February 13, 2025 • 47 mins
Text us your thoughts on the episode or the show!
Join us as we unravel the transformative power of AI in business with Kobi Stok, the visionary entrepreneur behind Forwrd.ai. Discover how his journey from WalkMe to launching Forwrd.ai is reshaping the landscape of data science automation. This episode promises insights into how AI can act as a team of data scientists, empowering businesses to turn complex data into clear, actionable strategies and enhanced performance. Kobi provides an insider's view into current challenges and solutions, highlighting the need for accessible tools that revolutionize decision-making processes.
Episode Brought to You By MO Pros
The #1 Community for Marketing Operations Professionals
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Michael Hartmann (00:00):
Hello everyone .
Welcome to another episode ofOpsCast brought to you by
MarketingOps.
com, powered by all the MoProsout there.
I'm your host, michael Hartman,joined today by Mike Rizzo.
Say hello, mike.
Mike Rizzo (00:10):
Hey, what's happening?
This is a new thing for us.
We're going to do video guys.
Michael Hartmann (00:15):
I know right.
Mike Rizzo (00:16):
MoPros, all you people out there, we're going to
use video from now on.
Michael Hartmann (00:27):
I don't know how this is going.
I've got a face for radio, youknow.
So tying into the music talk wehad before we started recording
, all right.
So joining us today to talkabout the potential impacts of
AI in business operations isKobe Stock.
Kobe is a serial entrepreneurwith two decades of experience
in building software.
He currently leads Forward AI,an automated data science
platform built for RevOps.
Previously, he was with WalkMe,where he led the global product
(00:48):
strategy with a focus on scale,growth and innovation.
He joined WalkMe through theacquisition of Abbeyio, a mobile
AI company, where he wasco-founder and CTO, leading the
product and technology.
Prior to WalkMe, he was thefounder and CEO of a consumer
music tech company with millionsof users.
Early in his career, he was asoftware engineer, architect and
a manager at multiple startupsas well as SAP.
(01:09):
So, kobe, thanks for joining ustoday Late in the evening for
you.
Kobi Stok (01:14):
Hello, hello.
It's a good thing that I justchanged my camera as we're doing
video, so full HD here.
There you go, full transparency.
Michael Hartmann (01:23):
If I showed my background fully, you would see
the in-process taking down ofour holiday decorations.
Mike Rizzo (01:29):
So I had to move a few things before Nice Stuff
that I haven't had to deal within the past.
Michael Hartmann (01:37):
So all good.
So you founded Forwardai nottoo long ago.
It was recently.
You started it.
You call it an automated datascience platform.
So for us and our listeners,what does that mean to you and
how should we be thinking aboutthat?
Kobi Stok (01:52):
Yeah, so actually every time.
So that's my third company thatI'm starting right, and every
time that I start a company Itry to, before I check the
market and check the need, I tryto solve my own problem.
That's like number one.
And in my previous role we weretrying to build a model that
(02:14):
will predict churn based onproduct usage.
Think about it we work for acompany collecting tons of data
and we're trying to build amodel that will basically use
this data and actually surfaceaccounts that are at risk.
So we get a heads up and we cancreate some sort of a
mitigation.
(02:36):
And it took us a year to do it Ayear.
It took us a year, one year todo it.
It and it was complex.
We needed to involve like atleast 20 people.
Half were technical, half werefrom the business.
You know we've talked about itso much and the process was so
(02:59):
long right, but when we finishedit after a year plus even and
we launched it, then what we did?
We basically surfaced thoseinsights to the customer success
team in our case, and theimpact was amazing.
It really changed how thecustomer success team operates
on the day-to-day, not on aspecific moment, and even
(03:22):
afterwards, we modified thecompensation of the reps based
on the model.
So I saw the impact and I sawthe investment and I said it's
too long to make thoseinvestments.
I mean, most companies won'teven start.
And if you think about it,companies are collecting so much
(03:45):
data these days and the datacosts money.
It takes time to arrange it, ittakes an effort to arrange it.
They're building tons ofdashboards right, but by the end
of the day, people are notreally data driven in the day to
.
People are not reallydata-driven in the day-to-day.
Michael Hartmann (04:07):
No.
It's just something I think alot of people either claim that
they especially marketers claimthat they either are or want to
be data-driven.
I'm not even sure they knowwhat that means.
Yeah.
Kobi Stok (04:20):
The best-case scenario is that companies have
good management dashboards thatprovide visibility on the
C-level, right On the leadershipteam, but when you go down to
the people who actually managethe business right Manage the
(04:41):
product, the day-to-day with thecustomers, support finance they
don't have this informationright, and they don't have this
information because it's reallyhard.
Yeah, and let's imagine that acompany has unlimited resources
(05:02):
of data science internally, then, using data science methods,
they wouldn't be able to surfacethose insights or data or
whatever, right, but they don'thave it.
So automating data sciencemeans that we are selling you a
piece of software that acts likehundreds of data science
(05:27):
professionals that are workingon your data to build models
that are fueling businessprocesses in your organization,
to make them all automated, tosurface insights or signals that
(05:47):
will help the team to performbetter, to surface allots or
emails right, with specific callto actions to improve
forecasting and basically everybusiness KPI that you can
imagine.
So that's what we do today.
That's interesting.
Michael Hartmann (06:06):
What really strikes me as interesting is
that when I first made mytransition from doing IT
management consulting mostlywith accounting and finance,
real estate type environmentsinto marketing, it was to do
database marketing so datawarehousing into marketing.
It was to do database marketingso data warehousing.
(06:27):
And we had a team of today thatwould be called data scientists
but you know statisticians whodid things like churn prediction
and likelihood to convert andall those kinds of models.
This was a big telecom companyand it was.
My job was to build a databaseto help with go-to-market
activity, so a 50 millionhousehold database and the data
(06:47):
processing and the amount ofdata that we had to acquire and
then analyze was huge and itcost a lot of money.
It was a huge effort.
I mean what you're describing.
It's almost surprising to methat that is still a challenge
at many places.
Kobi Stok (07:05):
You know what I mean Very, very big challenge.
I think that data became morecomplex.
So every time that someone, anycompany, improved something,
then another complexity layerthing happened and then it
became more complex.
Improved something, thenanother complexity layer thing
happened and then it became morecomplex.
(07:28):
And I think that in the past,let's say five to seven years,
we're in a race.
We were in a race before thatof hey, I need to collect data,
right, everyone was in a race, Ineed to collect more data, I
need to, right.
And now everyone stoppedcollecting data.
They have enough data.
Now, how do I get ROI on thosehuge investments that we just
(07:54):
did?
And we bought Snowflake and webought Salesforce and we have
HubSpot and we have Marketo andwe have Segment and we have
product analytics, like so manythings that essentially so much
value, right, there's inside somany things you can drive, but
you just don't have theresources and it's so complex to
(08:14):
even start.
So my goal was how do I makepeople start?
Forget about the price.
Mike Rizzo (08:23):
Yeah, yeah, forget about right.
Yeah, yeah, I think, and it'sso, um, it's so interesting to
be at this intersection of thislike data aggregation and
collection phase that we've beengoing through for the last
couple decades.
Where it was it, I don't know.
(08:45):
I think for some companies it'sgoing to feel a little bit like
a retirement plan, like you'vebeen piling this data away for a
while and eventually it'll havea payoff Hopefully a good one,
exactly and for others it'llfeel like let's just scrap the
last 10 years and then take themost recent five, and you know
(09:09):
so we'll look at the last 15 andtake the most recent five and
roll with that, because ourbusiness changed a lot.
But there's this, the.
The opportunity now iseverybody went.
It's a little bit of a cop out.
I don't know, maybe, maybe,maybe you don't feel this way,
kobe, but it's a little bit of aoh well, now I don't have to
(09:32):
think quite as hard Like AI'shere.
So cool, let's figure out howto use the data now, which I
think is great, right?
We've all kind of just beenwaiting for this moment to go.
How on earth am I supposed toget my arms around this massive
mishmash of information?
(09:53):
And today we finally, like fora number of us, we're going oh,
this is the way right, we'redoing a research project right
now in marketing ops.
We've We've decided to breakapart our state of the mo, pro
research, to go deeper on somesubjects, and one of them is on
this area of the state of go tomarket data.
(10:14):
And, to your point, michael,it's fascinating that, like
we're still faced with so muchof these challenges so far, the
responses to the question of thetop three priorities in the
next 12 months.
Number one is unifying dataacross platforms, so like data
integration in general, whichmakes a ton of sense.
Here we are.
(10:34):
Number two is I'm sorry, dataquality and hygiene is number
one.
Number two is unifying dataacross the platforms.
And number three, so far in thestate and the research, is
automating reporting anddashboards.
Are you kidding me?
This all sounds like 10 yearsago.
Michael Hartmann (10:52):
So a thousand and ten, yeah, that was like
what?
Mike Rizzo (10:55):
We've been doing this for a long time.
Why are we still worried aboutautomating?
Like that's crazy and yet it'sstill a priority.
