice it without knowing all the
(03:58):
different types of risks and howthose play into the model. And
so I'm curious, how did you findyourself? Did you start yourself
as an actuary and then go intotechnology? Or did you come from
the other side as a technologistinto the space?
How did you find yourself inthis field?
Igor (04:17):
It was actually quite an interesting story. So I was
programming since I was the ageof 13 because I told my parents
that I must have computer tolearn programming. That was a
ruse. I just wanted to playvideo games like all the other
cool kids. But I think they getthe last laugh because here I am
a software engineer.
(04:38):
So it was a hobby of mine. ThenI moved to US and I kind of lost
my circle of friends who wereinto programming. And so
thinking what I want to do, myparents were telling me, well,
maybe you should be a lawyer. SoI looked up, okay, to be a
lawyer, what major do I need?And it was English.
(05:01):
I did one semester of English. Ihated it with lots of passion
because somehow there is nowrong answer, but all of my
answers were somehow wrong. Andso I did a bit more soul
searching and I decided that Iwant to become a high school
teacher for mathematics. At theend of my college, I taught for
(05:22):
one semester in American highschool and I decided that I need
to switch a profession. So nextthing I know, I'm sitting in a
coffee shop Googling what to dowith a mathematics degree.
And that's when I discoveredactuaries. And I said, oh, okay.
Interesting. So there are someexams to take. I took my first
exam.
(05:43):
It was probability. And I waslike, oh, probability. I'm great
at probability. So I went in. Igot two out of 10 on that exam.
That is not a good grade. And Iwas like, Oh, you know, that is
actually challenging enough of aprofession for me. I like it. So
how do people pass this thing?It is absolutely brutal.
(06:04):
Turns out you get a manual thatis about thousand pages sick
with about 2,000 practiceproblems. You solve through all
of that and then you show up toexamine, it's easy at that
point. So I did that. So Iapplied to a bunch of entry
level actuarial jobs, but thatwas in January. And some of you
may remember that there was afinancial crisis right at about
(06:26):
that time
Chris (06:27):
when some Yes. People got laid
Igor (06:30):
So I discovered that getting a job as an actuary was
very, very difficult as an entrylevel actuary. And then I said,
Okay, what else do I know how todo programming? So I applied to
a bunch of insurance companieswith my resume for both actuary
and programming jobs. And I gotinvited by Prudential Financial
(06:52):
actually for an interview for aprogramming position. On my
resume, it said, I haveintroduction to C plus plus as
the only programming course.
And then I also brought aportfolio of things I've done.
And they said, Okay, you havedone a whole lot of stuff. And
so they hired me as aprogrammer. Pretty quickly,
realized that as an actuaryfocusing on modeling, I can
(07:14):
continue to do programming butget paid twice as much. And so I
joined their actual leadershipdevelopment program and rotated
through a bunch of differentdepartments, learning how
insurance works, won aninnovation competition along the
way there with the idea of kindof future of modeling, what does
(07:34):
it look like, and what should afirm do to prepare for that?
So that's how I ended up beingan actuary.
Chris (07:42):
Gotcha. So I'm curious, while the current age of AI had
kicked off again by that time,it was really before anybody was
doing it. So I'm kinda curiousas you were looking at models of
the day that you're actuallyworking with in real life at the
job, just to draw a distinction,you know, as we as we talk of
(08:03):
modeling over this past decadeplus. What were you typically
working with? What kinds ofmodels, what were the
algorithmic things that youcared about in your day job when
you were doing that, as youmoved into software development,
but within that actuarialmodeling context?
Igor (08:19):
So coming from a software background, I was actually very
surprised by a lot of thingsthat I've seen. Because quite
often models, instead of beingthis kind of software that is
very rigorously tested andoptimized, you discover that
there is a lot of qualitycontrols that is commonplace in
(08:41):
the software industry that justdoesn't exist. Part of it is
because how do you versioncontrol an Excel workbook? And
that is very challenging. Butcoming from background of using
Git and DevOps, you're like,Well, there is a way.
We just need to get away fromExcel somehow. But that runs
(09:02):
into a different problem ofthere was no programming
training for actuaries. Therewas no required programming
training for actuaries prior to2018. In 2018, there was
actually programming added tothe actuarial exams. And since
then, the profession becameoverall much more knowledgeable
(09:23):
on programming and programmingtools in general.
Because the situation was kindof interesting that we all do
modeling, but none of us hadsome sort of like very hands on
modeling course that is verykind of operations based as
opposed to understanding thetheory behind insurance and all
(09:45):
of the probability calculationsand things.
Chris (09:47):
I'm curious, is the field in 2018 updated itself, as you
said, and you started seeingprogramming as a requirement,
how did that lead into, now thatit's kind of recognized as part
of the field, and over the yearssince 2018, not too long here,
(10:08):
seven years, how has thataffected the field itself in
terms of getting up to the kindsof things that you're doing at
this point where you're havingAI capabilities and kind of
modern algorithms directlyimplemented for actuarial
fields? What did that look like,that evolution? Because that's
(10:30):
not a lot of time between 2018and 2025 as we record this.
Igor (10:35):
It is actually very different. And it was actually
driven not so much bytechnology, but by regulation.
Regulations became much moresophisticated and that kind of
forced everybody to use bettertechnology. Because if you need
to run stochastic runs withthousands of scenarios, you
(10:55):
cannot possibly do that inExcel. It's going to run for
weeks.
And so that forced the industryto upgrade their systems and to
adopt quite a few of thepractices that are much more
typical in software development.So seeing things like DevOps and
(11:16):
unit tests, it became much morecommon nowadays. It's still not
on every insurance company. Andit really depends kind of what
insurance company how insurancecompany operates. Right?
Because some of them, they focusjust on sales and they kind of
outsource all of the productdevelopment to other companies.
And the others are much morefocused on kind of the mechanics
(11:37):
of producing the insurance andthen letting other companies
care about sales. As far as AI,there is still very little
adoption of it. Chief problem isthe fact that AI is not reliable
in a sense that it doesn'tnecessarily give you the truth.
And when you're complying withregulations, you do have to have
(11:59):
the truth.
So there is quite a lot ofexperiments of what is it useful
for in summarizing the data.There is some experiments in
model development assistance.And yeah, I would say and in
(12:21):
providing a second opinion tothe data analysis that is being
done. So I would say those arethe three areas where we see AI
actually having some real worldvalue that we observe.
Sponsor (12:53):
Well, friends, when you're building and shipping AI
products at scale, there's oneconstant, complexity. Yes.
You're wrangling models, datapipelines, deployment
infrastructure, and then someonesays, let's turn this into a
business. Cue the chaos. That'swhere Shopify steps in whether
you're spinning up a storefrontfor your AI powered app or
(13:14):
launching a brand around thetools you built.
Shopify is the commerce platformtrusted by millions of
businesses and 10% of all USecommerce from names like
Mattel, Gymshark to foundersjust like you. With literally
hundreds of ready to usetemplates, powerful built in
marketing tools, and AI thatwrites product descriptions for
(13:34):
you, headlines, even polishesyour product photography.
Shopify doesn't just get youselling, it makes you look good
doing it. And we love it. We useit here at Changelog.
