Latest / Agentic Conversations (formally mlops.community) / Leading Enterprise Data Teams // Sol Rashidi // #227

Transcript

Leading Enterprise Data Teams // Sol Rashidi // #227 · Agentic Conversations (formally mlops.community)

0:00
Hey everyone, my name is Sol Rashidi. I'm
0:03
a former slash recovering C -suite executive who's Not
0:09
in transition, but I've definitely Started my own
0:12
company and doing my own thing speaking advising
0:14
writing I got a book out which has done well super
0:17
happy about that I am so excited to be here and
0:21
Demetrios asked me how I like my coffee all black
0:24
no BS no fillers No, that's just straight black
0:28
get to the point and give me that caffeine rush so I
0:30
can keep doing good shit What is happening folks
0:34
welcome back good people to the ML Ops community
0:37
podcast if you don't know Sol I highly encourage you
0:41
to check out her newsletter and or By her book your
0:46
survival guide to AI because it is an absolute
0:51
gem We'll leave a link to it in the description. She
0:53
has so much experience leading data teams I mean
0:58
one thing that she said was that the size of the
1:01
teams have been anywhere from 120 people to 834
1:06
which is absolutely insane to me at 834 people so
1:10
she talked about what she does to Make sure that the
1:15
decisions and the initiatives that they are
1:19
taking When she goes into a company are the right
1:22
ones. She also Mentioned how to stay
1:25
out of perpetual POC purgatory I think If she
1:31
did not coin that term she should because it is
1:33
incredible. I love that term perpetual
1:36
POC purgatory Hopefully you have not been on a team
1:40
that has been stuck in that purgatory I gave her
1:44
a bit of a multiple choice Answer or question that she
1:49
could answer it was all around What's the most
1:51
important thing for you when it comes to? Building
1:55
these teams and going out there and gaining influence
1:58
in the workplace and She said without a doubt
2:04
relationships are Paramount it's relationships building
2:09
those relationships cultivating those
2:10
relationships above all else So this one's gonna stay
2:15
with me for a long time I know it already if you
2:18
like and enjoy it leave us a review drop in some
2:21
stars on Spotify and Let's get into this conversation
2:24
with soul by her book if you do not have it already
2:34
Talk about the wrong use cases for AI what have you
2:38
been seeing out there? You know, I think everyone in
2:40
general and whether you're running use cases for a
2:43
data product data ecosystem AI Everyone has this notion
2:47
of like let's pick use cases based off of business
2:49
value I don't know how other people have done
2:52
their job But like my job as sort of the CDO CAO CD
2:56
AI or whatever CXO title you want to be has always
3:00
been to support supply chain Manufacturing the
3:03
brands the label so like the business that runs a P
3:06
&L and also the functions and they all are like
3:09
fatigued or plagued with a problem that they want
3:12
fixed and I kind of get stuck in this position
3:14
where my team is yay so big my capex is yay so
3:18
large I can only put so many individuals on a
3:22
certain project and Quite frankly if they brought it
3:25
to my attention all of it has business value And so
3:28
for me it's been really hard to pick use cases
3:30
based on business value and I stopped doing that
3:32
because I actually Unconsequentially made
3:34
enemies along the way where I had to prioritize things
3:37
So I would pick one business's problem over
3:39
another business or I pick one fiction functions
3:41
problem over another and People just don't like this
3:45
shit being to prioritize part of my French So I
3:50
was like business value is not the way to go about
3:52
this because what's business value to manufacturing is
3:55
different than what's business value to supply
3:57
chain which is different than procurement HR brand X
4:00
brand Y region Y Market B and so I have this formula
4:06
that I created and It's complexity versus criticality
4:10
and I created this at IBM when I was like Yay here
4:15
to all the work that I had to do for everyone and
4:17
it's applied in every one of my C -suite position And
4:19
it helps me just sift through the clutter
4:22
prioritize in a way that's Both qualitative and
4:25
quantitative and then there is a method to the madness
4:28
so if people disagree It's not a problem, but they
4:30
can fundamentally understand why we came to the
4:33
conclusion that we came to and so the criticality of
4:36
it is how Critical is it for the business? I
4:39
understand it's a pain point for you and I
4:40
understand that's something you've been plagued with
4:42
but Is there a fine or regulation that's been set
4:45
our way and are we in trouble? Is
4:48
there a massive competitive threat? Do
4:50
we have excess inventory for example? That's
4:52
about to go stale or the shelf life's about to
4:54
expire Is there a new product launch where we've
4:57
invested a ton and we're not able to see the
5:00
consumer acquisition numbers it Like what is it that
5:03
makes it so critical and then you know You
5:06
fundamentally grade it based on these questions and it
5:09
naturally gets its own score But then there's the
5:11
complexity side of it the complexity to deploy Do we
5:15
have the basic infrastructure to even be
5:17
able to move something? Do we have the data and is
5:19
it accessible or am I gonna have to go through
5:21
this? massive data ecosystem around
5:24
cleansing and hygiene and accessibility and quality
5:27
and etc. Or is it ready to go? Do
5:30
the people I need involved in the project are they
5:32
even available? Because you find that in a lot of
5:35
organizations They're really smart ones are on multiple
5:37
projects and so they tend to be spread really really
5:40
thin Are those folks even around and so there's a
5:43
series of five to seven questions that I ask on
5:45
complexity So everything gets a weight and
5:48
everything gets a score and I have this quadrant where
5:51
on the x -axis I mark you against how critical is it
5:55
to the organization and on the complexity score How
5:57
complex is it to deploy if it's highly complex and low
6:02
criticality? It's a non -starter. I'm
6:05
not doing it get it out of there if it is low
6:08
complexity and highly critical We're prioritizing
6:11
you because I know I can actually execute on it and
6:14
I can do so with the resources I have so I do
6:17
this every quarter and I plot everything because you
6:20
know things always come up so even if you set your
6:23
Strategic initiatives for the year or you've got
6:25
your projects that you want to run personally
6:27
professionally for your startup for your BC You do
6:31
your yearly planning based on that method and then I
6:33
update it ever quarterly and I always Communicate
6:36
what's being prioritized so that people know what's in
6:39
the bucket and what's not in the bucket And if
6:41
things need to change then they got to go up higher
6:43
in the food chain for me to prioritize stuff But it
6:46
neutralizes the playing field so back to your
6:48
question or what are the wrong use cases? I
6:51
think a lot of people choose business value.