And like I'm not dissinganybody for not having figured
this out.
I mean, I still pull manualreports too, but I think that's
exactly why you're trying to towork on forward, right, like
you're saying hey, let me helpyou get started, let me start
(11:16):
showing you a path, right?
Kobi Stok (11:19):
yeah, yeah, exactly I , yeah, I think that you're
right and I think that theintuition that we had as
operators seven years ago and 10years ago, I think that the
intuition was very good.
Yes, it's like like when you'recalling customer service today,
(11:40):
uh, you get like this messagehey, we are recording this call
to improve our service.
Blah, blah, blah.
Right, no one is going throughthis call to improve right.
But now, but now, maybe rightNow, they have the ability to
actually do that.
So I think the intuition ofcollecting data right, it's like
(12:03):
me buying sneakers ofcollecting data right, it's like
me buying sneakers.
I don't know when I'm going toactually wear them, but I'm
buying just in case and I willfind, like the best you know
event or occasion to kind of usethem.
So I think that people usedlike the intuition was right,
(12:23):
let's collect data, we'll figureit out later.
Right, but I think that thecomplexity, um, also grew, yeah,
and I think that and the volumeand the volume grew the volume
is done.
Mike Rizzo (12:41):
Yeah, the integration layers got easier to
make, it became easier to addmore right, and everybody said I
mean we know from our researchyear over year martechorg has
talked about it, scott Brinker'stalked about it, our
state-of-the-moment researchtalks about it Integration is
the number one thing and it'sproven to propagate this big
challenge of you know.
(13:02):
Okay, now there's even more forus to access but big challenge
of you know.
Michael Hartmann (13:11):
okay, now there's even more for us to
access, but what do I do withthis stuff?
Well, so I have my theory onwhy a lot of companies struggle
on this, because we've talkedabout a lot of reasons why, but
I think it ultimately comes downto you in your point, mike,
about data hygiene and qualitybeing the number one concern, is
that this the fact of it is itgenerally doesn't get the kind
of attention because of at leasttwo reasons.
One it's not.
It's not, uh, uh, immediatelyobvious why it's valuable, right
(13:33):
, so doing that work on cleaningthe data doesn't generate
immediate and obvious results,so it's taking you away from
doing other stuff that is moreobvious, like building more
emails, building more campaigns,et cetera, et cetera, right, so
there's this like just like,how, how visible is it?
Is one.
The second part of that to meis I have yet to meet anyone who
(13:54):
goes our data is great, right.
So I think there's this fearwhen you say, oh, when you go
like, my data is not clean.
So therefore, I probably needto do that before I start doing
any kind of reporting, because Idon't trust the data.
The data is not complete, not,right, whatever the term you use
is, which I keep coming back to, going like, first off, that
(14:16):
assumption that you can't getany value out of the reporting
until your data is quote right,right, is a false one, right, I
think, like by the nature ofjust doing reporting it's gonna,
it's gonna really highlightwhere there are issues, which
then gives you the ammunition orthe case to go like we need to
go fix and clean the data.
(14:36):
And it could be an iterativeprocess, and I think I think
that's part one.
Part two is this desire fordashboards versus something
that's going to give me insightsthat I can make a decision
about.
I don't know that dashboardsare always great at going.
They may give you a good likehow do I feel about the general
health of the organization orwhat I'm doing for go-to-market,
(14:59):
but it doesn't really go like.
What should I do next?
What should I cut?
What should I add?
What should I?
You like?
What should I do next?
What should I cut?
What should I add?
What should I?
You know, you know how should Iadjust, and I don't.
I think this, this cause you gointo like I've gone into
organizations like, oh, we wantyou to do part of your job is
going to be build dashboards.
I'm always like how about we doa few reports first, like let's
get those right, see if they'dhelp us, before we start going
(15:21):
to build a dashboard.
That is not going to probablyhelp us and it's going to be
painful, expensive.
Yeah.
Mike Rizzo (15:28):
And the context that's needed.
I've been in countless SaaSorganizations that have, you
know, both a sales led motionand a PLG led side of the
business, and you know you'retrying to track free trials and
starts and stops and crossoversbetween lifecycle stages and you
inevitably get into a state ofwell, is that person on a paid
(15:51):
account or not?
And, like you know, dashboardsactually create more questions,
which is good.
I'm not saying it shouldn'thappen, but most of the time if
you don't have somebody tryingto filter out the noise and just
say like, if I have to explainwhere the data came from every
single time, then yeah, that's aproblem.
That's a data quality andhygiene issue that needs to be
(16:14):
solved yeah, right, I I yeah,sorry, mike, no, you're fine,
you're fine, I just I.
I was just like thinking throughhow AI's impact on our ability
to get to maybe some more focusand support around.
You know, I've got a couple ofcore priorities and what are the
(16:36):
?
What are the sort of key assetsthat I have in my data lake,
data warehouse, whatever myintegration layers, layers, what
are those that can help pointme in the right direction?
yeah and I think that's thething that's most exciting is to
try to like, use the thing, usethe phrase that I say all the
time right, aim small, misssmall.
Right, like, let's just likefocus on this thing.
(16:58):
Um, yeah, I don't know yourthoughts, yeah, your thoughts,
kobe, on the state of all thisstuff.
Kobi Stok (17:05):
Yeah, first of all, I think that I have so much to
say about dashboards, by the way, but I think that when my team,
my customers, whatever, definean initiative or a task, I try
to divide it into two classes,let's say Strategic or tactic.
(17:30):
I think that for strategicinitiatives, dashboards are, in
some cases are the best outputand people need to see visuals,
people need to see trends,people need to see stuff right.
But because dashboards arecomplex in high-velocity
(17:50):
businesses, now a very smartperson needs to define the
dashboards so they are readable.
It's not a simple thing to doTo build a dashboard.
Forget the technical aspects,but the usability.
It's not a simple thing to doTo build a dashboard, forget the
technical aspects, but theusability, it's not a simple
thing to do.
And second, to read a dashboard, it's not a simple thing to do.
(18:11):
To tell the story behind it,it's not a simple thing to do.
For example, me and Michael canlook at the same dashboard.
Michael can decide let's fire50% of our workforce and I can
decide, thanks for fire 50% ofour root force.
And I can decide Thanks forputting that on me, you know I
try and I can decide let's hire50% more on the same dashboards.
(18:36):
Why?
Because we saw two differentthings.
Because it's not deterministic.
We need to understand what'sgoing on right.
So this stays on the executivelevel, but on the operational
side of the business, thosedecisions can't be made Right.
(18:57):
You want people to makedecisions on the tactical
fragments of the business.
I have a chain problem.
This customer can upsell right,this lead may convert, this
opportunity is at risk, so onand so forth.
Those stuff are today unsolved,agreed.
(19:18):
Unsolved Agreed.
Companies think they solve it,but they just define a set of
rules that they think that arethe reality, but in most cases,
they don't know.
And I think that I also thinkthat there's the approach of
(19:40):
trying to prepare your databefore you know what you want to
do with it.
It's a problem.
People need to change, to flipthe cassette.
Going back to our music before,right.
I think that now people need tobe target-oriented rather than
(20:02):
think about how they would solveit.
Michael Hartmann (20:08):
I want to make sure I understand what you mean
by that Target-oriented versushow they want to solve it.
Kobi Stok (20:13):
So let's say that we have a problem in the business,
let's say churn, whatever, andwe want to figure out what's
causing churn, instead of tryingto guess the data points and
whatever.
Let's first understand if weknow how to identify churn in
(20:35):
our data, ie if we have a fieldthat's called churn, yes or no
or whatever the case is.
Obviously you don't have this,but I don't know like renewal
opportunity, closed loss,whatever.
Let's make sure that thisprocess every time that someone
churned, we have this in thedata, we know it's churned.
Right, when we think like that,then we can analyze it better,
(21:00):
right?
Then what we need to do, whatpeople need to do, is reverse
engineering and not forwardengineering, meaning let's trace
backwards what caused thisspecific event that we know how
to measure.
Obviously, it's hard.
It's CRM.
(21:22):
People will tell you hey,listen, it's CRM's crm.
People will tell you hey,listen, it's crm.
You know reps are not fillingthe data correctly.
You can solve this as well, butbecause it's so much data today
, so even if 20 are not fillingthe data correctly, or 30% and
70% does this right, you willget the answers right Because
(21:47):
you have so much data and everytime that a customer or prospect
survey, are coming to me, to us.
Everyone is like look, my datais shit.
I know that, I know that for afact and I'm telling them for
sure.
The question is what's theratio?
Michael Hartmann (22:07):
Yeah, you bring up such a good point.
Kobi Stok (22:09):
What's the use case right?
Michael Hartmann (22:12):
I love that you bring up this point of you
pick the ratios 30% bad, 40% badbut you need to.
I think this is like themindset shift of going like but
that means that 70 is still good, or 60 is still good or 80 like
, what, like that's the partthat I think people get caught
up and he caught up in the.