Check us outmerch.changelog.com. That's our
and it handles the heavy liftingtoo. Payments, inventory,
returns, shipping, even globallogistics. It's like having an
(13:57):
ops team built into your stackto help you sell. So if you're
ready to sell, you are ready forShopify.
Sign up now for your $1 permonth trial and start selling
today atshopify.com/practicalai. Again,
that is shopify.com/practicalai.
Chris (14:21):
So as you got to this point in your career where where
technology is changing, youknow, and it's evolving, How did
you get from that point intowhere you're at now with NICE
technologies and and thecapabilities that you provide?
What was the the remaining partof the journey into as you were
kind of getting closer to themodern point?
Igor (14:43):
It really was born out of necessity. So our platform is
kind of two pieces. There'scalculation engine and data, and
those are kind of two majorparts. The data management part
was really born at 2AM as I wasworking on pricing reinsurance
(15:04):
deals. And in that job, you dealwith a lot of reruns and with a
lot of tool chains that arerather manual.
So new piece of informationcomes in for a meeting that you
have tomorrow at 9AM. This pieceof information comes in at 5PM
the day before. And then thetool chain to incorporate this
(15:30):
piece of information into thefinal quote can take four or
five hours of differentcombinations of Excel models,
Python models, differentproprietary models. One reason
for that is that reinsurancepricing is very bespoke. And so
it's difficult to automate aprocess that is bespoke.
(15:51):
And because of that, you end upactually sitting in the office
until you're done incorporatingthis and checking this. And then
you send your boss a message atlike 2AM saying, hey, it's done.
Can you review it by 9AM? Andthey would actually be there
knowing that they do need toreview it because there's a lot
(16:11):
of money, a lot of eyes on thequote and a lot of
responsibility. So that's when Icame up with the idea of, all
right, how there must be somesort of engineering solution to
this.
And the solution is really whatif you make your tool chain
remember the steps you took inyour original pricing. Right?
(16:33):
Because if you have that and itcan recreate those steps without
making mistakes that humans tendto make when they work at 2AM
for the fifth night in a row.That's where that idea was born.
The other part of it was I'vedone modeling in several
(16:54):
different systems.
And I realized that for certainproducts, some systems are far
easier and quicker and cheaperto use than others. The
challenge there is what if thecheapest system and the best
system to use for a givenproduct is an in house system?
And that's where I took where Irealized that we can take a page
(17:16):
from the video game developmentindustry. So in video game
development industry in the'90s, they had the same
problems. They had a bunch ofproprietary software.
As complexity of games took offwith compute power, you said,
Oh, three d graphics are great.Your development costs also
skyrocketed, but you can't sellvideo games more expensively
because your customer refuses topay more than $50 So how do you
(17:40):
solve it? Well, Unreal Engineand Unity, right? They basically
provide a core functionalitythat everybody builds off of to
build the actual video games.And that makes the process much
cheaper, and hence, you canactually have games that cost
$50 as opposed to $5,000 So wesaid, okay, can we take that
similar concept and apply it toinsurance?
(18:01):
So we would need a software thatprovides access to the
underlying code, but at the sametime, we need to make sure that
it is commercially viable andwell supported and industrial
grade. And that's how we came upwith idea for our calc engine,
which is source availablebusiness model. Meaning we
(18:22):
provide source code to ourclients, but not to general
public.
Chris (18:25):
Gotcha. So it's interesting that I I love you're
kind of there's a little bit ofkind of a classic
entrepreneurial story there, andthat you're a, kind of
scratching your own itch. Like,you're recognizing my job would
be easier if I were to recreatewhat I was doing before and have
that immediately available. Andthen that's interesting
(18:48):
inspiration. Would not I don'tthink I would have expected you
to the analogy between theactuary in field and also gaming
is pretty cool.
So it sounds like you'rebuilding an app kind of in the
same spirit as you mentioned,Unreal Engine and Unity, which
are the two biggies in thatindustry. So in the end, could
(19:10):
you talk a little bit about whatyour calc engine became and how
it was received in the industry?Like what you know, I know, a
sense, Daniel, who is my cohost,who is not able to make it here
today, is also the CEO of asmall company that's growing
very fast. And I love hearingkind of how he's gotten through
the steps of how his company hasdeveloped from kind of the the
(19:33):
challenges at the verybeginning, and and he gets
traction and stuff. Could youtalk a little bit about what
your what your pathway has beenlike as you've been developing
the the calc engine and theother features at NICE
Industries?
What has that pathway lookedlike for you?
Igor (19:48):
So we we always leaned on feedback. Even before we started
a company, we we did went toseveral executives in the
industry and said, would this beinteresting? You know? We went
to heads of pricing, heads ofmodeling, and said, you know,
what if this thing existed?Would you care?
And the answer was we would loveto see a demo. Pretty much every
(20:11):
single person we spoke withsaid, we would love to see a
demo. We said, okay, great. Butto build a demo, we need like a
year of work because this ispretty complicated thing to to
build if you want a demo thatlooks somewhat realistic. So
with that knowledge, we said,okay, well, let's let's see if
(20:32):
anybody else believes in this.
We reached out to several peoplethat we wanted to cofound a
company with, and only two ofthem did not join. One of them
did not join because she had ababy five days before we called
her. So that was understandable.
Chris (20:48):
That's a pretty good reason for anything, quite
honestly, in the world.
Igor (20:51):
But the others joined, and that was a very defining moment
of saying, Oh, okay, I'm not theonly person who believes in
this. And then we went toinvestors and we said, Well,
would you be willing to investin this kind of venture? And we
discovered that the answer wasalso yes. So we said, okay,
(21:11):
well, apparently there's quite afew people who are interested.
So then we developed a demo.
We showed it to the originalgroup of people that we were
working with. They said, okay,great. When is it going to be
commercially available? We said,well, probably for another year
or so because there's a lot ofwork to go from prototype to a
(21:32):
production grade system. Andthen we launched publicly in
this February with a productiongrade system.
And right now we're in talkswith a lot of different
insurance companies in variousstages of negotiations.
Chris (21:49):
So it sounds like right now you're kind of the hot new
thing in that field. Because ifyou just came out a few months
ago, you've got this great newthing that's gonna revolutionize
the industry with a capabilityit didn't have before. I'm just
curious, as an industry thathasn't had this kind of
capability in the past, what'sthe reception like on that? I
(22:10):
see in a lot of industries thatwe'll talk to people who, you
know, there's a bit of astruggle to shift mindset into
recognizing it, and then thereare other industries where
people are just waiting for theleap forward and stuff. Is it,
or some combination of the two,what is it like changing an
industry, you know, in that wayat such a fundamental level as
(22:31):
you're trying to get traction onthat and bring in for you as a
as a business owner, bringingnew customers on board, but kind
of giving them a new way of ofdoing that?
Igor (22:41):
It's a lot of explaining how we're different. Because on
one hand, people quickly catchon that our business model is
completely different fromeverybody else in the market. On
the other hand, the most commonquestion I get is, do I require
to know Python to use yoursystem? The answer is no. But
that is like the number onequestion the most common
(23:03):
question that I get.
And so it is a bit of a process,especially at big companies
where there are manystakeholders, a lot of different
vetting processes, and a lot ofcoordinating between a lot of
parties. So you need to havepricing on board and valuation
on board and modeling on boardand IT on board. And it just
(23:25):
takes a lot of time tocoordinate everything between
those. Just like life insurancepolicies, insurance companies
usually don't wake up in themorning and say, Oh, you know
what? We're going to swap ourmodeling platform.
