Subscribe
Share
Share
Embed
A podcast that explores how AI is transforming careers, businesses, and industries. Hosts Greg Boone and Erica Rooney deliver real-world use cases and actionable AI strategies to help professionals stay ahead of the curve.
Learn more about your hosts:
Erica Rooney, author of The AI Gap: Women, AI, and the Next Great Leap Forward
Greg Boone, author of AI at the Speed of Trust (available for pre-order)
Erica Rooney: AI is here.
Everybody's very excited about it.
Maybe they're anxious about
it, but there's this urgency.
To be involved and to
be doing all the things.
And at the same time, I feel like we need
to take a few steps back to make sure
that that data is where we need it first.
And so where do you
think that balance lies?
Pooja Basu: I'm a data person
and I'm a perfectionist.
She's like, who you talking to?
So I might say you have to wait for the
right data, but I also know, right, like
you can't always wait for various reason
you are trying to get in the market.
You might lose competitive advantage
if you are not the first one to enter.
So you'll have to think like anything
that I've built, is it consumable?
We talked about the pizza earlier, right?
Where you are throwing everything.
If you're just trying to get speed and
throw everything, can somebody eat it?
No.
If they can't eat teeth,
then what's the point?
Then you are just wasting
your time and money.
Greg Boone: AI isn't the future, it's now.
And whether you're in hr, sales,
operations, or leadership, the
choices you make today will determine
whether you thrive or get left behind.
Erica Rooney: Welcome to the
AI Voicer Victim Podcast.
We are live at the AI Powered Women
Conference here at MIT, and I am
sitting with none other than Puja.
Vasu.
Welcome to the podcast, Puja.
How are you?
I am great.
Very excited to be here.
Oh my goodness.
We also have Greg Boone in the house,
but Puja, we wanna hear about you
because I've heard you're an AI badass.
Pooja Basu: Thank you.
I'm Puja.
I have been in it.
Technologist for years, I started my
career in data, and right now I am
a VP of data platforms at Fidelity.
Been using ai, a very avid user of
Erica Rooney: ai.
Mm. I'm so excited that you're here
because we're talking about why it is so
important to have the right data, and I
think this is something that really makes
people very nervous for a lot of reasons.
Number one, they're scared
to put their data in.
Number two, a lot of
people have dirty data.
That's, and so then they don't
even know where to start.
So we always hear this
garbage in, garbage out.
Can you kind of break down
for me why it is so important
to have clean, accurate data?
Pooja Basu: So imagine you are
making smoothie, and now you
have sour yogurt, rotten bananas,
black apples, filthy oranges.
How do you think that?
Smoothie would taste like,
I'm not gonna enjoy it puja.
So it's the same thing when you
are making any AI solution or
analytical solution where having
clean data is very, very important.
You need to know who your customers
are, whom are you building the data.
Do you have not only accurate
data, but fresh data, people's
preference change over the time.
For example, I might like Banana
earlier, but I don't like them anymore.
So you, it's very, you need to make
sure not only you have clean data, you
need to have accurate data as well.
The third is you need to make sure
you have some governance around it.
You mentioned people
get very uncomfortable.
Nobody wants to share their data if they
are not getting enough value out of it.
If I'm getting enough value out of
my data share, I might be willing,
but I wanna know how my data is
being used, is it used against me?
So having a proper governance framework is
very critical for building an AI solution.
Erica Rooney: Mm. When you're
thinking about that governance
framework, what are some of the key
components that should be in there?
Pooja Basu: So one of the key component
that immediately comes to my mind
is, or two I would say, one is,
are you basically masking the data?
Like is, are people going to be
identify that this is Puja Basu who
said it, or it's more generalized.
The more you make it generalized and
you remove the personally identifiable
information, that's critical.
The second one is how long
you are retaining my data.
And again, I'll give you
a live example, right?
For example, if I got caught speeding,
DMV stores my data, I don't want DMV to
store that data more than three years.
That's what they were supposed to.
I don't want any AI model in future that's
trying to predict my driving insurance.
Use that drive back one bad
driving incident in future.
So this example in a way, ties both
personally identified the example
where I want, after three years,
they dem mask the data completely,
and if possible remove it.
Mm. That's such a good point.
Erica Rooney: I didn't
even think about that.