This is the portion that'squestionable, bad, invalid,
(22:32):
whatever, as opposed to how muchis right, and I think to me,
this is what I get again getthis data quality and going like
this is what's stopping peoplefrom moving forward with this
kind of stuff.
Kobi Stok (22:44):
Yeah, because the problem is that people can't
measure it today.
They don't know how much.
How bad, slash good is yourdata?
By the way, my enterprise datacan be good for leads closing
accounts, going in, bad forreturn whatever, right.
So even today, when you have somany touch points, you have PNG
(23:05):
, right.
You have the calls all recorded, right.
You have the emails withtracking calls all over the
place.
You know everything.
So if you're using like simpleplatforms, right.
If you use SSDC, HubSpot,mercado, you already have most
of it and people don't utilizeit.
(23:27):
People pay license, people paybig money and they don't utilize
the data.
And then they come to us.
They say we have a problem,right, and then we kind of get
solve the problem faster.
It's not like we kind of getsolve the problem faster.
It's not like we investedanything new, but we just
(23:47):
automated like a workflow thattook so long, it was super
complicated.
We just we broke it down topieces, to 80% pieces and we
just simplified it.
So, instead of 10 people andnine months with a failure rate
of 90% by the way, the fail rateat enterprises of internal data
science projects is 90%.
That's crazy.
(24:10):
It's crazy.
We had a prospect.
I won't mention their name andthey are figuring out health
score.
Health score is like anindicator that indicates the
health of a customer right andthey spend six months, two
people, building a health score,two data scientists.
After six months they stop andthey check the correlation of
(24:34):
the health score to churn nocorrelation.
They're like, oops, six months,two people.
Michael Hartmann (24:43):
They're like oops, six months, two people
Plus technology and otherresources.
Mike Rizzo (24:47):
right yeah all kinds of layers.
Kobi Stok (24:51):
Even that, even that, right Today, with the public
markets, with AI, withcompetition, you can't afford
that In 2021, in 2018, like youknow, I can throw 10 people in a
problem.
If they want solve it, noworries, I have like it's, it's
really better right, yeah allgood, zero interest, I'll get
(25:13):
more money from my vc.
But now people need to changehow they think and and if people
want hyper automate, notautomate, hyper automate, right.
If people won't understand howto consume and how to utilize
their data, they will lose thebattle.
Michael Hartmann (25:33):
Yeah.
So, kavi, to me, this is whatI'm really interested to hear
when we start talking about whatyou're doing with Forward.
I've always thought about thisas like to get reporting
analytics in go to marketactivities, in particular for
B2B, where it's complicated andmessy and incomplete, all those
things we talked about.
(25:53):
There's sort of two componentsthat I've been hopeful that AI
would be a part of solving.
One is you know pulling,helping to pull that data
together and normalize it andclean it up kind of in an
automated way.
That right now requires usuallytypically a lot of human
capital and technology, and youknow exception rules and all
(26:14):
that kind of stuff right andgetting it cleaner over time.
The other part is, I think whatyou just touched on right, you
hired two data scientists to toto build out a hypothesis and
then go test it, as opposed toyou know letting and this is my
ideal state, I think is that youhave ai that can go.
Rather than testing ahypothesis, it can go like
(26:37):
identify the patterns that youmight not see or be able to come
up with on your own, to thentest, and that might be able to
do that more effectively.
I still think there's a placefor humans in this process to be
able to make sense of what isgenerated, because it still may
be nonsense or something youcan't action on but like are you
(26:58):
seeing?
Is that are you addressing oneor both of those generic sides
of this problem?
Kobi Stok (27:06):
Yeah, good comments.
I think that we're solving allof that, from raw data to
predictive models, tosegmentation models, to
forecasting models up throughactivating your CRM and creating
alerts and all of that and thereason that we are doing
everything, because thecomposition and having
(27:30):
everything in one piece, that'sthe value that we bring to the
table.
That's the value thataccelerates the process, because
if we were doing only one thing, we didn't get to the finish
line.
It will be stuck in the middlein the organization, right?
So I, with ai, the, the datamodeling piece, as you're
(27:53):
referring to multiple datasources, how do we integrate the
data modeling thing is reallyum in a way that it will be much
more simpler not fullyautomated, but much more simpler
, right?
So, mike, could write, like youknow, a prompt, maybe drag some
(28:14):
elements on the screen and setto okay, now, in every step that
you are doing as a builder,right, you get the human in the
loop.
In every step that you're doingas a builder, right, you get
the human in the loop, meaningyou're validating every step,
because in some cases, you wouldneed to tweak or tune the
(28:34):
initial output that the AI did.
In some cases, you'll be able totweak it using prompts and in
some cases prompts won't helpyou, because when you go low
resolution, in most casesprompts will kind of um, it's
like um I'm trying to find an uhanalogy, but it's like in golf,
(28:55):
right, when you're so close tothe hole and you're and you're
trying to make it, maybe theball will be much far from the
hole.
Then you started.
This is how I, this is howprompts are, with really small
refinements, right, and in somecases you would need to do those
small refinements in a UI,right In a dropdown or in a box
(29:20):
or in a drag and drop quiz,whatever, right.
So that's number one.
Right On the data modelingpiece, number two what you asked
on the actual like human in themiddle, of course, the AI, or
(29:43):
your data or statistics or mathdon't know anything about your
business or statistics or mathdon't know anything about your
business.
They don't know that.
The field that you've used inAppSpot this is just a copy of
another field that's used forunit testing, whatever.
Mike Rizzo (30:00):
I've been trying to say this to people for so long.
You articulated that reallywell, great example.
Michael Hartmann (30:07):
I was just like.
Every Salesforce instance I'vebeen involved with has had some
existing field that is supposedto be used for one thing be used
for something else because theprocess of adding a new field
and all the integrations was sopainful.
But if you didn't know that youcould misinterpret that data?
Kobi Stok (30:28):
Yeah, but the thing is is that you are now in a
position to validate the dataand to validate the points much
better than you were before, andit's crazy.
It's crazy.
Now you can actually do it.
I see people like marketing ops, rev ops, rev ops, cs ops
(30:51):
making an impact, like oneperson making an impact of 10.
I'm seeing it day in and dayout.
It's really crazy what you cando with utilizing new technology
and rethink and think differentabout how we would solve it.
Mike Rizzo (31:13):
To use it, and that's yeah, and that's the part
that, like I've been definitelyon the pedestal about, without
a doubt, is the opportunity forfolks in this type of a role
marketing operations, rev ops,sales ops, cs ops, whatever ops
(31:35):
role you have.
I love the opportunity for themarketing ops folks because, in
large part, you tend to have alot more access to a lot more
touch points than the otherfunctional areas yeah, more
touch points than the otherfunctional areas.
Yeah, everything from you knowanonymized, de-anonymized,
(31:58):
visitors and tracking at the topof the funnel through.
Did somebody become a closedone customer or churn?
Yeah, oftentimes you get tointeract at some level with
product data as well, becauseproduct marketers want to know
usage and adoption of featuresand all those things.
So and I think you made acomment earlier, kobe, about you
know, starting with this,ensuring we understand the steps
(32:22):
and the process to identifychurn right, and that we're all
in agreement that those are thethings that we think or need to
have in place to identify churnto then go after solving solving
the problem, you know later, um, and that is no one else, like,
(32:44):
no one else can have theseconversations like, like, really
, truly no one is as close tothe systems as you are in
marketing operations and revenueoperations in general, using
that term broadly.
All you ops folks out there,yeah, it is a really, really
exciting time for you to go makean impact and I love that.
(33:05):
You just said you had one personeffectively making the impact
of 10, right, and I think rightnow is the time to you know
whatever, whatever phrase youwant to use, grab the bull by
the horns or seize the day orwhatever.
Absolutely sink your teeth into the art of the possible and
(33:29):
I'm going to say again startthinking like a product manager
and you know you're going tohear it over and over again from
our community and for me, andyou're all going to probably get
annoyed with me at some pointbut you are the product manager
of the go to market tech stackand your job is to figure out
(33:51):
how to best leverage all ofthese pieces of information to
enable your internal teams to goto market, to enable your
buyers who are buying yourproducts and services to engage
with your brand and yourinternal teams in a meaningful,
effective and scalable way, andyou also, unfortunately for you
(34:12):
have to be aware of is the brandbeing represented properly and
are we legally compliant?
Those are a lot of fun thingsthat you have to manage, but
they're all stakeholders thatyou have visibility across and
it sounds daunting, but forthose of you that are in this
role and you're having fun withit it's actually really exciting
, yeah.
Michael Hartmann (34:33):
I would add a third one.
Mike Rizzo (34:34):
That's how I translated for Kobe yeah, go
ahead.
Michael Hartmann (34:36):
Yeah, I would just add a third one to your
list there, which is being anadvocate for the customer,
because I think sometimes thatgets lost.
I mean, I don't know about you.
Mike Rizzo (34:46):
Sorry if I didn't say that clearly.