6:54
That's great But if you only use
6:56
business value as your marker, so let's say
6:58
criticality, but not complexity That's where a
7:01
lot of companies get into trouble and they get stuck
7:03
into perpetual POC purgatory You have a lot of AI use
7:06
cases that are just in PSE mode and they're wondering
7:09
why they even have a hard time shifting things Into
7:13
production and the reason they have a hard time is
7:15
because they don't even have the basic
7:17
infrastructure They can't even calculate the cost of
7:20
workloads to be able to push something in POC Could
7:22
you imagine pushing it into production or they've got
7:25
your classic devops? Mechanisms
7:27
in place, but they don't have ML ops or devops for
7:30
AI Which is totally different because when you
7:33
push when you do software development You push a
7:35
piece of code into production. It
7:37
is static once that code is containerized and it
7:40
lives in the ether It's alive and well That's not
7:43
how it works with AI. It is a live wire It's a
7:46
live stream. You're only as good as your last
7:48
piece of information so Building data pipelines and
7:52
intricacies associated with that the orchestration of
7:55
data being fed into the arms and then managing the
7:58
security protocol It's just a different beast and
8:00
people aren't prepared for that So I always say if
8:03
you have the maturity great If not, you have to take
8:06
complexity into consideration to see what's realistic or
8:10
not to get out of that POC perpetual Or perpetual
8:13
POC purgatory Remus folks are great term by the way,
8:16
that is the three P's The three P's get out of the
8:20
three P's pick a use case. You
8:22
can actually deploy I was recently on a client site
8:26
because I'm advising now and One of the major
8:29
management consulting firms had done a strategy. They
8:31
were like 12 use cases and they're like, we don't know
8:33
where to start I was like well I'm gonna ask you
8:37
series of questions and after I did I was like
8:38
nine of these 12 you can't even do I'm like you you
8:41
won't even be able to soft launch this stuff Because
8:45
you're missing this and you're missing this and
8:46
you're missing this I'm like so I would even
8:48
consider it So the answer is these three it's kind
8:50
of easy and like they were pissed off that they paid
8:53
nearly a million dollars for this extra and cheap
8:57
For use cases, they wouldn't even be able to
8:59
get a launch All right, we're analyzing a quarter
9:04
later We thought this was gonna be done in a quarter
9:06
or we thought this was gonna be done two quarters
9:08
ago Yeah, and it still is happening It's really hard
9:13
for us to say stop now because it feels like it's
9:16
just around the corner but you know that like maybe
9:21
in in your heart of hearts it Potentially isn't going
9:24
to be finished this quarter either. I
9:26
Think there's two key decisions you have to make
9:30
call the shelf life or Reassess and do better
9:33
forecasting the shelf life. I
9:35
learned oddly enough like through an experience I had
9:39
playing college sports, you know I was a great high
9:42
school athlete and I went played water polo for a
9:45
division one school And I was like second string to
9:48
second string. I was still training six
9:50
hours a day And I was working my ass off And I
9:54
talked to my coach mo and I'm like am I ever gonna
9:56
get in the water? I played water polo It's just
9:58
like listen. I love your tenacity. I
10:00
love your effort, but You know the girls are five
10:04
ten you're like five three on a good day And like
10:06
you can make up a certain portion of it with power
10:09
But you're just not naturally built to be
10:10
professional at the sport So you gotta make a choice
10:13
you want to keep swimming six hours a day and train
10:15
with the team or do you want to move on other
10:17
things? So It was like I'm not a
10:20
quitter, but this is point of diminishing returns I
10:24
was like, maybe I'll just refocus on academics. So
10:27
I'll just be you know a professional student And
10:29
then randomly a week later I got recruited by the
10:32
rugby team She she tapped me on the shoulder Kathy
10:34
and Kathy's like you ever played a sport. I
10:36
was like, yeah until a week ago I was playing
10:37
water polo. She's like have you ever tried
10:39
rugby? I was like what on earth is
10:40
rugby? She's like well, it's kind of
10:43
like football without pads and I'm like, whoa, whoa,
10:44
whoa Why would women like this? We
10:46
just are built for this. She's
10:48
like, I just have a feeling you're gonna be a
10:50
natural at it. Well, I Had great ball handling
10:53
skills because what you do treading in water. I
10:55
now had land What was a deficiency in water polo?