I wanna talk about though, having the
right data, because I know, while I'm
thinking about creating all of these
agents, sometimes it's like, well, I'm
just gonna give it everything I have.
Like, here's my book, here's all
my trainings, here's transcripts,
and, and it's really too much.
So speaking
Greg Boone: from personal experience,
it sounds like, I mean, IM.
Erica Rooney: Speaking
hypothetically, y'all.
Hypothetically speaking.
Okay, but so how does it, how does
it make a difference from like
the right data to too much data?
Again, we'll go back to the food
Pooja Basu: example,
Erica Rooney: but I'm more smoothie.
I love it.
I love it.
Pooja Basu: Imagine you are making a
pizza and do you wanna add chocolates
and tomatoes and onions and olives and
that cupcake you have in the fridge?
Erica Rooney: Not really.
Who wants to Exactly.
Yeah.
Not really.
I love these, these are real tangible
examples though that, you know, it's like,
talk to me like I'm a fifth grader, but
that's what makes sense for people today.
Exactly.
So it's so important.
I mean, this thing about data, what
really makes me nervous about it
too though, especially as an HR
professional, is like we just have so
much dirty data and I don't know that
people under understand that it's dirty
or they don't know how dirty it is.
And so like what is that
first step when it comes to.
Cleaning it all up.
Pooja Basu: See, when you are trying
to clean the data, right, you need to
make sure it has the right context.
Again.
For example, my first name is Puja.
My last name is Basu.
If you mix it up, that doesn't
make for a good uh, solution.
The next solution, I'll get a little
bit technical, is around master
data management and when, again,
I'll go with example, I might have.
I do my banking at Bank of America, and
I have three different accounts there.
I might have one account as just P Basu.
The other might be with Puja B, and the
third one is like my full name, Puja Basu.
If Bank of America Teka cannot tie that
all my three accounts are together,
suppose one of them is brokerage account.
They might just reach out to me later
on saying that, Hey, you haven't
opened up a brokerage account with me.
Can you open an account with us?
And I'm like scratching my head,
you don't even know me as a person.
I'm already your customer.
I already have that type
of account with you.
So as you are trying to build
clean data, you really need to
figure out, right, like how do you
remove duplication from the data?
How do you remove stale data?
How do you remove unclean data?
We talked about masking the data.
So can you remove per personally
identifiable information
if you really don't need.
So those are few steps you need to
take for making your data clean.
Greg Boone: Yeah.
We have a talk about like
data harmonization right?
To to that point, right?
You use your name.
For example, in, in previous companies,
we would talk about the idea of like, in
one system it's IBM and another system,
it's international business machines.
And another, and then people don't
understand the redundancy and the
challenges that come from that, you know?
Pooja Basu: Yeah.
And you might not even then
understand how much total business
you are doing with that company.
Right.
Just because they are
in different, uh, names.
And if you are going and meeting somebody
senior there, you might just give a
random number and the other person
that your client there is scratching,
like, oh, I do billion of dollars of
business, and they don't even know.
How much business I'm doing
with them, they might lose trust
Greg Boone: a hundred percent.
Erica Rooney: But here's where like my
brain is also going is because AI is here.
Everybody's very excited about it.
Maybe they're anxious about it, but
there's this urgency to be involved
and to be doing all the things.
And at the same time, I feel like we need
to take a few steps back to make sure
that that data is where we need it first.
And so where do you
think that balance lies?
Pooja Basu: I'm a data person
and I'm a perfectionist.
She's like, who talking to?
So I might say, you have
to wait for the right data.
But I also know, right, like you
can't always wait for various reason
you are trying to get in the market.
You might lose competitive advantage
if you are not the first one to enter.
So you'll have to think like anything
that I build, is it consumable?
We talked about the pizza earlier, right?
Where you are throwing
everything if you're just.
Trying to get speed and throw everything.
Can somebody eat it?
No.
If they can't eat, eat,
then what's the point?
Then you are just wasting
your time and money.
So you'll have to think about who are
your consumers, what problem you are
trying to solve, and do you have right
enough data to at least build an MVP?
So if you take an MVP approach and
keep customer at mind first, what are
the problems you are trying to solve?
Then you might not have to wait
for making all the data, right?