Yeah, you're absolutely rightIf I didn't say it clearly
enough when I said hey, you haveto service the internal team
members as well as the customersthat are buying your products
and services.
Yeah, that's the part where youneed to hone in on that
customer experience.
How are they able to actuallyengage with us in an effective
way?
(35:06):
That's also brand safe andcompliant right.
Brand safe and compliant Right.
Kobi Stok (35:09):
I would even add two more things in general that I
think that are must-haves forevery ops person out there, and
especially these days.
I think the number one is thinklike a builder.
Think like a builder.
That's what you need to do.
That's the mindset that youneed to be in in order to be
(35:31):
very successful.
Right, you need to build stuff.
You don't need to build anotherlike, build like.
You need to orchestrate thosesystems together.
So, yeah, right.
And second, it's really workout, because I think that, again
, doing the impact of 10 people,if you are successful in
(35:57):
achieving that impact, I thinkthat's the whole grand for
everyone at Ops.
And I think that Ops gets morelike a center stage not
necessarily the singer of theband, but maybe the bass player
or the lead guitarist stillreferring to our earliest
conversation Yep, but that's ascenter as they would, as they
(36:20):
ever end and currently and Ithink it will be even even
further down the road, becausethis is the platform that the
business runs on yeah, Withoutit, it won't run.
It won't run without it.
Michael Hartmann (36:37):
Yeah Well, I think it's interesting that
you're seeing a 10x productivitygain using some of these tools,
specifically around dataanalytics and science reporting,
whatever you want to call it,because I've believed for a long
time and this is why I've beenreally hopeful that AI would
play a part in making this Idon't even want to say easier,
(36:57):
but more scalable is because Ibelieve for a long time that it
is an effort-based thing to doright.
It takes time, effort and someexpertise to be able to do
effective data analysis andreporting and to get insights
that can then be used to impacta business and so go ahead.
(37:18):
No, no, I think that's what I'mreally excited about is hearing
that this is maybe somethingthat's actually happening.
Mike Rizzo (37:25):
I think it is, and it's funny here a word that we
very rarely use in our uh, sortof day-to-day world here I
suppose I don't think I hear itall that often is creativity.
And and this is an interestingtime because, while we don't
(37:45):
fall into that creative marketercategory, the one that we all
think of traditionally, right,right All, the fun ads and the
design and the and the coolcampaign ideas Although I will
say there's a lot of marketinghouse people that come up with
great ideas like campaign wise.
Okay, um, this type ofcreativity is it's a jigsaw
(38:08):
puzzle that you are, you arestitching together, you're
actually making the jigsawpuzzle that you are, you are
stitching together, you'reactually making the jigsaw
puzzle while you complete it,like all at the same time.
Yeah, and it's, it'sfascinating, right, it's like
you're seeing possible, you'reasking questions that could lead
to down paths, that that couldanswer questions that the
(38:30):
business never even knew it had.
Right, and I just I think it'sa new type of creativity that,
um, in some ways, is veryfreeing it keeps coming.
Michael Hartmann (38:41):
I can't go back to one of our earliest
guests, brandy sanders, talkingabout how, like thinking about
this like chess, right?
So I didn't even go beyond apuzzle, right, because it's it's
a puzzle that has made probablymultiple dozens of potential
solutions and I think that's,yes, how I think about it.
Um, but here's my, here's myconcern.
(39:02):
It could be maybe you can chimein here too, because I've said
this many times here, but I lovethe idea that the technology
could make it so that someonewho just doesn't have the
bandwidth to be able to providethose insights, but has access
to the data and understands itand the flaws in it and
everything else is that I amconcerned that there's not
enough knowledge or expertise onhow to actually interpret the
(39:25):
analytics and statistics, orwhatever you want to call it,
across most organizations andeven in inside ops.
So are you seeing some of thesame stuff with your clients and
customers or in your ownexperience, and do you have any
thoughts on how to address that?
Kobi Stok (39:42):
yeah, that's that.
That's um.
That's a very interesting uh umthing that you said just,
Michael.
I think that talented peopleacross any profession are really
hard to find.
That's in general.
I think that analysts and datascientists are really unique
(40:08):
because they need to understand.
They need to have technicalchops right, but in order to
really find insight, they needto understand the business.
The combination of one personbeing super technical and
understanding the business isunique.
What were my theory right?
(40:31):
And we are building the productas we built an AI agent inside
of the product.
That's an analyst.
We built an AI agent inside ofthe product.
That's a scientist, Meaning wekind of lead the ops person to a
journey where we ask them thequestions.
(40:55):
Right, there's a think about avery precise process that they
go through and then the AIexecute the job of an analyst
and scientist, because theproblem is that in most cases,
(41:18):
you don't have enough of them ata specific point.
So every question that you asktoday let's say we don't have AI
let's go back three years orfour years Every question that
you ask on the go-to-marketfront, you would need someone
else to help you around.
You couldn't answer questionson your own.
Michael Hartmann (41:48):
Yeah.
Kobi Stok (41:50):
I mean, it's so complex, right, in most cases
you need to export data, thenyou need to have a Python
notebook and then blah, blah,blah.
It's really complex.
Yeah, right, but today the sameprocess is happening, but
someone else is driving thewheel and it's not a person.
(42:10):
So what we're seeing is that weare freeing, we are freeing the
organization from thoseheadcounts that are expensive
and can't operate in the twofronts of the data and the
business, and we're enabling thebusiness to run it.
(42:31):
Now a little kind ofclarification here For the very
strategic initiatives, ai is notstill there.
But for the tacticalinitiatives, let's understand
why customers churn, let'sunderstand apps Like all those.
Ai can definitely do this, andwe do day in, day out.
(42:53):
So I hope that I answered thequestion, michael.
Michael Hartmann (43:01):
I think so.
I'm sure I think we couldprobably carry on for quite a
bit longer, but I think we'regoing to have to wrap it up here
.
Kavi, if there's, if folks wantto keep up with you, connect
with you or learn more aboutwhat you're doing at Forward AI,
what's the best way for them todo that?
Kobi Stok (43:18):
Oh, so maybe before I answer that, I want to add one
sentence on the data hygienepiece that Mike mentioned.
So while we're implementingforward, we automate it in
reprised.
So there are many problems thatwe see along the way.
What Mike said about dataquality or data hygiene, that's
(43:47):
the number one problem that wesee as well in every case out
there, Every case out there, andI'm excited to that's the first
time that I talk about it butwe are launching an agent that
puts data hygiene on autopilot.
Mike Rizzo (44:06):
Okay, I like it All right.
Kobi Stok (44:09):
Okay, I like it.
All right.
So every data hygiene problem,you can use the agent to solve
it.
I'm not sure it can solveeverything right, but we mapped
200 problems that we have heardfrom our customers and prospects
(44:29):
and we solved all of them.
Not saying that it will be ableto do everything, but that's
just that.
That's a new product that we'relaunching soon.
We are super excited.
We have better users and theresults are amazing.
(44:51):
If you want to um kind of um, uh, I'm learning about forward, so
it's forward or the I, withoutthe a, um and um.
You're welcome to follow us andmyself on linkedin where we
talk about ai and the differencebetween an LLM to machine
learning and I think thateducating the people around what
(45:13):
is AI because AI is big, it'sfrom the 50s, it's like you know
, it's like guitar.
Innovations are happeningacross the board and people
don't understand what AI is,what LLM is, what OpenAI can do,
what they can't do, what thedifference between an OpenAI to
an entropic, I don't know.
Deepseq, all of this peopleneed to.
(45:35):
And again, ops I encourage Opspeople to really be curious and
learn more about it and I inviteyou to follow us myself on
LinkedIn, you know, on all thosechannels.
Mike Rizzo (45:49):
Fantastic.
Yeah, I'm really excited aboutwhat you're doing at Forward,
kobe, and I'm glad we were ableto reconnect this year and have
you on the podcast, and we'llprobably, I know we have a few
more things lined up to in termsof educating the audience.
So, for those of you listening,stay tuned to what Forward's
doing.
It's F-O-R-W-R-D, that's howyou spell it.
(46:13):
Uh, you can go visit thewebsite at forwardai and um and
then pay attention to some ofthe other posts, emails and
slacks and social things thatwe're doing on the marketing
upside, because you're going tohear a few more different
renditions and going deeper indifferent areas around AI and
data that forward's sort ofhelping to solve.
And then Kobe I got one ofthese days when we get a chance
(46:35):
to sit down for a little longer.
I got to show you what I wasdoing with audit hub as a
product that I built.
It was totally around this datahygiene thing and I feel like
you probably already solved it,but I'm so excited by what
you're what you're telling us.
Kobi Stok (46:49):
I am too, we.
Michael Hartmann (46:51):
but I'm so excited by what you're telling
us, I am too.
Kobi Stok (46:53):
We will meet and you will show me Promise that Sounds
good, kavi, thank you.
Michael Hartmann (46:56):
Thanks for staying up late for us and, as
always, mike, thank you forjoining Absolutely.
And thanks to our audience forcontinuing to support us.