That's a big undertaking, and itis not taken lightly. And having
(23:45):
been in those shoes, I realizethat there's a lot of work and
you would only do it for very,very compelling business
reasons.
Chris (23:53):
It sounds I I like, when if you look at like, speaking
about one of your customers,that they're really it's really
to to get the benefits and toleap forward, they they have
their workflows, which is and asyou kinda pointed out, the
different components of theirbusiness, the different
departments, you will, have toall kinda get on board. It
(24:14):
reminds me a little bit as yousay that, and I don't know that
this is a good analogy, so I'llit out of kinda when you put in
like like an SAP implementationin a in a company, and it hits
many different functions of it.It seems to have where the
workflows themselves at thosethe at the customers will
change. It sounds like differentsoftware, but within your
(24:36):
industry, like you're, there's abunch of plugin points where
people have to buy in andworkflows change so that they
can get the benefits of surgingforward. Is that an accurate way
of representing that?
Igor (24:48):
Oh, yeah, it is somewhat similar. I would agree that it
does change the way people workfor some people more than
others. So if you look at itfrom a model developer
perspective, you used to useproprietary software where you
had a bunch of dropdowns whereyou could edit some formulas.
(25:09):
Fun fact, those formulas looksuspiciously similar to code.
And now you use something likeVisual Studio, for example.
Now, looking at it fromperspective of an actuary, you
would say, okay, that's a bigdifference. Now I have to
understand Python. But lookingat it from kind of an actuary
who also has experience inprogramming, I see it as, oh, I
(25:33):
finally have Git for versioncontrol. I finally have DevOps.
I finally can automaticallymerge code from multiple
developers.
I finally have amazingcapabilities and breakpoints and
a lot of these tools that I didnot have before that allow me to
(25:54):
understand things quicker. Andin today's world, that also
means that you can have an AIcode assistance that can explain
how things work, that can makevariants of different
calculations, that can createunit tests for you, and do all
of those things actually quitewell.
Chris (26:14):
So that is pretty fascinating to me in terms of
the description. As you'retalking about tools like Git and
Visual Studio and stuff likethat, and those are things that
I obviously know coming at it asa developer, Could you talk a
little bit about recognizingthat not every actuary is a
developer and everything, couldyou do a deep dive into kind of
(26:37):
what your workflow looks like?Like from the customer's
perspective, what does that turntheir reality into? And can you
draw out some of those pointsabout why this is better than it
was before by doing that?Because we've talked a bit about
the investment in learning newstuff, the investment in getting
the different departments of theinsurance company working
(26:59):
together.
Take me through it, and if youcould draw out some of the
benefits of the technology andspecifically the AI bits that
you have into it, I'd love tounderstand what that looks like.
What does a modern actuarialworkflow look like for a company
that's now out on the leadingedge of that?
Igor (27:19):
Sure, so a simple example would be, let's say you have an
Excel model that does something.Right? So imagine a dozen or so
tabs full of Excel formulas, andthat does some sort of
calculation. Right? So thenumber one question you would
get is, well, how does it work?
If I want to understand how thisworks, I would spend some time
(27:42):
digging through all of theseformulas and tracing and kinda
understanding how they allconnect. Now in the in our
platform, right, you literallyjust ask the AI system, hey, can
you explain this how thiscalculation works? Can you point
me to where in the code is it?And you get an exact location in
the code and the calculation.Right?
(28:05):
Then you say, okay. Well, I needto modify this model to support
a new product. And there ismaybe several pieces of that
product that would need to bemodified. In Excel, it's very
difficult to merge models. So ifyou have three people working in
parallel, it's very difficult toput all of their changes
together and make sure that youdidn't miss anything.
(28:28):
Versus if you work in Python,you have Git. You just say, hey,
can you merge this? And if yourplatform is well organized,
meaning it's object oriented,you would have very little
problems with that. And there'sgood chance it'll just auto
merge in seconds and say, okay,there is no conflicts. So it
(28:50):
enables development offunctionalities as a team as
opposed to being forced to be ina single developer because
working in parallel is simplytoo complicated to merge.
And then you say, okay, so now Ihave this functionality. It
kinda sorta works. I wanna makesure that it's production grade,
so I need to test it. In Excel,you have a limited number of
(29:15):
tests that you would hopefullycome up with. I've seen dozens
usually being kind of a decentnumber.
With the help of AI, you cancreate a lot more than that. And
with unit test functionality inVisual Studio or DevOps, You can
have literally thousands or evenmillions of unit tests to make
(29:38):
sure that your software doesexactly what you want and can
handle all of the edge casescorrectly. And then the next
step would be to say, okay,well, it works in Excel, but
this is glacially slow. What canI do? And at that point, you're
quite often just stuck becausethere's simply no way of getting
(30:00):
it faster.
Now, in our world, you can askyour AI assistant, Hey, can you
check my code for Well, first ofall, you have the performance
benchmarks and those would gettriggered if you've done
something that tripled theruntime all of a sudden. But
second, you can ask AI assistantto say, hey, can you tell me
(30:22):
which part of my code is slowand can you give me some ideas
of how to do it quicker? Andquite often you will get, at the
very least, will identify whereyou're doing same calculation
over and over or doing otherthings that are suboptimal.
Overall, it's just much nicerexperience, which requires some
(30:46):
knowledge of Python to beeffective at it from modeling
perspective. But at the sametime, with AI systems, that
threshold of how much Python youneed to know is much lower.
As well as it is just so muchfaster to ask AI assistant for
like, Hey, can you explainsomething? Or can you find
(31:07):
something in the code base?
Chris (31:08):
It sounds like, just as an analogy on another topic that
we brought in, common buzzphrase these days is vibe
programming, vibe development,where you're working with a LLM
on your code, and that it soundslike there's sort of an analogy
in that the actuary is going toknow some Python to use the
(31:29):
system, but the system helpswith that. Is that is that a
fair assessment to where thesystem itself is kind of working
in that vibe coding manner ofsaying, I can show you the
challenge in the code that youneed to address or and maybe I
can help you fix it and thingslike that. What are can you
differentiate a little bit aboutwhich parts of the workflow the
(31:52):
actuary is getting into thePython and what kinds of things
are able to be done through aGUI or through an agentic
approach, for instance, whereyou have an agent that may have
taken on a task that washistorically a very manual task
or anything like that. Do youhave anything that you can share
along those lines just in termsof how workflows evolve?
Igor (32:13):
Right, so I'm mostly talking from a modeler
perspective. Now, if you are auser like pricing actuary or
valuation actuary, so you usemodels, but you don't alter
their functionality. You're justputting inputs in and getting
outputs out. For that, youactually don't need to know any
(32:34):
Python at all because all ofthat is done through a GUI. Now,
a great thing about a sourceavailable approach is that you
can bring your own GUI.
And in fact, you can havemultiple GUIs. So instead of a
single desktop applicationthat's used by pricers and
modelers and tries to do wellfor both of them, we say, well,
(32:58):
why not allow use of third partytools? So a modeler can use
Visual Studio or PyCharm orwhichever IDE they prefer. Now
the pricer can use Excel or webinterface or some other media
that integrates well with theirparticular systems and
(33:20):
workflows. And you can swapthose out or you can even build
your own.