Mm. So you're saying start
with that MVP in mind.
Start small, have a very clearly
identified problem, statement in mind,
clearly targeted consumers of that
solution, and go after that small data.
Taking that step by step
approach might be more, uh.
Realistic than saying, I'll wait
for all the data to be ready
and then I'll do something.
So having more of a use case based
solution might be more, uh, achievable.
Erica Rooney: Yeah.
So Puja, I love this because we had a
conversation earlier with a woman who was
all on the branding side of the house, and
so she was very much so talking about you
have to understand your problem and your
client, but you can use AI to do that.
Yes.
So I think instead of.
People looking at it like this, I've
gotta have clean data that's gonna take
me six months to really get it all clean.
Like use AI to do what you're saying.
Build that problem statement, get the MVP,
go a little slower when it comes to making
sure you have accurate data, good clean
data, and then use AI to accelerate again.
Pooja Basu: Yeah.
You can't build a zenga tower without
having the right building blocks in place.
I love that.
Greg Boone: I love these analogies.
Pooja Basu: I know she's,
she's like, just spin around.
Greg Boone: I, I wanted to go back
to like the, the whole like, uh,
consumable, uh, part of it I've been
telling folks I just got back from
San Francisco, um, last night from
the HubSpot's, uh, inbound conference.
And one of the things I was telling
some of the sales folks there is like,
you can't sell things in the age of AI
that people don't know how to consume.
Right.
One of the most challenging things
are folks who are trying to push
these tools and these technology.
But the people on the other
end, like, who is it for?
Right?
If you don't know how to consume
these things, like it's never
gonna go anywhere, you know?
And then I think Erica was very, uh, uh, I
don't know what the right word is, but the
idea that all this data's gonna be clean
in six months, that's very optimistic.
View, optimistic.
Consider
she, you know, Puja mentioned,
uh, MDM or Master Data Management,
and people have been talking about
big data for the last 15 years.
Right.
And people are still like,
my data's still not clean.
I'm like, it's, you
can't be a perfectionist.
So I, I like that advice as well, you
know, and, and being here at, uh, MIT,
you know, uh, this is just amazing space
to see kind of this level of innovation.
Like what.
When it, what are you hearing
as it relates to, what are
some of the challenges to get
people to adopt or consume?
I don't mean to take Erica off of
track of the questions, but the
consume piece is something I keep
talking to folks about, so I'm trying
to understand what you're saying.
Pooja Basu: So we just came out
of a panel, right, where they
were talking about uncertainty.
So I think one of the
biggest fear that I see.
People not wanting to use it
is they are really uncertain.
How will it impact my job if I let
too much AI creep into my space?
Will I lose my job?
So that's one piece.
The second is literacy.
You won't believe.
Like recently I was going for a
walk with my friends and we were
talking about chat, GPT and two of
my friends had never used cha g pity.
And I was like, how come?
Like, I was just amazed.
Baled, I know, I, I gave them
tips like, you gotta go and try
it just for small things because
it genuinely makes life easier.
So I'm still amazed.
Like even in today's world, there
are people who just don't even
know how to use the of people.
The basic, A lot of people.
Yeah, a lot of people.
Right.
So how do we train them?
The second is, the last is like the, are
they getting the value that they want?
If I go to an AI solution,
I'll give another example.
I love it.
Greg Boone: It's gotta be food though.
Is it gonna be food?
Pooja Basu: No, it's not food this time.
Oh man.
So I was using charge GPT and I wanted
to create an image about something.
And I kept on asking Chad
GPT to create an image.
At first it created in some
language, I don't know, German.
I thought it was German.
Then I asked Chad gt, Hey,
can you create it in English?
And it wrote something.
I knew it was not English, and I'm
like, Chad, g, pt, you are wrong.
So if your solution is not giving
consumable product that I can't use,
people will eventually lose faith.
Like I still use chat GPT,
but I've learned my lesson.
Don't ask to.
Don't ask it to create images.
So is your product, uh, are your
consumers able to consume the product?
If not, there will.
They will.
There will be no adoption.
Greg Boone: And I think the
training is, is critical, right?
Yeah.
Because the, the point I would always,
my, my counter argument would be.
You know, it's, it's
garbage in, garbage out.
But it's also context.
In context.