10:58
Which was I was short was a
11:00
strength in rugby because I was 5 -3 lower center
11:02
gravity harder tackle by calling my shelf life in
11:06
Water polo I found my calling and rugby and I
11:10
was a natural at it and I became captain in three
11:11
months and that taught me that Sometimes you just
11:14
have to reassess something and it's not that you're
11:17
giving up and it's not that you're failing it just
11:20
for different parameters variables Constitutions it
11:25
wasn't meant to be so I think we've all gone
11:28
through projects where we are spending endless nights
11:31
endless weekends and that Goal marker just keeps
11:35
moving and you have to truly assess and go are we
11:38
ready to do this or is the team putting effort
11:40
into a bunch of things that We just got to call
11:44
it shelf life. So I think that's one if
11:47
There's hope and if there's faith then I think the
11:51
alternative option is Redo the forecasting and be
11:54
realistic about what it's going to take We assumed
11:57
these folks were going to be available They are not
12:00
available and when they are it takes us six weeks to
12:03
get on their calendar We assume we had this data we
12:06
have it It's not accessible and the only people who
12:09
can grant us access are the application leads and
12:11
it's not two application leads There's eight
12:13
application leads because the data happens to be
12:15
fragmented Not only do we have to get permission and
12:17
approvals, but we now have to aggregate it that's
12:20
going to take an extra three months and Just being
12:23
very honest about what the new forecast is going to
12:26
be because the last thing you want to do is keep
12:28
pushing the timeline Because you lose credibility and
12:31
when you talk about business value not being
12:34
the right metric I Instinctively see that just
12:38
because it is fuzzy and it feels like Like you said
12:42
it's like okay business value means one thing to
12:45
one person It may mean something totally different
12:48
to another person and the numbers that they are
12:51
thinking about could be completely On two sides of
12:55
the universe So they're very far apart and the
12:59
numbers are sometimes made up Forgive me for saying
13:01
that that could be kind of bold, but the numbers are
13:04
sometimes made up You sometimes are being thrown
13:08
numbers at you and it's like well if we put this
13:11
into production We're going to be making or saving X
13:15
millions which obviously is much more important than
13:20
anything else we're doing right now So this should
13:22
be highly critical, right? So
13:25
I guess the question is how can you stress test
13:29
those numbers? I'm a big believer of the
13:33
Reagan quote trust but verify I get numbers thrown
13:36
at me all the time and you know Numbers are very
13:42
manipulative and so for me I understand why we need
13:46
them just because it proves a point and proves a case
13:48
and it makes it really convincing Unfortunately
13:51
because I'm a data person. I
13:52
don't take anything in face value. So
13:54
someone's giving me Statistics if we do this
13:57
we're gonna gain X amount of market share if we do
13:59
this we're gonna increase our efficiency by X amount
14:02
I love the percentage points. I'm
14:04
like, okay great walk me through the method in which
14:05
we calculated this because I want to make sure I
14:07
Understand the context the assumptions the players of
14:11
all and that I could once I understand your
14:13
measurement I could track towards that measurement. It
14:16
has to do with understanding How those
14:19
numbers were created so that when we get to a
14:22
point where we need to measure and monitor and
14:25
report on our progress We're doing so according to
14:28
based on the original metrics and baseline And I
14:32
also do it to validate and make sure that people
14:34
aren't making shit up which is wise very wise I'm sure
14:37
that you have a reason for that because you've been
14:40
through enough cycles where people were making shit up
14:42
And so you saw they saw what happened That's right
14:46
And it became a number that stuck and no one
14:48
asked the question of where did this come from? How
14:51
was it calculated? What were the assumptions that
14:53
went in and what do we need to be conscientious
14:56
of? That is a hundred percent true
14:59
and I hate to say it but sometimes I was the analyst
15:01
doing the calculation I'm like making a bunch of
15:04
stuff up But I hope it sticks this and then
15:06
sometimes it was the person who had to execute because
15:08
something got bought in from above and I was like,
15:11
but this doesn't make any sense. The
15:12
most we can get is 13 percent. Why
15:14
are they promising 37 percent? I
15:17
am we keep going and we keep going and we kill
15:18
ourselves because some executive said this number
15:21
was achievable We don't want to disappoint them So
15:24
fortunately with my career because things have kind of
15:26
progressed and I have a seat at the table now I'm
15:28
like, okay, wonderful. We absolutely want to support
15:30
this first thing. I need to know is how is
15:32
this number generated? Talk to me about Inheriting
15:36
teams and how you've gone about that especially data
15:40
teams building teams is easier in The sense that
15:46
you understand what you need to do The talent that
15:50
you need to do it You're often restrained with capex
15:53
because it's not like you get an endless bucket of
15:55
funds to hire and build but you build really good
15:57
ones You hire really good teammates and if they've
16:01
got that drive that grit that to nasty like a lot
16:03
a lot can happen But you're starting from scratch
16:07
and the recruiting process takes a long time So it's
16:09
you yourself and I for really a long time until
16:12
you can do the right hires and you get some support
16:14
So progress up front is really slow and then you
16:17
hit critical mass I've inherited teams as small as
16:22