So it helps a lot when yoursystem is source available with
the ideas that you can buildanything you want in it, and
that could include your GUIs.
Chris (33:37):
So it sounds like you're really focused on kind of
maximizing the actuary's timeand minimizing the amount of
effort to get compared to whatit was prior to them having this
kind of capability available tothem. How does that, like if
you're the actuary using it,like what is the impression of
(33:57):
how it changes jobs? I ask thatin terms of the requirements of
doing the job because that's oneof the first things that people
are always asking is kind of howdoes the technology change what
it means to be an actuary? Likein as a software developer
myself, that's in the AI space,people are always talking about
what the quality of softwaredevelopment is with our with the
(34:19):
all these cool new tools wehave. And often the results
there's been a lot ofconversations about it, you
know, depressing the softwaredevelopment need.
But in actuality, I actuallydisagree. I find the tools to be
very helpful, and I'm still veryengaged in it. What is the
actuary experiencing now, havingramped up on your process versus
(34:41):
the process of doing it fairlymanually with spreadsheets and
everything before? What's thefeedback you get on that in
terms of what people are how dothey perceive this new way of
doing it now that they're kindafully ramped in and up to speed?
Igor (34:55):
So from pricing perspective, I think this would
be most vivid. So as a pricer,you basically take the request
for proposal for a certaininsurance contract. You figure
out how to plug it into yourmodels. And ultimately, you
produce a quote that says, okay,we will do this reinsurance for
(35:17):
this much money on these terms.Right?
If you ask a pricing actuarytoday, what are they doing
during their day? They wouldsay, Oh, I'm pricing. Now, if
you ask them, Okay, so whatexactly does that mean?
Mechanically, what are youdoing? They would say, Well,
about 50% to 70% of my time, I'mliterally moving numbers from
(35:42):
one model into another model.
And then run the next model andthen takes the outputs and plug
them into the next model. Theremaining time I'm spending
checking the results and makingsure that they make sense and
that everything looks good andthere is no strange things
(36:03):
happening as far as the resultsthat I expect versus what I
actually got. And then I'd sayabout 10 to 15% of time is also
spent on hunting down answers tovarious ambiguities that were
presented in this contract. Soyou would reach out to the party
requesting and saying, Hey,there is a bunch of people
(36:24):
missing gender. Do you know whatgender they are or their
birthdays are in the future?
We're pretty sure that's wrong.So that's the world today.
Right? Once you implement aplatform like ours, what happens
is you set up your pricingprocess once. And then from
there, at any point, if you getany data updates or changes or
(36:47):
you want to try something, youchange your original inputs, you
click a button.
And because the system remembersthe steps you took, it'll
actually do the full repricingfor you in about ten, fifteen
minutes. So it's a good amountof time to go get a coffee. And
that frees up your time to domore analytical things or to
(37:13):
price more quotes. Managementloves the second one. But as a
pricer myself, I like the thirdoption.
It gives you an option to gohome at 5PM as opposed to 2AM.
Chris (37:28):
That's a good answer, actually, a really good answer.
So you've taken it this far,you've really made an impact in
terms of the capabilities you'reoffering to the companies in
your industry as you're takingon new customers and stuff. As
you're looking forward in time,and and over some period, you
(37:48):
know, not not immediately, but,know, over a year or two, maybe
longer, depending on what yourhorizon is. How are you
envisioning the industrychanging, and do you have any
ideas for the future that youwould like to see either
yourself implement or theindustry at large change and
develop into? What kind offuture you know, when you're
(38:10):
when you're kind of done for theday and you're having, you know,
a glass of wine or just kind ofchilling and your mind's
wandering a little bit, what'sthe future that you're excited
to see that you're wanting tohelp build or be a part of?
Igor (38:25):
Yes. So I see the future we wanna change the way actual
modeling is done. Instead ofcurrent proprietary softwares
that are being used, we envisiona future that is built on common
technologies. So particularlyPython and tooling around that.
And we think that that futurewill be much more efficient and
(38:48):
it will make our profession evenmore valuable than it is today.
Reason for that is if you learna proprietary software today,
you go to a different company oreven a different department
within the same company and theyuse different software. And then
you have to relearn everythingyou learn because all the
buttons are different, all themenus are different, the logic
(39:10):
is different, and many thingsare different. If you build
modeling on a common technologylike Python and Visual Studio
and DevOps, you learn it onceand those skills will be just as
applicable to any other companywhere you go. In fact, they
(39:31):
apply beyond just modeling. Youcan do automation, you can do
data science.
And in fact, they apply evenoutside of actuarial
professions. They're justgenerally more and more valuable
as our society becomes much moretechnologically advanced.
Chris (39:47):
Fantastic. Well, Igor, thank you so much for coming on
Practical AI today. I learned alot and really appreciate you
sharing, not only kind ofteaching us a bit about the
industry you're in, but tellingus how NICE Technologies is
starting to change the face ofthe industry in terms of
actuaries are able to kindalevel up and go home at 05:00
(40:09):
instead of 2AM. Definitely a bigbonus there. Thank you very much
for coming on the show.
Igor (40:21):
All
Jerod (40:21):
right. That's our show for this week. If you haven't
checked out our website, head topracticalai.fm and be sure to
connect with us on LinkedIn, X,or Blue Sky. You'll see us
posting insights related to thelatest AI developments, and we
would love for you to join theconversation. Thanks to our
partner Prediction Guard forproviding operational support
for the show.
Check them out atpredictionguard.com. Also,
(40:44):
thanks to Breakmaster Cylinderfor the Beats and to you for
listening. That's all for now,but you'll hear from us again
next week.
Welcome to the Practical AI podcast, where we break down
the real world applications ofartificial intelligence and how
it's shaping the way we live,work, and create. Our goal is to
help make AI technologypractical, productive, and
accessible to everyone. Whetheryou're a developer, business
leader, or just curious aboutthe tech behind the buzz, you're
(00:24):
in the right place. Be sure toconnect with us on LinkedIn, X,
or Blue Sky to stay up to datewith episode drops, behind the
scenes content, and AI insights.You can learn more at
practicalai.fm.
Now, onto the show.
Chris (00:48):
Welcome to another episode of the Practical AI
podcast. My name is ChrisBenson, your co host. And with
me today, we have Igor Nikotin,who is the CEO and co founder of
NICE Technologies. Welcome tothe show, Igor.
Igor (01:02):
Thank you, Chris.
Chris (01:04):
So one of the things that we love doing on the show is to
get into different industriesand explore how different
industries are implementing andusing AI in different ways. And
you are an actuary, or at leastthe CEO of an actuarial modeling
company. And that, we may havecertain folks in the audience
(01:25):
who are not intimately familiarwith your industry, and I was
wondering if we could startwith, kinda tell us a little bit
about what your industry is, aswell as kind of how you got into
it, and then kind of into whythe technology around it
specifically?
Igor (01:39):
Sure. So Actuary, not to be confused with ACTR, is a risk
management professional. We workat insurance companies, and we
are the people who figure outhow much insurance costs, what
are the terms of insurance, andvarious aspects of managing
insurance companies. So if yougo to any insurance company,
(02:01):
people there would tell you thatit is really run by actuaries in
in that company.