And if you have suggestions fortopics or guests or want to be
a guest, feel free to reach outto naomi, mike or me.
We'd be happy to chat with youabout it until next time.
Kobi Stok (47:12):
Bye, everybody bye, bye guys.
Hello everyone .
Welcome to another episode ofOpsCast brought to you by
MarketingOps.
com, powered by all the MoProsout there.
I'm your host, michael Hartman,joined today by Mike Rizzo.
Say hello, mike.
Mike Rizzo (00:10):
Hey, what's happening?
This is a new thing for us.
We're going to do video guys.
Michael Hartmann (00:15):
I know right.
Mike Rizzo (00:16):
MoPros, all you people out there, we're going to
use video from now on.
Michael Hartmann (00:27):
I don't know how this is going.
I've got a face for radio, youknow.
So tying into the music talk wehad before we started recording
, all right.
So joining us today to talkabout the potential impacts of
AI in business operations isKobe Stock.
Kobe is a serial entrepreneurwith two decades of experience
in building software.
He currently leads Forward AI,an automated data science
platform built for RevOps.
Previously, he was with WalkMe,where he led the global product
(00:48):
strategy with a focus on scale,growth and innovation.
He joined WalkMe through theacquisition of Abbeyio, a mobile
AI company, where he wasco-founder and CTO, leading the
product and technology.
Prior to WalkMe, he was thefounder and CEO of a consumer
music tech company with millionsof users.
Early in his career, he was asoftware engineer, architect and
a manager at multiple startupsas well as SAP.
(01:09):
So, kobe, thanks for joining ustoday Late in the evening for
you.
Kobi Stok (01:14):
Hello, hello.
It's a good thing that I justchanged my camera as we're doing
video, so full HD here.
There you go, full transparency.
Michael Hartmann (01:23):
If I showed my background fully, you would see
the in-process taking down ofour holiday decorations.
Mike Rizzo (01:29):
So I had to move a few things before Nice Stuff
that I haven't had to deal within the past.
Michael Hartmann (01:37):
So all good.
So you founded Forwardai nottoo long ago.
It was recently.
You started it.
You call it an automated datascience platform.
So for us and our listeners,what does that mean to you and
how should we be thinking aboutthat?
Kobi Stok (01:52):
Yeah, so actually every time.
So that's my third company thatI'm starting right, and every
time that I start a company Itry to, before I check the
market and check the need, I tryto solve my own problem.
That's like number one.
And in my previous role we weretrying to build a model that
(02:14):
will predict churn based onproduct usage.
Think about it we work for acompany collecting tons of data
and we're trying to build amodel that will basically use
this data and actually surfaceaccounts that are at risk.
So we get a heads up and we cancreate some sort of a
mitigation.
(02:36):
And it took us a year to do it Ayear.
It took us a year, one year todo it.
It and it was complex.
We needed to involve like atleast 20 people.
Half were technical, half werefrom the business.
You know we've talked about itso much and the process was so
(02:59):
long right, but when we finishedit after a year plus even and
we launched it, then what we did?
We basically surfaced thoseinsights to the customer success
team in our case, and theimpact was amazing.
It really changed how thecustomer success team operates
on the day-to-day, not on aspecific moment, and even
(03:22):
afterwards, we modified thecompensation of the reps based
on the model.
So I saw the impact and I sawthe investment and I said it's
too long to make thoseinvestments.
I mean, most companies won'teven start.
And if you think about it,companies are collecting so much
(03:45):
data these days and the datacosts money.
It takes time to arrange it, ittakes an effort to arrange it.
They're building tons ofdashboards right, but by the end
of the day, people are notreally data driven in the day to
.
People are not reallydata-driven in the day-to-day.
Michael Hartmann (04:07):
No.
It's just something I think alot of people either claim that
they especially marketers claimthat they either are or want to
be data-driven.
I'm not even sure they knowwhat that means.
Yeah.
Kobi Stok (04:20):
The best-case scenario is that companies have
good management dashboards thatprovide visibility on the
C-level, right On the leadershipteam, but when you go down to
the people who actually managethe business right Manage the
(04:41):
product, the day-to-day with thecustomers, support finance they
don't have this informationright, and they don't have this
information because it's reallyhard.
Yeah, and let's imagine that acompany has unlimited resources
(05:02):
of data science internally, then, using data science methods,
they wouldn't be able to surfacethose insights or data or
whatever, right, but they don'thave it.
So automating data sciencemeans that we are selling you a
piece of software that acts likehundreds of data science
(05:27):
professionals that are workingon your data to build models
that are fueling businessprocesses in your organization,
to make them all automated, tosurface insights or signals that
(05:47):
will help the team to performbetter, to surface allots or
emails right, with specific callto actions to improve
forecasting and basically everybusiness KPI that you can
imagine.
So that's what we do today.
That's interesting.
Michael Hartmann (06:06):
What really strikes me as interesting is
that when I first made mytransition from doing IT
management consulting mostlywith accounting and finance,
real estate type environmentsinto marketing, it was to do
database marketing so datawarehousing into marketing.
It was to do database marketingso data warehousing.
(06:27):
And we had a team of today thatwould be called data scientists
but you know statisticians whodid things like churn prediction
and likelihood to convert andall those kinds of models.
This was a big telecom companyand it was.
My job was to build a databaseto help with go-to-market
activity, so a 50 millionhousehold database and the data
(06:47):
processing and the amount ofdata that we had to acquire and
then analyze was huge and itcost a lot of money.
It was a huge effort.
I mean what you're describing.
It's almost surprising to methat that is still a challenge
at many places.
Kobi Stok (07:05):
You know what I mean Very, very big challenge.
I think that data became morecomplex.
So every time that someone, anycompany, improved something,
then another complexity layerthing happened and then it
became more complex.
Improved something, thenanother complexity layer thing
happened and then it became morecomplex.
(07:28):
And I think that in the past,let's say five to seven years,
we're in a race.
We were in a race before thatof hey, I need to collect data,
right, everyone was in a race, Ineed to collect more data, I
need to, right.
And now everyone stoppedcollecting data.
They have enough data.
Now, how do I get ROI on thosehuge investments that we just
(07:54):
did?
And we bought Snowflake and webought Salesforce and we have
HubSpot and we have Marketo andwe have Segment and we have
product analytics, like so manythings that essentially so much
value, right, there's inside somany things you can drive, but
you just don't have theresources and it's so complex to
(08:14):
even start.
So my goal was how do I makepeople start?
Forget about the price.
Mike Rizzo (08:23):
Yeah, yeah, forget about right.
Yeah, yeah, I think, and it'sso, um, it's so interesting to
be at this intersection of thislike data aggregation and
collection phase that we've beengoing through for the last
couple decades.
Where it was it, I don't know.
(08:45):
I think for some companies it'sgoing to feel a little bit like
a retirement plan, like you'vebeen piling this data away for a
while and eventually it'll havea payoff Hopefully a good one,
exactly and for others it'llfeel like let's just scrap the
last 10 years and then take themost recent five, and you know
(09:09):
so we'll look at the last 15 andtake the most recent five and
roll with that, because ourbusiness changed a lot.
But there's this, the.
The opportunity now iseverybody went.
It's a little bit of a cop out.
I don't know, maybe, maybe,maybe you don't feel this way,
kobe, but it's a little bit of aoh well, now I don't have to
(09:32):
think quite as hard Like AI'shere.
So cool, let's figure out howto use the data now, which I
think is great, right?
We've all kind of just beenwaiting for this moment to go.
How on earth am I supposed toget my arms around this massive
mishmash of information?
(09:53):
And today we finally, like fora number of us, we're going oh,
this is the way right, we'redoing a research project right
now in marketing ops.
We've We've decided to breakapart our state of the mo, pro
research, to go deeper on somesubjects, and one of them is on
this area of the state of go tomarket data.
(10:14):
And, to your point, michael,it's fascinating that, like
we're still faced with so muchof these challenges so far, the
responses to the question of thetop three priorities in the
next 12 months.
Number one is unifying dataacross platforms, so like data
integration in general, whichmakes a ton of sense.
Here we are.
(10:34):
Number two is I'm sorry, dataquality and hygiene is number
one.
Number two is unifying dataacross the platforms.
And number three, so far in thestate and the research, is
automating reporting anddashboards.
Are you kidding me?
This all sounds like 10 yearsago.
Michael Hartmann (10:52):
So a thousand and ten, yeah, that was like
what?
Mike Rizzo (10:55):
We've been doing this for a long time.
Why are we still worried aboutautomating?
Like that's crazy and yet it'sstill a priority.
And like I'm not dissinganybody for not having figured
this out.
I mean, I still pull manualreports too, but I think that's
exactly why you're trying to towork on forward, right, like
you're saying hey, let me helpyou get started, let me start
(11:16):
showing you a path, right?
Kobi Stok (11:19):
yeah, yeah, exactly I , yeah, I think that you're
right and I think that theintuition that we had as
operators seven years ago and 10years ago, I think that the
intuition was very good.