170 and as large as 832 and I'm very specific Wow,
16:28
I knew the number and Oftentimes I get brought in
16:32
not as the first transformational leader Well,
16:35
I did with Royal Caribbean Sort of with Sony but like
16:38
in the other positions They've had a few leaders
16:40
prior to and they've had mixed results or it wasn't
16:43
the transformation they wanted And so they'll bring
16:45
me in like okay, you know how to do this You've done
16:47
this four times over at least in the data and
16:50
analytics in a space Change It's really interesting It's
16:57
a mix back of nuts and you inevitably get the
17:01
doers the evangelists the cheerleaders the naysayers
17:06
the Chromogens like you get these archetypes of
17:10
personalities and you could pick it up in the first
17:12
week And it's amazing how so many people don't want
17:15
to impress the new boss They do not give a damn
17:19
and that's just not the camp I grew up in and
17:22
there's others who are so like We're so excited about
17:26
this. We've been wanting this. We've
17:27
been praying for this Let me walk you through every
17:30
issue we have So I go through a few things one
17:34
assuming it's not 800 Let's say it's a couple hundred
17:38
I spend the first four to five months getting to know
17:41
everyone and I will literally kill myself doing
17:43
one -on -ones Well first with my leadership team
17:45
then LT plus one LT plus two and then I'll get to
17:48
the middle managers and then the practitioners And
17:51
I ask some basic questions. What
17:53
should we start doing as a team that we haven't done?
17:55
What should we stop doing that
17:57
we continue to do and what should we continue to do?
18:00
That's working really well, and
18:01
we're building good reputation and we have
18:03
credibility about it And I will ask everyone that same
18:06
question And then I'll understand is there clarity
18:09
on what my job is? What do you think my job is?
18:12
Is there clarity on what our
18:13
team does? What do you think our team is
18:14
supposed to be doing? And then I'll ask questions of
18:17
who should we be partnering with who've been our allies
18:19
who've been our advocates Who is not our ally and
18:22
who's not our advocate and what are the reasons why?
18:25
I'll ask those questions and
18:26
then all I do is listen take notes listen take
18:28
notes in About four to five months I will come up
18:32
with a map of the start stop continues partnerships
18:35
allies advocates the ones not so much and I will
18:39
create a sort of like a mind map of where we are
18:43
and Then what I'll do is it throughout that time in
18:46
parallel I'm having conversations with key
18:48
stakeholders executives middle managers and I will
18:52
ask him who from my team is helping you Who has
18:55
created the greatest impact? Who
18:57
do you consider a teammate of yours? And
18:59
if I've got people consistently saying no one
19:01
or if I've got people saying the same five people
19:04
Then I'll start understanding. Okay,
19:06
where the issues are From there like I will actually
19:10
create what I call a will skill map And it's like
19:13
those that have the skill sets and are willing to do
19:16
the work. They're hungry. They're
19:17
motivated They're excited me coming on board is a
19:20
breath of fresh air for them And they've been
19:23
wanting to do something right so there's purpose
19:25
and impact and they're just hungry They haven't been
19:27
they just haven't had the right leadership Then
19:29
there's those that have the skill but don't have the
19:31
will they're complacent. They're
19:33
tired They've been with the company 20 plus years.
19:36
They're not worried about change
19:37
because they're not going to change They've been
19:38
there too long, right? Then there's those that have the
19:42
will but don't have the skill you have to assess
19:44
and go are they coachable is This potential can I
19:47
put them in this role and will they flourish if I
19:50
put enough time and investments? And
19:52
then there are those that don't have the skill and
19:53
don't have the will Honestly, the ones that
19:56
don't have the will for me are automatically into a
19:59
bucket of assessing Do I need to make an
20:01
organizational change? because
20:02
I can't create change without making change and
20:06
Everything starts in the mindset the mentality and
20:08
the team that I have to work with um, so inevitably
20:12
unfortunately when I inherit a team Some people are
20:15
naturally just a little bit apprehensive and scared like
20:17
what changes is she gonna make but I take a really
20:19
long time I know about six to seven months of what
20:23
changes I want to take but I usually take up to about
20:26
a Year to 14 months before I thoroughly make the
20:29
changes because there's a lot that goes into it It's
20:32
it's not an easy task But for me, it's all about how
20:36
hungry are you and do you seriously want to be here
20:38
for the right reasons? And if the answer is no
20:41
There's just no room on my team. So
20:43
the first option I give is transfer to another one
20:46
Let's find you another job team or worst worst comes
20:50
to worse We may have to discuss a package because
20:53
they clearly don't want to be here All right Let's
20:55
take a minute to thank our sponsors of this episode
20:58
weights and biases Elevate your machine learning skills
21:02
with weights and biases free courses including the
21:05
latest additions enterprise model management and LLM
21:10
engineering Structured outputs choose your own
21:14
adventure with the wide offering of free courses
21:17
ranging from traditional MLops LLMs and CI CD to
21:22
data management and cutting edge tools That's not all
21:26
you get to learn from industry experts like Jason
21:29
Louis Jonathan Frank Shreya Shankar and More all those
21:35
people. I will 100 % vouch for they
21:38
are incredible friends of the pod Enroll now to
21:41
build better models better and faster and better and
21:45
faster and just get your Education game on check the