Chris (02:06):
You're the folks with the power in it there. You're you're
the ones that are actuallymaking the wheels turn.
Igor (02:11):
Well, us and sales. Because people don't wake up in
the morning thinking, Oh, youknow what, today I'm gonna buy
myself some life insurance.
Chris (02:21):
That's good. Given the fact that it's risk management
of this, could you talk a littlebit about what is involved? If
your job is to wake up, you'rean actuary, and you're going to
work at an insurance company,what does that look like? What
is it that you're doing for theinsurance company to do, like
how do you implement riskmanagement? What are the
(02:43):
traditional tools, aside fromsome of the technologies that
we're about to get into withyou?
What does that world look like?
Igor (02:50):
It usually looks like a lot of Excel spreadsheets. And
for various items, there aremore sophisticated models. So
you may see proprietary softwarebeing used for certain things
like projecting benefits. Youmay see some software
specialized on the projecting ofyour investments and on your And
(03:14):
in general, your day looks likedoing a lot of financial
modeling to answer questions orto satisfy various regulations
or to develop new products. Butin the end of the day, most of
what actuaries do are some formof modeling or regulation
(03:34):
research or compliance.
Chris (03:36):
Gotcha, so that is the wheels behind the scene in that
industry that makes it turn. SoI guess that when an insurance
company, for instance, is goingto price its products, know, the
insurance that a consumer mightbuy, I assume that all the
aspects that go into that arepart of what the actuary figures
out in terms of like, they can'tprice it without knowing all the
(03:58):
different types of risks and howthose play into the model. And
so I'm curious, how did you findyourself? Did you start yourself
as an actuary and then go intotechnology? Or did you come from
the other side as a technologistinto the space?
How did you find yourself inthis field?
Igor (04:17):
It was actually quite an interesting story. So I was
programming since I was the ageof 13 because I told my parents
that I must have computer tolearn programming. That was a
ruse. I just wanted to playvideo games like all the other
cool kids. But I think they getthe last laugh because here I am
a software engineer.
(04:38):
So it was a hobby of mine. ThenI moved to US and I kind of lost
my circle of friends who wereinto programming. And so
thinking what I want to do, myparents were telling me, well,
maybe you should be a lawyer. SoI looked up, okay, to be a
lawyer, what major do I need?And it was English.
(05:01):
I did one semester of English. Ihated it with lots of passion
because somehow there is nowrong answer, but all of my
answers were somehow wrong. Andso I did a bit more soul
searching and I decided that Iwant to become a high school
teacher for mathematics. At theend of my college, I taught for
(05:22):
one semester in American highschool and I decided that I need
to switch a profession. So nextthing I know, I'm sitting in a
coffee shop Googling what to dowith a mathematics degree.
And that's when I discoveredactuaries. And I said, oh, okay.
Interesting. So there are someexams to take. I took my first
exam.
(05:43):
It was probability. And I waslike, oh, probability. I'm great
at probability. So I went in. Igot two out of 10 on that exam.
That is not a good grade. And Iwas like, Oh, you know, that is
actually challenging enough of aprofession for me. I like it. So
how do people pass this thing?It is absolutely brutal.
(06:04):
Turns out you get a manual thatis about thousand pages sick
with about 2,000 practiceproblems. You solve through all
of that and then you show up toexamine, it's easy at that
point. So I did that. So Iapplied to a bunch of entry
level actuarial jobs, but thatwas in January. And some of you
may remember that there was afinancial crisis right at about
(06:26):
that time
Chris (06:27):
when some Yes. People got laid
Igor (06:30):
So I discovered that getting a job as an actuary was
very, very difficult as an entrylevel actuary. And then I said,
Okay, what else do I know how todo programming? So I applied to
a bunch of insurance companieswith my resume for both actuary
and programming jobs. And I gotinvited by Prudential Financial
(06:52):
actually for an interview for aprogramming position. On my
resume, it said, I haveintroduction to C plus plus as
the only programming course.
And then I also brought aportfolio of things I've done.
And they said, Okay, you havedone a whole lot of stuff. And
so they hired me as aprogrammer. Pretty quickly,
realized that as an actuaryfocusing on modeling, I can
(07:14):
continue to do programming butget paid twice as much. And so I
joined their actual leadershipdevelopment program and rotated
through a bunch of differentdepartments, learning how
insurance works, won aninnovation competition along the
way there with the idea of kindof future of modeling, what does
(07:34):
it look like, and what should afirm do to prepare for that?
So that's how I ended up beingan actuary.
Chris (07:42):
Gotcha. So I'm curious, while the current age of AI had
kicked off again by that time,it was really before anybody was
doing it. So I'm kinda curiousas you were looking at models of
the day that you're actuallyworking with in real life at the
job, just to draw a distinction,you know, as we as we talk of
(08:03):
modeling over this past decadeplus. What were you typically
working with? What kinds ofmodels, what were the
algorithmic things that youcared about in your day job when
you were doing that, as youmoved into software development,
but within that actuarialmodeling context?
Igor (08:19):
So coming from a software background, I was actually very
surprised by a lot of thingsthat I've seen. Because quite
often models, instead of beingthis kind of software that is
very rigorously tested andoptimized, you discover that
there is a lot of qualitycontrols that is commonplace in
(08:41):
the software industry that justdoesn't exist. Part of it is
because how do you versioncontrol an Excel workbook? And
that is very challenging. Butcoming from background of using
Git and DevOps, you're like,Well, there is a way.
We just need to get away fromExcel somehow. But that runs
(09:02):
into a different problem ofthere was no programming
training for actuaries. Therewas no required programming
training for actuaries prior to2018. In 2018, there was
actually programming added tothe actuarial exams. And since
then, the profession becameoverall much more knowledgeable
(09:23):
on programming and programmingtools in general.
Because the situation was kindof interesting that we all do
modeling, but none of us hadsome sort of like very hands on
modeling course that is verykind of operations based as
opposed to understanding thetheory behind insurance and all
(09:45):
of the probability calculationsand things.
Chris (09:47):
I'm curious, is the field in 2018 updated itself, as you
said, and you started seeingprogramming as a requirement,
how did that lead into, now thatit's kind of recognized as part
of the field, and over the yearssince 2018, not too long here,
(10:08):
seven years, how has thataffected the field itself in
terms of getting up to the kindsof things that you're doing at
this point where you're havingAI capabilities and kind of
modern algorithms directlyimplemented for actuarial
fields? What did that look like,that evolution? Because that's
(10:30):
not a lot of time between 2018and 2025 as we record this.
Igor (10:35):
It is actually very different. And it was actually
driven not so much bytechnology, but by regulation.
Regulations became much moresophisticated and that kind of
forced everybody to use bettertechnology. Because if you need
to run stochastic runs withthousands of scenarios, you
(10:55):
cannot possibly do that inExcel. It's going to run for
weeks.
And so that forced the industryto upgrade their systems and to
adopt quite a few of thepractices that are much more
typical in software development.So seeing things like DevOps and
(11:16):
unit tests, it became much morecommon nowadays. It's still not
on every insurance company. Andit really depends kind of what
insurance company how insurancecompany operates. Right?
Because some of them, they focusjust on sales and they kind of
outsource all of the productdevelopment to other companies.
