Latest / Founder Mode / AI + Automation: What to Ship First with Sangya Singh

Transcript

AI + Automation: What to Ship First with Sangya Singh · Founder Mode

0:00
And the bar is really high. I used to think good and great.
0:02
Now it's great. And next you have to say no to
0:04
great, to say yes to extraordinary.
0:22
You're listening to Founder mode.
0:23
Where we go inside the minds of builders, not just what they
0:26
launch, but also why they launch it.
0:28
Today's episode's all about sequencing.
0:30
When everything is possible, what should you ship first?
0:33
AI and automation offer huge potential, but without focus,
0:36
they can just create complexity. That's where our guest comes in.
0:39
But before we get to Sonya, who is an incredible leader inside
0:43
of Microsoft, Kevin, you, you're building multiple companies
0:46
right now that are ultra lean, moving fast, and maybe breaking
0:49
some stuff. How are you thinking about MVPS,
0:52
what to ship 1st and how to take new ideas to market in this AI
0:55
era that we live in? Yeah, it's.
0:56
Kind of cool and I'm excited to get Sangas take on this as we
0:59
talk about how she's building because, you know, obviously
1:02
before Microsoft, we built this company with a handful of people
1:06
accompli got into Microsoft, you know, and I kind of saw that and
1:09
then now been out of Microsoft for well, almost almost 10
1:13
years. And the the speed at which you
1:15
build kind of outside versus inside, I think has been, you
1:18
know, a topic that Satya has talked a lot about, you know,
1:21
kind of when he acquired us and then also over his a lot of his
1:24
public stuff. And so when we look at what
1:26
we're doing today, kind of with a tiny company, like less than
1:29
10 people, how are you building? The thing that I, I like to do
1:32
is like, almost like when you're thinking about growth or
1:35
thinking about any kind of an experiment is like, how can you
1:39
test and execute a lot of things?
1:41
Try not to find things that are one way doors where you can't go
1:44
back if they don't work out. But you know, one, I was talking
1:47
to somebody earlier and they're like, it's just kind of like
1:49
they're, you know, they're asking me like, how are you
1:51
doing so many things? And I was like, well, you know,
1:52
they're you're kind of just throwing stuff against the wall
1:54
and see what sticks. And then you're going to double
1:56
down, triple down on those things and.
1:58
Yeah, I think AI is, is, is in the same way that it's allowing
2:01
you to prototype and test even faster of the stuff that you're
2:03
building. It affords us the same
2:05
opportunities, growth, right? Like every channel, every
2:08
campaign you want to run, every idea, make door tests, All of it
2:11
just gets easier because you can spin up landing pages even
2:13
faster than before. You can spin up new channels,
2:15
you can create ads, you can do all these things really quickly.
2:19
And I think in the context of today's guest, it's interesting
2:21
because in the Satya era of Microsoft, he has really
2:25
embraced the kind of dev, you know, dev OPS, developer
2:30
relations, open source community.
2:32
They acquired GitHub, right? Like there's, there's just this
2:34
idea that like, just because we're a big company doesn't mean
2:37
we can't innovate and do stuff and, and move fast.
2:39
And I'm excited to talk to us on you about this because I think
2:42
in the same way that with products that we're building,
2:44
that you're building and the teams that you're running that
2:45
are super lean are creating things quickly and going to
2:48
market quickly. And you stole my Thunder.
2:49
You're like, I'm going to do this outbound thing and I'm
2:51
going to go test it really fast. It's true.
2:52
Like you can do that now and AI helps with a ton of it.
2:55
And so I think it just makes us get faster.
2:57
Big caveat is don't get like, don't make it sloppy.
3:00
Like it's so easy to just like write bad code.
3:03
It's so easy to like put out bad ads, like take the time to use
3:07
your human brain to review the stuff.
3:09
That's my only like quick like edit on that.
3:12
But you know, we probably probably need to get get her out
3:14
here. Should we should we get her out
3:16
here? Let's do it.
3:27
Sangya, welcome to founder mode. You spent years like, you know,
3:30
turning powerful ideas into real world problems, automation
3:33
systems, AI first workflows. And you know, you've done it all
3:36
with this focus on clarity, impact and inclusivity.
3:38
And today we're talking about like what it is to like ship 1st
3:41
and AI and automation because like sequencing, especially in
3:44
this space is everything. Yeah, Thank you so much, Kevin
3:47
and Jason, that's a it's a great opportunity to sync up with
3:50
Kevin again and get to know Jason too.
3:52
And you're amazing audience, so looking forward over to the
3:55
conversation. Let's let's get into it.
3:56
So you've worked across product innovation and intelligent
4:00
automation. What drew you to this kind of
4:03
intersection of AI and enterprise workflows?
4:05
Yeah. So just to bubble up a little
4:07
bit, I've been at Microsoft for about now almost 25 years and it
4:11
seems like such a long time. We come from a consulting
4:14
background, but I kind of was with Microsoft for five years
4:18
around one problem domain and my hip pocket skill has always been
4:22
to integrate something that the world hasn't seen into
4:26
applications. So when I first started, it was