21:49
link in the description to start your journey now
21:51
Let's get back into the show. How
21:53
do you assess if someone has the will but not the
21:56
skill if there's folks that I have Learned to respect
22:01
I Will assign them to the leaders or middle management
22:05
or non leaders and saying Assess their skill and see
22:09
what they if they can benefit you they can
22:10
benefit the team They can benefit the group or
22:13
sometimes if there's a rising star, but I'm not
22:15
quite sure I'll assign them as my chief of staff
22:18
Usually I'll have two or three because we will
22:20
always get random projects that come to me directly
22:22
They don't go through the normal funnel or filter and
22:25
I'm just gonna need help And I'll say listen, we've
22:27
got a strategy deck that's coming up. We've
22:29
got a percent with XYZ. Here's
22:31
the narrative they want to communicate See if you can
22:34
create a flow or there's an analysis that needs to
22:37
be done Marketing wants to do a customer acquisition
22:41
strategy They're gonna do an AB campaign testing
22:44
between this consumer cohort in this consumer cohort I
22:47
want to build a methodology for us to track and
22:49
monitor Put your thoughts on what we should do and
22:52
how we should communicate So oftentimes like I'll
22:55
know the answers to these but I will assign it to
22:57
this individual to see Critical thinking common
23:00
sense. Are they leveraging the tools
23:01
available to them? Do they know can they ask
23:04
questions from the people who are known for this
23:06
space just to see how resourceful they are So for
23:09
those that have the will but don't have the skill
23:11
I'll just assign them to different individuals or
23:13
they'll be my chief of staff So we can assess
23:15
their skill set level and where they're best suited
23:17
some people aren't strong engineers or have really
23:21
strong analytical mindset, but They can think through
23:25
how a storyline should connect and we all struggle
23:28
with storyline Or can they should think through how
23:31
something should be presented something we all
23:33
struggle with but if they're also a brilliant
23:35
engineer Then I'm gonna put him next to a brilliant
23:38
developer and group him up within the scrum team of a
23:41
product manager that I trust wholeheartedly do you
23:43
think that Holding people accountable on the data
23:47
team is any different than holding people accountable
23:50
on any other team Because there it feels like there
23:53
is the excuse of Data is messy. We
23:57
need exploration. It's a science sometimes So not
24:01
all hypotheses are gonna come out. I
24:04
think the accountability is worse Meaning there's more
24:08
pressure There's a lot of things that you can hide
24:12
with business processes You can't hide with data If
24:16
something isn't of good quality or of good hygiene
24:20
and let's say in most of my organizations I own the
24:22
enterprise data management team and some things just
24:25
not right like why isn't it right? That's
24:27
literally our job So then we'll have to go through
24:31
until the onion. I'm like what happened Is it
24:33
someone on our team that's not doing their job? Is
24:35
there a business process that hasn't been redefined?
24:38
Is there a tool that hasn't
24:39
been configured to make sure that we get this
24:42
format consistently and through our outputs? Where
24:46
is it broken because we're the end EDM team? So
24:49
if there's bad hygiene, I mean I'm blaming on us
24:51
Unless it's not in our scope. So
24:53
like in my last employer, we had product. We
24:57
had vendor. We had location But we didn't
25:00
have customer and we constantly were getting bad
25:02
customer data requests of like can you help fix
25:05
this? I'm like, it's not in our
25:06
scope. So in the following fiscal
25:08
year I said, you know, here's what is in our
25:10
scope and here's what's not and here's how many
25:13
requests I got in last year for these following
25:15
domains Should we change the scope? And
25:18
if so, I'm gonna need funding and resources and
25:22
here's what I can fix for you And here's what I'll
25:23
be able to do for the team If not, then we have
25:26
to make it very very clear that we've made an active
25:29
decision To not incorporate this into the enterprise
25:32
data management team for the following reasons and
25:34
it's what I'm gonna communicate on But if right
25:36
now it's sort of headless and it doesn't have an
25:38
owner I'm willing to take it on but I need the
25:41
resources to take it on and I will find a way
25:44
Integrating it into our operating model. And
25:46
so I'll actually make the case Because that often
25:49
happens and I'm not sure why different master data
25:52
domains get split out to different teams It doesn't
25:54
make any sense. Um, but sometimes it just never
25:56
makes it in but we'll always get the requests And
26:00
at some point people are like data's broken data's
26:01
broken like which data Yeah, it was not all
26:05
data's created equal if it's data that my team
26:07
owns, let me know specifically what's broken I
26:09
will have it fixed for you. Oh,
26:11
but it's this data. We don't own this Will you
26:13
come with me to the board and say that we now need
26:16
to own this so I can fix this for you and never
26:17
becomes an issue again And then that's where you start
26:20
building partnerships because you have a leader who's
26:22
being an active advocate And stepping up to the
26:24
table and saying that I'm a janitor. I'm
26:27
just here to clean shit I'm here to clean shit and
26:30
build really cool stuff from it. Um,
26:32
and that's how you build your credibility like Some
26:35
of my deepest relationships have nothing to do with
26:37
folks in the data analytics space or the AI space It's
26:40
with business partners that I've served in the past
26:42
and they're like, you're the only one who's willing
26:43
to get their hands dirty I'm like down straight.