Yes, it's like like when you'recalling customer service today,
(11:40):
uh, you get like this messagehey, we are recording this call
to improve our service.
Blah, blah, blah.
Right, no one is going throughthis call to improve right.
But now, but now, maybe rightNow, they have the ability to
actually do that.
So I think the intuition ofcollecting data right, it's like
(12:03):
me buying sneakers ofcollecting data right, it's like
me buying sneakers.
I don't know when I'm going toactually wear them, but I'm
buying just in case and I willfind, like the best you know
event or occasion to kind of usethem.
So I think that people usedlike the intuition was right,
(12:23):
let's collect data, we'll figureit out later.
Right, but I think that thecomplexity, um, also grew, yeah,
and I think that and the volumeand the volume grew the volume
is done.
Mike Rizzo (12:41):
Yeah, the integration layers got easier to
make, it became easier to addmore right, and everybody said I
mean we know from our researchyear over year martechorg has
talked about it, scott Brinker'stalked about it, our
state-of-the-moment researchtalks about it Integration is
the number one thing and it'sproven to propagate this big
challenge of you know.
(13:02):
Okay, now there's even more forus to access but big challenge
of you know.
Michael Hartmann (13:11):
okay, now there's even more for us to
access, but what do I do withthis stuff?
Well, so I have my theory onwhy a lot of companies struggle
on this, because we've talkedabout a lot of reasons why, but
I think it ultimately comes downto you in your point, mike,
about data hygiene and qualitybeing the number one concern, is
that this the fact of it is itgenerally doesn't get the kind
of attention because of at leasttwo reasons.
One it's not.
It's not, uh, uh, immediatelyobvious why it's valuable, right
(13:33):
, so doing that work on cleaningthe data doesn't generate
immediate and obvious results,so it's taking you away from
doing other stuff that is moreobvious, like building more
emails, building more campaigns,et cetera, et cetera, right, so
there's this like just like,how, how visible is it?
Is one.
The second part of that to meis I have yet to meet anyone who
(13:54):
goes our data is great, right.
So I think there's this fearwhen you say, oh, when you go
like, my data is not clean.
So therefore, I probably needto do that before I start doing
any kind of reporting, because Idon't trust the data.
The data is not complete, not,right, whatever the term you use
is, which I keep coming back to, going like, first off, that
(14:16):
assumption that you can't getany value out of the reporting
until your data is quote right,right, is a false one, right, I
think, like by the nature ofjust doing reporting it's gonna,
it's gonna really highlightwhere there are issues, which
then gives you the ammunition orthe case to go like we need to
go fix and clean the data.
(14:36):
And it could be an iterativeprocess, and I think I think
that's part one.
Part two is this desire fordashboards versus something
that's going to give me insightsthat I can make a decision
about.
I don't know that dashboardsare always great at going.
They may give you a good likehow do I feel about the general
health of the organization orwhat I'm doing for go-to-market,
(14:59):
but it doesn't really go like.
What should I do next?
What should I cut?
What should I add?
What should I?
You like?
What should I do next?
What should I cut?
What should I add?
What should I?
You know, you know how should Iadjust, and I don't.
I think this, this cause you gointo like I've gone into
organizations like, oh, we wantyou to do part of your job is
going to be build dashboards.
I'm always like how about we doa few reports first, like let's
get those right, see if they'dhelp us, before we start going
(15:21):
to build a dashboard.
That is not going to probablyhelp us and it's going to be
painful, expensive.
Yeah.
Mike Rizzo (15:28):
And the context that's needed.
I've been in countless SaaSorganizations that have, you
know, both a sales led motionand a PLG led side of the
business, and you know you'retrying to track free trials and
starts and stops and crossoversbetween lifecycle stages and you
inevitably get into a state ofwell, is that person on a paid
(15:51):
account or not?
And, like you know, dashboardsactually create more questions,
which is good.
I'm not saying it shouldn'thappen, but most of the time if
you don't have somebody tryingto filter out the noise and just
say like, if I have to explainwhere the data came from every
single time, then yeah, that's aproblem.
That's a data quality andhygiene issue that needs to be
(16:14):
solved yeah, right, I I yeah,sorry, mike, no, you're fine,
you're fine, I just I.
I was just like thinking throughhow AI's impact on our ability
to get to maybe some more focusand support around.
You know, I've got a couple ofcore priorities and what are the
(16:36):
?
What are the sort of key assetsthat I have in my data lake,
data warehouse, whatever myintegration layers, layers, what
are those that can help pointme in the right direction?
yeah and I think that's thething that's most exciting is to
try to like, use the thing, usethe phrase that I say all the
time right, aim small, misssmall.
Right, like, let's just likefocus on this thing.
(16:58):
Um, yeah, I don't know yourthoughts, yeah, your thoughts,
kobe, on the state of all thisstuff.
Kobi Stok (17:05):
Yeah, first of all, I think that I have so much to
say about dashboards, by the way, but I think that when my team,
my customers, whatever, definean initiative or a task, I try
to divide it into two classes,let's say Strategic or tactic.
(17:30):
I think that for strategicinitiatives, dashboards are, in
some cases are the best outputand people need to see visuals,
people need to see trends,people need to see stuff right.
But because dashboards arecomplex in high-velocity
(17:50):
businesses, now a very smartperson needs to define the
dashboards so they are readable.
It's not a simple thing to doTo build a dashboard.
Forget the technical aspects,but the usability.
It's not a simple thing to doTo build a dashboard, forget the
technical aspects, but theusability, it's not a simple
thing to do.
And second, to read a dashboard, it's not a simple thing to do.
(18:11):
To tell the story behind it,it's not a simple thing to do.
For example, me and Michael canlook at the same dashboard.
Michael can decide let's fire50% of our workforce and I can
decide, thanks for fire 50% ofour root force.
And I can decide Thanks forputting that on me, you know I
try and I can decide let's hire50% more on the same dashboards.
(18:36):
Why?
Because we saw two differentthings.
Because it's not deterministic.
We need to understand what'sgoing on right.
So this stays on the executivelevel, but on the operational
side of the business, thosedecisions can't be made Right.
(18:57):
You want people to makedecisions on the tactical
fragments of the business.
I have a chain problem.
This customer can upsell right,this lead may convert, this
opportunity is at risk, so onand so forth.
Those stuff are today unsolved,agreed.
(19:18):
Unsolved Agreed.
Companies think they solve it,but they just define a set of
rules that they think that arethe reality, but in most cases,
they don't know.
And I think that I also thinkthat there's the approach of
(19:40):
trying to prepare your databefore you know what you want to
do with it.
It's a problem.
People need to change, to flipthe cassette.
Going back to our music before,right.
I think that now people need tobe target-oriented rather than
(20:02):
think about how they would solveit.
Michael Hartmann (20:08):
I want to make sure I understand what you mean
by that Target-oriented versushow they want to solve it.
Kobi Stok (20:13):
So let's say that we have a problem in the business,
let's say churn, whatever, andwe want to figure out what's
causing churn, instead of tryingto guess the data points and
whatever.
Let's first understand if weknow how to identify churn in
(20:35):
our data, ie if we have a fieldthat's called churn, yes or no
or whatever the case is.
Obviously you don't have this,but I don't know like renewal
opportunity, closed loss,whatever.
Let's make sure that thisprocess every time that someone
churned, we have this in thedata, we know it's churned.
Right, when we think like that,then we can analyze it better,
(21:00):
right?
Then what we need to do, whatpeople need to do, is reverse
engineering and not forwardengineering, meaning let's trace
backwards what caused thisspecific event that we know how
to measure.
Obviously, it's hard.
It's CRM.
(21:22):
People will tell you hey,listen, it's CRM's crm.
People will tell you hey,listen, it's crm.
You know reps are not fillingthe data correctly.
You can solve this as well, butbecause it's so much data today
, so even if 20 are not fillingthe data correctly, or 30% and
70% does this right, you willget the answers right Because
(21:47):
you have so much data and everytime that a customer or prospect
survey, are coming to me, to us.
Everyone is like look, my datais shit.
I know that, I know that for afact and I'm telling them for
sure.
The question is what's theratio?
Michael Hartmann (22:07):
Yeah, you bring up such a good point.
Kobi Stok (22:09):
What's the use case right?
Michael Hartmann (22:12):
I love that you bring up this point of you
pick the ratios 30% bad, 40% badbut you need to.
I think this is like themindset shift of going like but
that means that 70 is still good, or 60 is still good or 80 like
, what, like that's the partthat I think people get caught
up and he caught up in the.
This is the portion that'squestionable, bad, invalid,
(22:32):
whatever, as opposed to how muchis right, and I think to me,
this is what I get again getthis data quality and going like
this is what's stopping peoplefrom moving forward with this
kind of stuff.
Kobi Stok (22:44):
Yeah, because the problem is that people can't
measure it today.
They don't know how much.
How bad, slash good is yourdata?
By the way, my enterprise datacan be good for leads closing
accounts, going in, bad forreturn whatever, right.