Without context,
Pooja Basu: it would make no sense.
Right?
Yeah.
Because
Greg Boone: it's, it's trying
to, you know, hallucination is
not a bug, it's a feature, right?
It's intended and
Pooja Basu: it's so confident,
like for me, that first image it
generated, it's so confidently
presented, this is the image you want.
And I'm like, don't know.
What language is this in?
Greg Boone: So, so what I would
say then my pushback would be then.
Training people on how to actually
create the right images, understanding
like, Hey, maybe you shouldn't
try to get texts as a part of it.
Because as I always try to explain to
folks, it's not writing, it's drawing.
Yeah.
Which is radically different and
it just continues to progress.
But, but, uh, the point is well
taken, it, it doesn't matter.
If your experience is one of which
you feel like you can't consume it
and going back to the point you made
earlier and there's no value in that
handoff, then you're gonna not adopt it.
You're not gonna use it.
And we're seeing this every day inside
of workbook, inside of the workplace.
Pooja Basu: So there
are two pieces, right?
One is the user need to have enough
context, like I learned, okay, charge GT
is good with text, but not with images.
The same is like, now how do you
keep on evolving your models so
that it's more consumer friendly?
And Chad g. Pity, maybe like
doesn't produce image, it stops
like, Hey, I'm better at this.
It might, it could
Erica Rooney: be very confident sometimes.
Pooja Basu: Yeah.
Erica Rooney: One question I do have for
you, because I mean, you've been in data.
15 years, you know that
stuff inside and out.
That is the scary part for me.
'cause I'm not that kind of girl.
What advice would you give for people
who are not in the data science,
in how can we protect our data?
Because I think that is a big,
you talked about uncertainty.
People not knowing what's happening.
What should I, as like a non-technical
person do to protect my data?
Pooja Basu: So there are
simple things, right?
I think everybody's data in some ways
will be used in digital marketplace.
I don't know if Will girl, I'll also
Erica Rooney: say, I think all my data's
Pooja Basu: already out there anyway.
Yeah.
So I'm not like that
worried, but go ahead.
Some people.
But yeah, the simple things you could
do, like we were having the earlier tech
session, looking at the settings of the
app you are using and if there is any data
privacy setting that you can turn it off.
Don't sign up for things
that you don't need.
So, uh, don't give lesson
number, number one.
Yeah.
Unnecessarily to somebody.
Explore all the settings.
I think those are few things you can
do to make sure your data is protected.
Is
Erica Rooney: that something
that you do just intuitively?
Is this data person, like, every
time you get an app, you're,
you're clicking the settings.
Tell, tell the truth.
Is that keeping up?
Because I'm like, accept, accept,
Pooja Basu: accept.
WhatsApp recently released a feature
right where you can turn off, uh.
So that it won't import all the chat
features and uh, up, like somebody cannot
use it in an LLM, you won't believe.
I literally went to every
single group message where I
was admin and I turned them off.
So yeah, I try to keep up as much
as possible so that my data is not
shared, at least unintentionally, but
at the same time, I'm aware that I
can't protect it completely as well.
Yes, I
Erica Rooney: love that we
need people like you in our
chats to keep us safe, Puja.
Pja, what are you most excited
for when it comes to women and ai,
especially being at this conference?
Pooja Basu: The one thing I really enjoyed
from all the things in conference today
is AI is going to make our life simpler,
and it's going to make complex in some
ways too, but some of the speakers talked
about empathy and compassion and how
we need more of that to compliment our
life that technology is taking over.
And as a woman, we are
very natural in that.
So I'm very excited about that piece,
that as a woman leader, I can combine
some of this technology advancement with
my empathy skills and compassion skill.
And be a better leader at work.
Oh, I
Erica Rooney: love it.
Oh my gosh.
Puja, thank you so much
for being on the podcast.
This is The Voice or Victim podcast,
and we are here live at the AI
Powered Women's Conference at MIT.
Thank you so much, Puja.
Thank you for having me.
Thank you.
Thanks for joining us
on AI, voice or victim.
If you want to stay competitive
in the AI age, start now.
Take one insight from today's episode and
put it into practice in the next 24 hours.
Make sure to follow us, share
your thoughts, and subscribe
for more actionable AI insights.
See you next time.