26:45
That's our jobs. It's
26:46
a dirty dirty job Talk to me more about the other
26:50
stakeholders that you have to deal with and how Being
26:54
a data professional and a data leader plays into that
26:59
This is probably the part that I've struggled with
27:02
the most We now have a world where we have a CIO
27:06
a CTO CIO a CDO chief digital officer another CDO
27:12
chief data officer Everyone has their VPs. Everyone
27:15
has a line of commands like oh you and even with
27:18
the chief technology officer like I know databases But
27:21
I'm not a database administrator. So
27:23
just because you have a CTO title doesn't
27:25
necessarily mean, you know, the modern data stack It's
27:29
a lot to know all of us are trying to be masters
27:31
of our craft, but it's impossible to keep up with
27:33
But the challenge that I've always had and it's never
27:36
gone a ways Where does one rule start and another one
27:39
end? So for example when I'm
27:41
developing data products like I don't want to build
27:43
dashboards anymore It's too time -consuming. There's
27:45
always tiny little changes I'm gonna build data
27:47
products that are multiple use cases multiple user
27:50
groups and persona and I call it the one -to -many
27:52
build I build once and it serves many But I
27:55
fundamentally have a reliance on infrastructure
27:58
on dev ops on security protocols Well, I don't own
28:02
those I have to partner up with my fellow CIO But
28:06
depending on what's on his or her list of things to
28:09
do Where we fall in line in terms of priorities and
28:12
they're barely able to keep up with everything the
28:14
business is lobbing over the fence My stuff barely
28:18
makes it in the frontline. You
28:19
know, like this just can't work So I'm constantly
28:22
having to ask the question if I can't get the
28:24
resources to own and I will follow all the
28:27
processes and procedures and protocols that the office
28:30
of the CIO have set But I want to control the pace
28:34
of development and the quality of development And
28:38
I'm gonna need funding and resources to be able to do
28:40
so if you choose an alternative path Something
28:43
that should take three months will take us nine
28:45
months because based on the priorities already outlined
28:48
My ass won't even slot in for another three quarters.
28:51
Are we okay? And
28:53
the thing is is I make it all very very clear for
28:57
the stakeholders the business The board the C
29:00
-suite and all teams and whether it's town halls
29:03
Whether it's group meetings whether it's one -on -one I
29:07
very much communicate the state of the union because
29:10
part I over communicate because people only absorb
29:13
about 10 % of what you say And Malcolm Gladwell had
29:16
this great saying if you got to repeat something 27
29:18
times before it actually is heard I need to everyone
29:21
to know of like the limitations oftentimes aren't
29:24
on our ends It's actually we have too many competing
29:26
too many competing priorities And we're just
29:29
waiting in line, but the darn thing is built. I
29:31
just can't push it into production So you ask that
29:35
question of working with everyone It's hard because
29:38
there's just too many priorities within companies
29:40
and you're constantly having to battle and prioritize
29:44
your stuff over others and it's not the right answer,
29:46
but we tend to over commit and Division of labor is
29:50
not clear amongst the CXO Suites and lines and you
29:53
don't know where one thing starts and another thing
29:55
ends like I'm the chief data officer But do I own
29:58
all data platforms and data ecosystems? Do
30:02
I know nope alright, so maybe I own the staff for
30:06
Analytics which now means the data lake do I own
30:09
the data lake or do I just own the architecture
30:11
that conforms the semantic layer like? All
30:14
those questions and it's messy speaking of the Amel
30:20
that you want to stay up to date on all the tech
30:24
Versus all the other stuff that you have to keep in
30:26
mind How much do you feel like it is valuable for
30:30
you to know the newest tool that comes out? Versus
30:34
I'm gonna spend instead of an hour researching the
30:37
newest tools and the cool ways to do things I'm
30:41
gonna spend an hour trying to figure out our strategy
30:44
or trying to network with the other stakeholders So
30:47
that I can get my stuff pushed through and is there
30:50
a hierarchy that you feel like in the needs of
30:55
executive 100 % and it will the priority is always the
31:01
relationships Always people don't do business with you
31:06
because you have the greatest tool They do
31:09
business with you because they like you they trust
31:11
you and you've shown them that you're Interested in
31:15
solving their problems That only comes when you have
31:18
the relationship now. I say that it sounds so
31:20
obvious I got my ass handed to me my first job.
31:23
I Went in guns slinging so
31:25
cocky about the position that I had gotten saying
31:27
like I'm gonna fix everyone's problems I had
31:30
no idea about the politics. I
31:33
had no idea about the relationship building I had
31:36
no idea that the higher up you go It has less to do
31:39
with your IQ and more to do with your EQ No idea
31:43
and I got my ass handed to me, but I learned a
31:47
lesson I've never made that mistake. So
31:49
the first thing I do when I go in is not call
31:51
people's baby ugly It's not say here's how we're gonna
31:54
fix things. It's not come out with a
31:56
mission statement and a vision statement I literally
31:58
spend the first four to five months Getting to know
32:02
as many individuals as possible doing the due
32:04
diligence asking. What do you think my job is?