So even today, when you have somany touch points, you have PNG
(23:05):
, right.
You have the calls all recorded, right.
You have the emails withtracking calls all over the
place.
You know everything.
So if you're using like simpleplatforms, right.
If you use SSDC, HubSpot,mercado, you already have most
of it and people don't utilizeit.
(23:27):
People pay license, people paybig money and they don't utilize
the data.
And then they come to us.
They say we have a problem,right, and then we kind of get
solve the problem faster.
It's not like we kind of getsolve the problem faster.
It's not like we investedanything new, but we just
(23:47):
automated like a workflow thattook so long, it was super
complicated.
We just we broke it down topieces, to 80% pieces and we
just simplified it.
So, instead of 10 people andnine months with a failure rate
of 90% by the way, the fail rateat enterprises of internal data
science projects is 90%.
That's crazy.
(24:10):
It's crazy.
We had a prospect.
I won't mention their name andthey are figuring out health
score.
Health score is like anindicator that indicates the
health of a customer right andthey spend six months, two
people, building a health score,two data scientists.
After six months they stop andthey check the correlation of
(24:34):
the health score to churn nocorrelation.
They're like, oops, six months,two people.
Michael Hartmann (24:43):
They're like oops, six months, two people
Plus technology and otherresources.
Mike Rizzo (24:47):
right yeah all kinds of layers.
Kobi Stok (24:51):
Even that, even that, right Today, with the public
markets, with AI, withcompetition, you can't afford
that In 2021, in 2018, like youknow, I can throw 10 people in a
problem.
If they want solve it, noworries, I have like it's, it's
really better right, yeah allgood, zero interest, I'll get
(25:13):
more money from my vc.
But now people need to changehow they think and and if people
want hyper automate, notautomate, hyper automate, right.
If people won't understand howto consume and how to utilize
their data, they will lose thebattle.
Michael Hartmann (25:33):
Yeah.
So, kavi, to me, this is whatI'm really interested to hear
when we start talking about whatyou're doing with Forward.
I've always thought about thisas like to get reporting
analytics in go to marketactivities, in particular for
B2B, where it's complicated andmessy and incomplete, all those
things we talked about.
(25:53):
There's sort of two componentsthat I've been hopeful that AI
would be a part of solving.
One is you know pulling,helping to pull that data
together and normalize it andclean it up kind of in an
automated way.
That right now requires usuallytypically a lot of human
capital and technology, and youknow exception rules and all
(26:14):
that kind of stuff right andgetting it cleaner over time.
The other part is, I think whatyou just touched on right, you
hired two data scientists to toto build out a hypothesis and
then go test it, as opposed toyou know letting and this is my
ideal state, I think is that youhave ai that can go.
Rather than testing ahypothesis, it can go like
(26:37):
identify the patterns that youmight not see or be able to come
up with on your own, to thentest, and that might be able to
do that more effectively.
I still think there's a placefor humans in this process to be
able to make sense of what isgenerated, because it still may
be nonsense or something youcan't action on but like are you
(26:58):
seeing?
Is that are you addressing oneor both of those generic sides
of this problem?
Kobi Stok (27:06):
Yeah, good comments.
I think that we're solving allof that, from raw data to
predictive models, tosegmentation models, to
forecasting models up throughactivating your CRM and creating
alerts and all of that and thereason that we are doing
everything, because thecomposition and having
(27:30):
everything in one piece, that'sthe value that we bring to the
table.
That's the value thataccelerates the process, because
if we were doing only one thing, we didn't get to the finish
line.
It will be stuck in the middlein the organization, right?
So I, with ai, the, the datamodeling piece, as you're
(27:53):
referring to multiple datasources, how do we integrate the
data modeling thing is reallyum in a way that it will be much
more simpler not fullyautomated, but much more simpler
, right?
So, mike, could write, like youknow, a prompt, maybe drag some
(28:14):
elements on the screen and setto okay, now, in every step that
you are doing as a builder,right, you get the human in the
loop.
In every step that you're doingas a builder, right, you get
the human in the loop, meaningyou're validating every step,
because in some cases, you wouldneed to tweak or tune the
(28:34):
initial output that the AI did.
In some cases, you'll be able totweak it using prompts and in
some cases prompts won't helpyou, because when you go low
resolution, in most casesprompts will kind of um, it's
like um I'm trying to find an uhanalogy, but it's like in golf,
(28:55):
right, when you're so close tothe hole and you're and you're
trying to make it, maybe theball will be much far from the
hole.
Then you started.
This is how I, this is howprompts are, with really small
refinements, right, and in somecases you would need to do those
small refinements in a UI,right In a dropdown or in a box
(29:20):
or in a drag and drop quiz,whatever, right.
So that's number one.
Right On the data modelingpiece, number two what you asked
on the actual like human in themiddle, of course, the AI, or
(29:43):
your data or statistics or mathdon't know anything about your
business or statistics or mathdon't know anything about your
business.
They don't know that.
The field that you've used inAppSpot this is just a copy of
another field that's used forunit testing, whatever.
Mike Rizzo (30:00):
I've been trying to say this to people for so long.
You articulated that reallywell, great example.
Michael Hartmann (30:07):
I was just like.
Every Salesforce instance I'vebeen involved with has had some
existing field that is supposedto be used for one thing be used
for something else because theprocess of adding a new field
and all the integrations was sopainful.
But if you didn't know that youcould misinterpret that data?
Kobi Stok (30:28):
Yeah, but the thing is is that you are now in a
position to validate the dataand to validate the points much
better than you were before, andit's crazy.
It's crazy.
Now you can actually do it.
I see people like marketing ops, rev ops, rev ops, cs ops
(30:51):
making an impact, like oneperson making an impact of 10.
I'm seeing it day in and dayout.
It's really crazy what you cando with utilizing new technology
and rethink and think differentabout how we would solve it.
Mike Rizzo (31:13):
To use it, and that's yeah, and that's the part
that, like I've been definitelyon the pedestal about, without
a doubt, is the opportunity forfolks in this type of a role
marketing operations, rev ops,sales ops, cs ops, whatever ops
(31:35):
role you have.
I love the opportunity for themarketing ops folks because, in
large part, you tend to have alot more access to a lot more
touch points than the otherfunctional areas yeah, more
touch points than the otherfunctional areas.
Yeah, everything from you knowanonymized, de-anonymized,
(31:58):
visitors and tracking at the topof the funnel through.
Did somebody become a closedone customer or churn?
Yeah, oftentimes you get tointeract at some level with
product data as well, becauseproduct marketers want to know
usage and adoption of featuresand all those things.
So and I think you made acomment earlier, kobe, about you
know, starting with this,ensuring we understand the steps
(32:22):
and the process to identifychurn right, and that we're all
in agreement that those are thethings that we think or need to
have in place to identify churnto then go after solving solving
the problem, you know later, um, and that is no one else, like,
(32:44):
no one else can have theseconversations like, like, really
, truly no one is as close tothe systems as you are in
marketing operations and revenueoperations in general, using
that term broadly.
All you ops folks out there,yeah, it is a really, really
exciting time for you to go makean impact and I love that.
(33:05):
You just said you had one personeffectively making the impact
of 10, right, and I think rightnow is the time to you know
whatever, whatever phrase youwant to use, grab the bull by
the horns or seize the day orwhatever.
Absolutely sink your teeth into the art of the possible and
(33:29):
I'm going to say again startthinking like a product manager
and you know you're going tohear it over and over again from
our community and for me, andyou're all going to probably get
annoyed with me at some pointbut you are the product manager
of the go to market tech stackand your job is to figure out
(33:51):
how to best leverage all ofthese pieces of information to
enable your internal teams to goto market, to enable your
buyers who are buying yourproducts and services to engage
with your brand and yourinternal teams in a meaningful,
effective and scalable way, andyou also, unfortunately for you
(34:12):
have to be aware of is the brandbeing represented properly and
are we legally compliant?
Those are a lot of fun thingsthat you have to manage, but
they're all stakeholders thatyou have visibility across and
it sounds daunting, but forthose of you that are in this
role and you're having fun withit it's actually really exciting
, yeah.
Michael Hartmann (34:33):
I would add a third one.
Mike Rizzo (34:34):
That's how I translated for Kobe yeah, go
ahead.
Michael Hartmann (34:36):
Yeah, I would just add a third one to your
list there, which is being anadvocate for the customer,
because I think sometimes thatgets lost.
I mean, I don't know about you.
Mike Rizzo (34:46):
Sorry if I didn't say that clearly.
Yeah, you're absolutely rightIf I didn't say it clearly
enough when I said hey, you haveto service the internal team
members as well as the customersthat are buying your products
and services.
Yeah, that's the part where youneed to hone in on that
customer experience.
How are they able to actuallyengage with us in an effective
way?
(35:06):
That's also brand safe andcompliant right.
Brand safe and compliant Right.
Kobi Stok (35:09):
I would even add two more things in general that I
think that are must-haves forevery ops person out there, and
especially these days.