And the others are much morefocused on kind of the mechanics
(11:37):
of producing the insurance andthen letting other companies
care about sales. As far as AI,there is still very little
adoption of it. Chief problem isthe fact that AI is not reliable
in a sense that it doesn'tnecessarily give you the truth.
And when you're complying withregulations, you do have to have
(11:59):
the truth.
So there is quite a lot ofexperiments of what is it useful
for in summarizing the data.There is some experiments in
model development assistance.And yeah, I would say and in
(12:21):
providing a second opinion tothe data analysis that is being
done. So I would say those arethe three areas where we see AI
actually having some real worldvalue that we observe.
Sponsor (12:53):
Well, friends, when you're building and shipping AI
products at scale, there's oneconstant, complexity. Yes.
You're wrangling models, datapipelines, deployment
infrastructure, and then someonesays, let's turn this into a
business. Cue the chaos. That'swhere Shopify steps in whether
you're spinning up a storefrontfor your AI powered app or
(13:14):
launching a brand around thetools you built.
Shopify is the commerce platformtrusted by millions of
businesses and 10% of all USecommerce from names like
Mattel, Gymshark to foundersjust like you. With literally
hundreds of ready to usetemplates, powerful built in
marketing tools, and AI thatwrites product descriptions for
(13:34):
you, headlines, even polishesyour product photography.
Shopify doesn't just get youselling, it makes you look good
doing it. And we love it. We useit here at Changelog.
Check us outmerch.changelog.com. That's our
and it handles the heavy liftingtoo. Payments, inventory,
returns, shipping, even globallogistics. It's like having an
(13:57):
ops team built into your stackto help you sell. So if you're
ready to sell, you are ready forShopify.
Sign up now for your $1 permonth trial and start selling
today atshopify.com/practicalai. Again,
that is shopify.com/practicalai.
Chris (14:21):
So as you got to this point in your career where where
technology is changing, youknow, and it's evolving, How did
you get from that point intowhere you're at now with NICE
technologies and and thecapabilities that you provide?
What was the the remaining partof the journey into as you were
kind of getting closer to themodern point?
Igor (14:43):
It really was born out of necessity. So our platform is
kind of two pieces. There'scalculation engine and data, and
those are kind of two majorparts. The data management part
was really born at 2AM as I wasworking on pricing reinsurance
(15:04):
deals. And in that job, you dealwith a lot of reruns and with a
lot of tool chains that arerather manual.
So new piece of informationcomes in for a meeting that you
have tomorrow at 9AM. This pieceof information comes in at 5PM
the day before. And then thetool chain to incorporate this
(15:30):
piece of information into thefinal quote can take four or
five hours of differentcombinations of Excel models,
Python models, differentproprietary models. One reason
for that is that reinsurancepricing is very bespoke. And so
it's difficult to automate aprocess that is bespoke.
(15:51):
And because of that, you end upactually sitting in the office
until you're done incorporatingthis and checking this. And then
you send your boss a message atlike 2AM saying, hey, it's done.
Can you review it by 9AM? Andthey would actually be there
knowing that they do need toreview it because there's a lot
(16:11):
of money, a lot of eyes on thequote and a lot of
responsibility. So that's when Icame up with the idea of, all
right, how there must be somesort of engineering solution to
this.
And the solution is really whatif you make your tool chain
remember the steps you took inyour original pricing. Right?
(16:33):
Because if you have that and itcan recreate those steps without
making mistakes that humans tendto make when they work at 2AM
for the fifth night in a row.That's where that idea was born.
The other part of it was I'vedone modeling in several
(16:54):
different systems.
And I realized that for certainproducts, some systems are far
easier and quicker and cheaperto use than others. The
challenge there is what if thecheapest system and the best
system to use for a givenproduct is an in house system?
And that's where I took where Irealized that we can take a page
(17:16):
from the video game developmentindustry. So in video game
development industry in the'90s, they had the same
problems. They had a bunch ofproprietary software.
As complexity of games took offwith compute power, you said,
Oh, three d graphics are great.Your development costs also
skyrocketed, but you can't sellvideo games more expensively
because your customer refuses topay more than $50 So how do you
(17:40):
solve it? Well, Unreal Engineand Unity, right? They basically
provide a core functionalitythat everybody builds off of to
build the actual video games.And that makes the process much
cheaper, and hence, you canactually have games that cost
$50 as opposed to $5,000 So wesaid, okay, can we take that
similar concept and apply it toinsurance?
(18:01):
So we would need a software thatprovides access to the
underlying code, but at the sametime, we need to make sure that
it is commercially viable andwell supported and industrial
grade. And that's how we came upwith idea for our calc engine,
which is source availablebusiness model. Meaning we
(18:22):
provide source code to ourclients, but not to general
public.
Chris (18:25):
Gotcha. So it's interesting that I I love you're
kind of there's a little bit ofkind of a classic
entrepreneurial story there, andthat you're a, kind of
scratching your own itch. Like,you're recognizing my job would
be easier if I were to recreatewhat I was doing before and have
that immediately available. Andthen that's interesting
(18:48):
inspiration. Would not I don'tthink I would have expected you
to the analogy between theactuary in field and also gaming
is pretty cool.
So it sounds like you'rebuilding an app kind of in the
same spirit as you mentioned,Unreal Engine and Unity, which
are the two biggies in thatindustry. So in the end, could
(19:10):
you talk a little bit about whatyour calc engine became and how
it was received in the industry?Like what you know, I know, a
sense, Daniel, who is my cohost,who is not able to make it here
today, is also the CEO of asmall company that's growing
very fast. And I love hearingkind of how he's gotten through
the steps of how his company hasdeveloped from kind of the the
(19:33):
challenges at the verybeginning, and and he gets
traction and stuff. Could youtalk a little bit about what
your what your pathway has beenlike as you've been developing
the the calc engine and theother features at NICE
Industries?
What has that pathway lookedlike for you?
Igor (19:48):
So we we always leaned on feedback. Even before we started
a company, we we did went toseveral executives in the
industry and said, would this beinteresting? You know? We went
to heads of pricing, heads ofmodeling, and said, you know,
what if this thing existed?Would you care?
And the answer was we would loveto see a demo. Pretty much every
(20:11):
single person we spoke withsaid, we would love to see a
demo. We said, okay, great. Butto build a demo, we need like a
year of work because this ispretty complicated thing to to
build if you want a demo thatlooks somewhat realistic. So
with that knowledge, we said,okay, well, let's let's see if
(20:32):
anybody else believes in this.
We reached out to several peoplethat we wanted to cofound a
company with, and only two ofthem did not join. One of them
did not join because she had ababy five days before we called
her. So that was understandable.
Chris (20:48):
That's a pretty good reason for anything, quite
honestly, in the world.
Igor (20:51):
But the others joined, and that was a very defining moment
of saying, Oh, okay, I'm not theonly person who believes in
this. And then we went toinvestors and we said, Well,
would you be willing to investin this kind of venture? And we
discovered that the answer wasalso yes. So we said, okay,
(21:11):
well, apparently there's quite afew people who are interested.
So then we developed a demo.
We showed it to the originalgroup of people that we were
working with. They said, okay,great. When is it going to be
commercially available? We said,well, probably for another year
or so because there's a lot ofwork to go from prototype to a
(21:32):
production grade system. Andthen we launched publicly in
this February with a productiongrade system.