32:07
What if I had a magic wand and
32:08
I could fix three problems for you? What
32:10
would it be while I'm asking my teammates to
32:12
start stop continues? What you think our role is?
32:18
Etc. It's all about the
32:19
relationships. And so I've always built my
32:21
reputation on that like she actually wants to listen
32:23
She's not coming in and trying to fix things off
32:26
the bat And and that gives me context into the
32:29
business and when I can get to know their language
32:31
And I get to know the business and I get to know
32:33
context Anytime I storytell what we're going to do and
32:37
how it's going to benefit them I use their words and
32:39
I that's that's probably the biggest advantage I've
32:42
ever had I call myself just like a glorious
32:44
translator But I get to know the words and the
32:46
language and the metrics that are important to them
32:48
and I use their language Not our language. I
32:51
don't talk to them about some layers and data
32:53
architecture engineering and orchestration and
32:55
observability new I'm like that thing that takes you
32:59
14 weeks to get here's why in very simple forms Here's
33:04
how we're gonna fix it and here's what you're gonna
33:06
get in four weeks It builds the relationship. So
33:09
I Rely on people a lot smarter than me to do the
33:14
tool discovery Because every six months, there's a new
33:17
tech stack that we can lean on and I've got a
33:19
lot of smart individuals who not only love it They
33:21
do it during the daytime, but they do it at nights
33:23
and during the weekends for fun And I don't ask one
33:26
person I ask a ton of individuals because based on
33:30
what they suggest and why they like it That's person
33:33
number one. I then go to person number
33:34
two I spend the hour to instead of researching
33:36
myself I spend the hour talking to about three or
33:39
four individuals to come up with an opinion and a
33:43
judgment and then I'll come up with a perspective and
33:46
present it and then we'll have a discussion from
33:47
there So I'd rather go straight to the source of
33:50
the people who work with it day in and day out Mm
33:53
-hmm. Let them Be the ones that can
33:57
synthesize that information for you and then you get
34:00
higher level. It's what they love They want
34:02
to feel valued it's sort of their shtick while I'm
34:06
building relationships to make sure whatever we built
34:08
has credibility They get to do with their love and the
34:12
fact that I come to them, you know, sometimes they do
34:14
feel valued about it They're like, okay, she
34:15
works, but she respects my opinion now It doesn't mean
34:18
that what they say I'm gonna take and be like,
34:19
okay, this is our answer. I'll
34:21
do I'll ask many individuals I'll always form
34:23
an opinion on my own, but I'll always say, okay, I
34:25
heard you guys say this There seems to be some
34:28
contradictions here Let's come do a conclusion
34:30
together and like yeah, they tend to respect that
34:35
You know, it's hilarious when you were talking about
34:37
how you don't use the words that were used to
34:41
using and You just say it's almost like an
34:44
abstraction above and you say things like you see,
34:48
you know that thing that takes you a while It
34:51
reminded me of a one of my favorite physical
34:54
therapist. Yeah, who would Give me all of
34:58
the muscles and just call them that guy or those
35:02
guys or this guy There was no like Not of me because
35:11
the muscles are super complicated you have And
35:15
all the names there are to think about and it's so
35:20
easy as the expert to dive into these Names and then
35:25
you because it like makes you feel good like you
35:28
look how much I know it's almost showing off Yeah,
35:31
the end user the person you're trying to help is
35:35
Sitting there and they're like wait, what did they
35:37
just say and they're thinking about what muscle
35:40
it is And so with this it was exactly that they would
35:43
just go like that guy right there You want to
35:45
feel it burning in that guy and there was no
35:49
naming the muscle nothing like that And so it's very
35:52
much it just reminded me of that story where you do
35:54
the same thing 100 % like that we have to and I
35:58
think that's part of the challenge that we as an
36:00
industry have like as an industry I think we have
36:03
three challenges data analytics and let's loop AI
36:05
into it because at the end of the day It's not the
36:08
AI applications that are complicated to be able to
36:09
supporting the ecosystem because it's still
36:11
fundamentally a data ecosystem The pace of
36:15
change is extremely fast We can't all keep up and so
36:18
like our heads are being blown off with just the
36:20
latest advent and tools and startups like we can barely
36:23
keep up Second we have a bunch of technical
36:25
individuals who are learning the business language, but
36:28
haven't invested in the business language So we we
36:30
struggle With storytelling we struggle with the
36:34
narrative like if I go and talk to a CMO I'm not
36:39
talking about Digital campaigns and a B testing
36:43
and data pipelines and consumer cohorts. I'm
36:45
not using that language I'm talking about our consumer
36:48
acquisition cost. I'm talking about our CRM I'm
36:50
talking about our consumer lifetime value and I'm
36:53
using the metrics They presented in the deck that
36:55
I saw as a means of describing what we're going
36:59
to help with and which of those percentage Dingies are
37:01
gonna go up and which of those percentage dingies are
37:04
not gonna go up and why I use their language and we
37:08
just don't do that and I think the third issue with
37:12
our space. It's really really hard to deal