I think the number one is thinklike a builder.
Think like a builder.
That's what you need to do.
That's the mindset that youneed to be in in order to be
(35:31):
very successful.
Right, you need to build stuff.
You don't need to build anotherlike, build like.
You need to orchestrate thosesystems together.
So, yeah, right.
And second, it's really workout, because I think that, again
, doing the impact of 10 people,if you are successful in
(35:57):
achieving that impact, I thinkthat's the whole grand for
everyone at Ops.
And I think that Ops gets morelike a center stage not
necessarily the singer of theband, but maybe the bass player
or the lead guitarist stillreferring to our earliest
conversation Yep, but that's ascenter as they would, as they
(36:20):
ever end and currently and Ithink it will be even even
further down the road, becausethis is the platform that the
business runs on yeah, Withoutit, it won't run.
It won't run without it.
Michael Hartmann (36:37):
Yeah Well, I think it's interesting that
you're seeing a 10x productivitygain using some of these tools,
specifically around dataanalytics and science reporting,
whatever you want to call it,because I've believed for a long
time and this is why I've beenreally hopeful that AI would
play a part in making this Idon't even want to say easier,
(36:57):
but more scalable is because Ibelieve for a long time that it
is an effort-based thing to doright.
It takes time, effort and someexpertise to be able to do
effective data analysis andreporting and to get insights
that can then be used to impacta business and so go ahead.
(37:18):
No, no, I think that's what I'mreally excited about is hearing
that this is maybe somethingthat's actually happening.
Mike Rizzo (37:25):
I think it is, and it's funny here a word that we
very rarely use in our uh, sortof day-to-day world here I
suppose I don't think I hear itall that often is creativity.
And and this is an interestingtime because, while we don't
(37:45):
fall into that creative marketercategory, the one that we all
think of traditionally, right,right All, the fun ads and the
design and the and the coolcampaign ideas Although I will
say there's a lot of marketinghouse people that come up with
great ideas like campaign wise.
Okay, um, this type ofcreativity is it's a jigsaw
(38:08):
puzzle that you are, you arestitching together, you're
actually making the jigsawpuzzle that you are, you are
stitching together, you'reactually making the jigsaw
puzzle while you complete it,like all at the same time.
Yeah, and it's, it'sfascinating, right, it's like
you're seeing possible, you'reasking questions that could lead
to down paths, that that couldanswer questions that the
(38:30):
business never even knew it had.
Right, and I just I think it'sa new type of creativity that,
um, in some ways, is veryfreeing it keeps coming.
Michael Hartmann (38:41):
I can't go back to one of our earliest
guests, brandy sanders, talkingabout how, like thinking about
this like chess, right?
So I didn't even go beyond apuzzle, right, because it's it's
a puzzle that has made probablymultiple dozens of potential
solutions and I think that's,yes, how I think about it.
Um, but here's my, here's myconcern.
(39:02):
It could be maybe you can chimein here too, because I've said
this many times here, but I lovethe idea that the technology
could make it so that someonewho just doesn't have the
bandwidth to be able to providethose insights, but has access
to the data and understands itand the flaws in it and
everything else is that I amconcerned that there's not
enough knowledge or expertise onhow to actually interpret the
(39:25):
analytics and statistics, orwhatever you want to call it,
across most organizations andeven in inside ops.
So are you seeing some of thesame stuff with your clients and
customers or in your ownexperience, and do you have any
thoughts on how to address that?
Kobi Stok (39:42):
yeah, that's that.
That's um.
That's a very interesting uh umthing that you said just,
Michael.
I think that talented peopleacross any profession are really
hard to find.
That's in general.
I think that analysts and datascientists are really unique
(40:08):
because they need to understand.
They need to have technicalchops right, but in order to
really find insight, they needto understand the business.
The combination of one personbeing super technical and
understanding the business isunique.
What were my theory right?
(40:31):
And we are building the productas we built an AI agent inside
of the product.
That's an analyst.
We built an AI agent inside ofthe product.
That's a scientist, Meaning wekind of lead the ops person to a
journey where we ask them thequestions.
(40:55):
Right, there's a think about avery precise process that they
go through and then the AIexecute the job of an analyst
and scientist, because theproblem is that in most cases,
(41:18):
you don't have enough of them ata specific point.
So every question that you asktoday let's say we don't have AI
let's go back three years orfour years Every question that
you ask on the go-to-marketfront, you would need someone
else to help you around.
You couldn't answer questionson your own.
Michael Hartmann (41:48):
Yeah.
Kobi Stok (41:50):
I mean, it's so complex, right, in most cases
you need to export data, thenyou need to have a Python
notebook and then blah, blah,blah.
It's really complex.
Yeah, right, but today the sameprocess is happening, but
someone else is driving thewheel and it's not a person.
(42:10):
So what we're seeing is that weare freeing, we are freeing the
organization from thoseheadcounts that are expensive
and can't operate in the twofronts of the data and the
business, and we're enabling thebusiness to run it.
(42:31):
Now a little kind ofclarification here For the very
strategic initiatives, ai is notstill there.
But for the tacticalinitiatives, let's understand
why customers churn, let'sunderstand apps Like all those.
Ai can definitely do this, andwe do day in, day out.
(42:53):
So I hope that I answered thequestion, michael.
Michael Hartmann (43:01):
I think so.
I'm sure I think we couldprobably carry on for quite a
bit longer, but I think we'regoing to have to wrap it up here
.
Kavi, if there's, if folks wantto keep up with you, connect
with you or learn more aboutwhat you're doing at Forward AI,
what's the best way for them todo that?
Kobi Stok (43:18):
Oh, so maybe before I answer that, I want to add one
sentence on the data hygienepiece that Mike mentioned.
So while we're implementingforward, we automate it in
reprised.
So there are many problems thatwe see along the way.
What Mike said about dataquality or data hygiene, that's
(43:47):
the number one problem that wesee as well in every case out
there, Every case out there, andI'm excited to that's the first
time that I talk about it butwe are launching an agent that
puts data hygiene on autopilot.
Mike Rizzo (44:06):
Okay, I like it All right.
Kobi Stok (44:09):
Okay, I like it.
All right.
So every data hygiene problem,you can use the agent to solve
it.
I'm not sure it can solveeverything right, but we mapped
200 problems that we have heardfrom our customers and prospects
(44:29):
and we solved all of them.
Not saying that it will be ableto do everything, but that's
just that.
That's a new product that we'relaunching soon.
We are super excited.
We have better users and theresults are amazing.
(44:51):
If you want to um kind of um, uh, I'm learning about forward, so
it's forward or the I, withoutthe a, um and um.
You're welcome to follow us andmyself on linkedin where we
talk about ai and the differencebetween an LLM to machine
learning and I think thateducating the people around what
(45:13):
is AI because AI is big, it'sfrom the 50s, it's like you know
, it's like guitar.
Innovations are happeningacross the board and people
don't understand what AI is,what LLM is, what OpenAI can do,
what they can't do, what thedifference between an OpenAI to
an entropic, I don't know.
Deepseq, all of this peopleneed to.
(45:35):
And again, ops I encourage Opspeople to really be curious and
learn more about it and I inviteyou to follow us myself on
LinkedIn, you know, on all thosechannels.
Mike Rizzo (45:49):
Fantastic.
Yeah, I'm really excited aboutwhat you're doing at Forward,
kobe, and I'm glad we were ableto reconnect this year and have
you on the podcast, and we'llprobably, I know we have a few
more things lined up to in termsof educating the audience.
So, for those of you listening,stay tuned to what Forward's
doing.
It's F-O-R-W-R-D, that's howyou spell it.
(46:13):
Uh, you can go visit thewebsite at forwardai and um and
then pay attention to some ofthe other posts, emails and
slacks and social things thatwe're doing on the marketing
upside, because you're going tohear a few more different
renditions and going deeper indifferent areas around AI and
data that forward's sort ofhelping to solve.
And then Kobe I got one ofthese days when we get a chance
(46:35):
to sit down for a little longer.
I got to show you what I wasdoing with audit hub as a
product that I built.
It was totally around this datahygiene thing and I feel like
you probably already solved it,but I'm so excited by what
you're what you're telling us.
Kobi Stok (46:49):
I am too, we.
Michael Hartmann (46:51):
but I'm so excited by what you're telling
us, I am too.
Kobi Stok (46:53):
We will meet and you will show me Promise that Sounds
good, kavi, thank you.
Michael Hartmann (46:56):
Thanks for staying up late for us and, as
always, mike, thank you forjoining Absolutely.
And thanks to our audience forcontinuing to support us.
And if you have suggestions fortopics or guests or want to be
a guest, feel free to reach outto naomi, mike or me.
We'd be happy to chat with youabout it until next time.
Kobi Stok (47:12):
Bye, everybody bye, bye guys.
Popular Podcasts
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
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. Follow now to get the latest episodes of Dateline NBC completely free, or subscribe to Dateline Premium for ad-free listening and exclusive bonus content: DatelinePremium.com