And right now we're in talkswith a lot of different
insurance companies in variousstages of negotiations.
Chris (21:49):
So it sounds like right now you're kind of the hot new
thing in that field. Because ifyou just came out a few months
ago, you've got this great newthing that's gonna revolutionize
the industry with a capabilityit didn't have before. I'm just
curious, as an industry thathasn't had this kind of
capability in the past, what'sthe reception like on that? I
(22:10):
see in a lot of industries thatwe'll talk to people who, you
know, there's a bit of astruggle to shift mindset into
recognizing it, and then thereare other industries where
people are just waiting for theleap forward and stuff. Is it,
or some combination of the two,what is it like changing an
industry, you know, in that wayat such a fundamental level as
(22:31):
you're trying to get traction onthat and bring in for you as a
as a business owner, bringingnew customers on board, but kind
of giving them a new way of ofdoing that?
Igor (22:41):
It's a lot of explaining how we're different. Because on
one hand, people quickly catchon that our business model is
completely different fromeverybody else in the market. On
the other hand, the most commonquestion I get is, do I require
to know Python to use yoursystem? The answer is no. But
that is like the number onequestion the most common
(23:03):
question that I get.
And so it is a bit of a process,especially at big companies
where there are manystakeholders, a lot of different
vetting processes, and a lot ofcoordinating between a lot of
parties. So you need to havepricing on board and valuation
on board and modeling on boardand IT on board. And it just
(23:25):
takes a lot of time tocoordinate everything between
those. Just like life insurancepolicies, insurance companies
usually don't wake up in themorning and say, Oh, you know
what? We're going to swap ourmodeling platform.
That's a big undertaking, and itis not taken lightly. And having
(23:45):
been in those shoes, I realizethat there's a lot of work and
you would only do it for very,very compelling business
reasons.
Chris (23:53):
It sounds I I like, when if you look at like, speaking
about one of your customers,that they're really it's really
to to get the benefits and toleap forward, they they have
their workflows, which is and asyou kinda pointed out, the
different components of theirbusiness, the different
departments, you will, have toall kinda get on board. It
(24:14):
reminds me a little bit as yousay that, and I don't know that
this is a good analogy, so I'llit out of kinda when you put in
like like an SAP implementationin a in a company, and it hits
many different functions of it.It seems to have where the
workflows themselves at thosethe at the customers will
change. It sounds like differentsoftware, but within your
(24:36):
industry, like you're, there's abunch of plugin points where
people have to buy in andworkflows change so that they
can get the benefits of surgingforward. Is that an accurate way
of representing that?
Igor (24:48):
Oh, yeah, it is somewhat similar. I would agree that it
does change the way people workfor some people more than
others. So if you look at itfrom a model developer
perspective, you used to useproprietary software where you
had a bunch of dropdowns whereyou could edit some formulas.
(25:09):
Fun fact, those formulas looksuspiciously similar to code.
And now you use something likeVisual Studio, for example.
Now, looking at it fromperspective of an actuary, you
would say, okay, that's a bigdifference. Now I have to
understand Python. But lookingat it from kind of an actuary
who also has experience inprogramming, I see it as, oh, I
(25:33):
finally have Git for versioncontrol. I finally have DevOps.
I finally can automaticallymerge code from multiple
developers.
I finally have amazingcapabilities and breakpoints and
a lot of these tools that I didnot have before that allow me to
(25:54):
understand things quicker. Andin today's world, that also
means that you can have an AIcode assistance that can explain
how things work, that can makevariants of different
calculations, that can createunit tests for you, and do all
of those things actually quitewell.
Chris (26:14):
So that is pretty fascinating to me in terms of
the description. As you'retalking about tools like Git and
Visual Studio and stuff likethat, and those are things that
I obviously know coming at it asa developer, Could you talk a
little bit about recognizingthat not every actuary is a
developer and everything, couldyou do a deep dive into kind of
(26:37):
what your workflow looks like?Like from the customer's
perspective, what does that turntheir reality into? And can you
draw out some of those pointsabout why this is better than it
was before by doing that?Because we've talked a bit about
the investment in learning newstuff, the investment in getting
the different departments of theinsurance company working
(26:59):
together.
Take me through it, and if youcould draw out some of the
benefits of the technology andspecifically the AI bits that
you have into it, I'd love tounderstand what that looks like.
What does a modern actuarialworkflow look like for a company
that's now out on the leadingedge of that?
Igor (27:19):
Sure, so a simple example would be, let's say you have an
Excel model that does something.Right? So imagine a dozen or so
tabs full of Excel formulas, andthat does some sort of
calculation. Right? So thenumber one question you would
get is, well, how does it work?
If I want to understand how thisworks, I would spend some time
(27:42):
digging through all of theseformulas and tracing and kinda
understanding how they allconnect. Now in the in our
platform, right, you literallyjust ask the AI system, hey, can
you explain this how thiscalculation works? Can you point
me to where in the code is it?And you get an exact location in
the code and the calculation.Right?
(28:05):
Then you say, okay. Well, I needto modify this model to support
a new product. And there ismaybe several pieces of that
product that would need to bemodified. In Excel, it's very
difficult to merge models. So ifyou have three people working in
parallel, it's very difficult toput all of their changes
together and make sure that youdidn't miss anything.
(28:28):
Versus if you work in Python,you have Git. You just say, hey,
can you merge this? And if yourplatform is well organized,
meaning it's object oriented,you would have very little
problems with that. And there'sgood chance it'll just auto
merge in seconds and say, okay,there is no conflicts. So it
(28:50):
enables development offunctionalities as a team as
opposed to being forced to be ina single developer because
working in parallel is simplytoo complicated to merge.
And then you say, okay, so now Ihave this functionality. It
kinda sorta works. I wanna makesure that it's production grade,
so I need to test it. In Excel,you have a limited number of
(29:15):
tests that you would hopefullycome up with. I've seen dozens
usually being kind of a decentnumber.
With the help of AI, you cancreate a lot more than that. And
with unit test functionality inVisual Studio or DevOps, You can
have literally thousands or evenmillions of unit tests to make
(29:38):
sure that your software doesexactly what you want and can
handle all of the edge casescorrectly. And then the next
step would be to say, okay,well, it works in Excel, but
this is glacially slow. What canI do? And at that point, you're
quite often just stuck becausethere's simply no way of getting
(30:00):
it faster.
Now, in our world, you can askyour AI assistant, Hey, can you
check my code for Well, first ofall, you have the performance
benchmarks and those would gettriggered if you've done
something that tripled theruntime all of a sudden. But
second, you can ask AI assistantto say, hey, can you tell me
(30:22):
which part of my code is slowand can you give me some ideas
of how to do it quicker? Andquite often you will get, at the
very least, will identify whereyou're doing same calculation
over and over or doing otherthings that are suboptimal.
Overall, it's just much nicerexperience, which requires some
(30:46):
knowledge of Python to beeffective at it from modeling
perspective. But at the sametime, with AI systems, that
threshold of how much Python youneed to know is much lower.
As well as it is just so muchfaster to ask AI assistant for
like, Hey, can you explainsomething? Or can you find
(31:07):
something in the code base?