37:14
with as you know my analogy is When's the last
37:17
time you flushed the toilet and called your plumber and
37:19
said thank you When's the last time you flipped a
37:22
switched and called your electrician and said thank
37:24
you you don't you just assume It's gonna work and
37:28
that's like that's our jobs It's very thankless and
37:32
it's not commodity based but You hear news when
37:36
things aren't working when things are working. You're
37:38
not getting a pat on the back You're not getting a
37:40
thank you or a hallelujah You're in the business like
37:45
the janitorial business. We're
37:46
here to fix things and build things and make
37:48
things We we are construction workers and
37:50
janitors No news is good news and everyday news
37:53
means something else is broken We have to fix and
37:56
that's just kind of like our industry Pace of change
37:58
we suck at storytelling. It's
38:01
very hard Because we got to keep up with our own
38:03
space and all the businesses we serve and in
38:06
general It's a very thankless position and so I
38:09
feel like you sometimes got to be a masochist to
38:10
choose a space I love it, but it is so frustrating
38:14
day -in -day out, but I still love it. Do
38:16
you think that? Any company can be turned
38:19
around? When you're a company that was
38:24
born out of disrupting existing companies in a pre
38:28
-established vertical industry It's in your DNA to
38:33
disrupt and so therefore Changing adapting Pivoting
38:39
saying whoop see that didn't work. Let's
38:41
go this way They come very natural at you change is a
38:45
constant For companies and industries who are within
38:48
certain verticals or domains that have been around part
38:51
of the S &P family 10 15 20 50 going on a hundred
38:55
years and while there be disruptions Tiny little
38:59
micro disruptions every five to ten years, but they
39:02
haven't fundamentally had to rethink their entire
39:05
business model And usually they introduce a new
39:07
product to the market or a new service to the market
39:10
Or they have a new brand or a new name and that's
39:12
helped them to survive and do acquisition Those
39:17
companies have a hard time They just do have a harder
39:19
time because their culture isn't inherently built on
39:23
disruption and change And it's so funny you meet you
39:26
assume a laugh, but like in my interview process I
39:29
asked you questions to understand whether it's a
39:31
culture that's fit for change or a culture that's
39:33
not Do you have a pension fund and what's the average
39:37
tenure ship of your employee? Do
39:39
you know why I ask this questions? well,
39:41
is it if there's a pension fund and the average tenure
39:45
is long they are Less likely to change Really?
39:51
Oh, wow your legacy based very
39:53
few companies offer pension funds nowadays. It's
39:55
it's my insurance But those that do means they
39:58
fundamentally have employees That have stayed there
40:01
because their interest is more the pension fund
40:04
pension fund versus Evolving and pivoting and adapting
40:08
and changing and so by default if your answer is
40:11
yes on the pension fund You're gonna get individuals
40:14
the average tenure ship is greater than 20 years as
40:16
the as the majority mean and That's not a crew I
40:20
can change you learn these things, but not because you
40:22
read it in a book You're like, why is this so hard
40:26
like oh Oh Unless it's your book that is the
40:31
perfect place to mention that yeah, if you do want
40:35
to read and learn these things in a book We're
40:37
gonna be we we're gonna leave the link to your
40:40
book in the Description so everyone can grab it and
40:44
also Your newsletter is it is for the leaders out
40:49
there 100 % Your wisdom and learnings are unmatched and
40:55
I really appreciate you coming on here. Thank
40:58
you. I'm super super honored You've
41:00
done amazing with this podcast by the way I found
41:03
out about it just because a bunch of Friends of mine
41:06
who have either worked for me or I truly respect
41:08
because they're amazing at their craft or like this
41:11
is an awesome podcast You got to get to know the
41:13
Demetrios. He's super amazing and they were
41:15
not wrong. So thank you. I'm
41:17
humble. Oh, that is so cool. Yeah
41:19
Hold up wait a minute We got to talk real fast
41:22
because I am so excited about the ML Ops community
41:26
Conference that is happening on June 25th in San
41:29
Francisco. It is our first in -person
41:31
conference ever Honestly, I'm shaking in my boots
41:33
because it's something that I've wanted to do for ages
41:36
We've been doing the online version of this and
41:39
hopefully I've gained enough of your trust for you to
41:42
be able to say that I know when this guy has a
41:45
conference It's going to be quality funny enough. We
41:49
are doing it. The whole theme is about AI
41:53
Quality I teamed up with my buddy Mo at Kalina who
41:57
knows it thing or two about AI quality and We
42:01
are going to have some of the most Impressive speakers
42:05
that you could think of I'm not going to list them
42:08
all here because it would probably take the next two
42:10
to five minutes But just know we've got the CTO of
42:14
crews coming to give a little keynote. We've
42:17
got the CEO of U .com coming. We've
42:21
got Chip. We've got Linus. We've
42:23
got the whole crew that you would expect and I am
42:28
going to be doing all kinds of extracurricular
42:30
activities that will be fun and Maybe a little bit
42:34
cringe. You may see here or see me
42:38
playing the guitar Just come it's gonna be an
42:40
awesome time. We'd love to have you there and
42:42
that is again June 25th in San Francisco See you all
42:47
there