Chris (31:08):
It sounds like, just as an analogy on another topic that
we brought in, common buzzphrase these days is vibe
programming, vibe development,where you're working with a LLM
on your code, and that it soundslike there's sort of an analogy
in that the actuary is going toknow some Python to use the
(31:29):
system, but the system helpswith that. Is that is that a
fair assessment to where thesystem itself is kind of working
in that vibe coding manner ofsaying, I can show you the
challenge in the code that youneed to address or and maybe I
can help you fix it and thingslike that. What are can you
differentiate a little bit aboutwhich parts of the workflow the
(31:52):
actuary is getting into thePython and what kinds of things
are able to be done through aGUI or through an agentic
approach, for instance, whereyou have an agent that may have
taken on a task that washistorically a very manual task
or anything like that. Do youhave anything that you can share
along those lines just in termsof how workflows evolve?
Igor (32:13):
Right, so I'm mostly talking from a modeler
perspective. Now, if you are auser like pricing actuary or
valuation actuary, so you usemodels, but you don't alter
their functionality. You're justputting inputs in and getting
outputs out. For that, youactually don't need to know any
(32:34):
Python at all because all ofthat is done through a GUI. Now,
a great thing about a sourceavailable approach is that you
can bring your own GUI.
And in fact, you can havemultiple GUIs. So instead of a
single desktop applicationthat's used by pricers and
modelers and tries to do wellfor both of them, we say, well,
(32:58):
why not allow use of third partytools? So a modeler can use
Visual Studio or PyCharm orwhichever IDE they prefer. Now
the pricer can use Excel or webinterface or some other media
that integrates well with theirparticular systems and
(33:20):
workflows. And you can swapthose out or you can even build
your own.
So it helps a lot when yoursystem is source available with
the ideas that you can buildanything you want in it, and
that could include your GUIs.
Chris (33:37):
So it sounds like you're really focused on kind of
maximizing the actuary's timeand minimizing the amount of
effort to get compared to whatit was prior to them having this
kind of capability available tothem. How does that, like if
you're the actuary using it,like what is the impression of
(33:57):
how it changes jobs? I ask thatin terms of the requirements of
doing the job because that's oneof the first things that people
are always asking is kind of howdoes the technology change what
it means to be an actuary? Likein as a software developer
myself, that's in the AI space,people are always talking about
what the quality of softwaredevelopment is with our with the
(34:19):
all these cool new tools wehave. And often the results
there's been a lot ofconversations about it, you
know, depressing the softwaredevelopment need.
But in actuality, I actuallydisagree. I find the tools to be
very helpful, and I'm still veryengaged in it. What is the
actuary experiencing now, havingramped up on your process versus
(34:41):
the process of doing it fairlymanually with spreadsheets and
everything before? What's thefeedback you get on that in
terms of what people are how dothey perceive this new way of
doing it now that they're kindafully ramped in and up to speed?
Igor (34:55):
So from pricing perspective, I think this would
be most vivid. So as a pricer,you basically take the request
for proposal for a certaininsurance contract. You figure
out how to plug it into yourmodels. And ultimately, you
produce a quote that says, okay,we will do this reinsurance for
(35:17):
this much money on these terms.Right?
If you ask a pricing actuarytoday, what are they doing
during their day? They wouldsay, Oh, I'm pricing. Now, if
you ask them, Okay, so whatexactly does that mean?
Mechanically, what are youdoing? They would say, Well,
about 50% to 70% of my time, I'mliterally moving numbers from
(35:42):
one model into another model.
And then run the next model andthen takes the outputs and plug
them into the next model. Theremaining time I'm spending
checking the results and makingsure that they make sense and
that everything looks good andthere is no strange things
(36:03):
happening as far as the resultsthat I expect versus what I
actually got. And then I'd sayabout 10 to 15% of time is also
spent on hunting down answers tovarious ambiguities that were
presented in this contract. Soyou would reach out to the party
requesting and saying, Hey,there is a bunch of people
(36:24):
missing gender. Do you know whatgender they are or their
birthdays are in the future?
We're pretty sure that's wrong.So that's the world today.
Right? Once you implement aplatform like ours, what happens
is you set up your pricingprocess once. And then from
there, at any point, if you getany data updates or changes or
(36:47):
you want to try something, youchange your original inputs, you
click a button.
And because the system remembersthe steps you took, it'll
actually do the full repricingfor you in about ten, fifteen
minutes. So it's a good amountof time to go get a coffee. And
that frees up your time to domore analytical things or to
(37:13):
price more quotes. Managementloves the second one. But as a
pricer myself, I like the thirdoption.
It gives you an option to gohome at 5PM as opposed to 2AM.
Chris (37:28):
That's a good answer, actually, a really good answer.
So you've taken it this far,you've really made an impact in
terms of the capabilities you'reoffering to the companies in
your industry as you're takingon new customers and stuff. As
you're looking forward in time,and and over some period, you
(37:48):
know, not not immediately, but,know, over a year or two, maybe
longer, depending on what yourhorizon is. How are you
envisioning the industrychanging, and do you have any
ideas for the future that youwould like to see either
yourself implement or theindustry at large change and
develop into? What kind offuture you know, when you're
(38:10):
when you're kind of done for theday and you're having, you know,
a glass of wine or just kind ofchilling and your mind's
wandering a little bit, what'sthe future that you're excited
to see that you're wanting tohelp build or be a part of?
Igor (38:25):
Yes. So I see the future we wanna change the way actual
modeling is done. Instead ofcurrent proprietary softwares
that are being used, we envisiona future that is built on common
technologies. So particularlyPython and tooling around that.
And we think that that futurewill be much more efficient and
(38:48):
it will make our profession evenmore valuable than it is today.
Reason for that is if you learna proprietary software today,
you go to a different company oreven a different department
within the same company and theyuse different software. And then
you have to relearn everythingyou learn because all the
buttons are different, all themenus are different, the logic
(39:10):
is different, and many thingsare different. If you build
modeling on a common technologylike Python and Visual Studio
and DevOps, you learn it onceand those skills will be just as
applicable to any other companywhere you go. In fact, they
(39:31):
apply beyond just modeling. Youcan do automation, you can do
data science.
And in fact, they apply evenoutside of actuarial
professions. They're justgenerally more and more valuable
as our society becomes much moretechnologically advanced.
Chris (39:47):
Fantastic. Well, Igor, thank you so much for coming on
Practical AI today. I learned alot and really appreciate you
sharing, not only kind ofteaching us a bit about the
industry you're in, but tellingus how NICE Technologies is
starting to change the face ofthe industry in terms of
actuaries are able to kindalevel up and go home at 05:00
(40:09):
instead of 2AM. Definitely a bigbonus there. Thank you very much
for coming on the show.
Igor (40:21):
All
Jerod (40:21):
right. That's our show for this week. If you haven't
checked out our website, head topracticalai.fm and be sure to
connect with us on LinkedIn, X,or Blue Sky. You'll see us
posting insights related to thelatest AI developments, and we
would love for you to join theconversation. Thanks to our
partner Prediction Guard forproviding operational support
for the show.
Check them out atpredictionguard.com. Also,
(40:44):
thanks to Breakmaster Cylinderfor the Beats and to you for
listening. That's all for now,but you'll hear from us again
next week.
Popular Podcasts
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.
The Breakfast Club
The World's Most Dangerous Morning Show, The Breakfast Club, With DJ Envy, Jess Hilarious, And Charlamagne Tha God!
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.