4:27
RFID into business solution. Then it was like when Kevin was
4:31
part of the journey at Microsoft, it was really about
4:34
how do you make emails social again with, you know, likes and
4:36
AT mentions. And about six years ago, I was
4:40
asked to come into this world of, of low code, no code
4:44
platform actually. And I've never worked in dev
4:46
tools before. And it was for me, it was about
4:49
how do you democratize coding, which, as you know, in the past
4:52
three years have taken a whole new meaning with the Gen. tech
4:55
automation. And so a, my hip pocket skill
4:58
has always been about when an invention is done, in this case
5:02
at a large language model, how do you really apply to business
5:06
applications and truly create innovation?
5:08
So for me, it was a natural brand that I had that I was
5:12
called called to pull in. But it's, it's a similar
5:15
systematic pattern that I've done.
5:17
And as a product builder, it is going back to Thunder mode, you
5:22
know, Ground Zero to say, Oh my God, I can, we can totally
5:25
rethink how people actually solve business problems.
5:28
So that's kind of what drew me into it.
5:30
Yeah. So anytime I think I go back
5:33
into a moment where you have to truly transform a product line,
5:37
I think the number one thing goes down to a couple meta
5:41
points #1 is really focusing on the strategy to win and strategy
5:46
to win as you know. But if you have been in this in
5:49
this world is really a hypothesized logic of success,
5:54
how am I going to create value? And for that, you have to just
5:57
fall in love with the problems faced first.
5:58
So first, like I go back to before you when I hire my other
6:02
founding members is what is the strategy to win?
6:04
So when I'm called by a leader, I give away like give up tons of
6:08
big orgs and I just focus the first two weeks to really
6:11
defining the problem space and the strategy to win.
6:14
So design thinking was always about that.
6:17
It was always about the why behind it, not you know, when
6:21
and you know, revenue and all of that stuff is really, really
6:24
cool. So I would say defining the
6:27
problem space strategy, which is a hypothesis regardless of
6:31
resources, right? And then really, really then
6:34
deciding what kind of talent you want to bring in.
6:37
So you kind of build the organization once you know what
6:41
the strategy to win is, which is very opinionated.
6:43
And then then you know, you know, the basic drill around
6:46
coming up with the MVP. And that's just a start to how
6:49
you actually find product market fit.
6:51
And then kind of what I'm known for is 10X once the product
6:54
market good fit is there, then actually mainstream adoption
6:58
because mainstream adoption requires very different things
7:00
to get there. So it very naturally falls into
7:03
it. You kind of also want to know
7:05
your blind spots, you know, great plans can gather dust for
7:08
a long time and you know, strategy and we've all heard it.
7:10
So it is giving time constraint a little bit and you don't have
7:15
all the answers, but how do you make the best decision around
7:20
it? So for me, really spending time
7:22
with the customers, really, really understanding in the as a
7:25
gentic world, like truly understanding what their
7:27
business process is, what the struggling moments are, what the
7:31
forces against the jobs to be done is, you know, going back to
7:34
those first principle basics around it.
7:37
And then really, really saying, wow, now I understand all the
7:40
players. Where should I focus on?
7:42
What is my niche going to be? That's great.
7:45
I will simplify it a little bit just to repeat it back and make
7:48
sure I got it. But I think what I'm hearing you
7:50
say is like you kind of need to know what your problem is that
7:52
you're solving and then it's what is the strategy to help you
7:55
win in that problem space? And then what are the resources,
7:58
people, team systems you need to do that?
8:00
And then maybe product market fit is one thing, but how you
8:03
scale and you know, kind of hit broad market adoption totally
8:06
different. So I love the kind of the
8:08
thinking there. So you mentioned your customer
8:11
pain, is it technical feasibility?
8:13
Is it market timing? Is it all of the?
8:15
How do you think about that? Yeah.
8:16
So think of it as in my mind and the way I think of it is like an
8:20
equation of impact come from when you figure out the value
8:24
that you're creating kind of what we just talked about.
8:26
And agility is very important, especially when you're going to
8:30
such a big platonic platform shift.
8:32
Because no matter how much I can hypothesize, it's only the
8:35
shipping that you learn and just the bar of agility is just
8:38
increased so much. Why?
8:40
Because you all know like the frontier models every week just
8:43
gets just potentially just multi, just better.
8:46
Sonnet 4.5. Holy moly, 30 hours of
8:49
autonomously coding. And we were just talking just
8:52
few months ago. Oh, replit agent 3 can code for
8:55
three hours and come up with an app and we thought that was
8:58
amazing. And 30 hours to build something
9:00
like a slack or a teams that it can do that.
9:03
It says way beyond. So it's like so for for I have
9:06
to always keep in mind value and agility has gone up like 10X100X
9:12
time. So everyday, daily, a reminder
9:15
for that. And you know, there's, there's
9:17
this thing about will to innovate and skill to innovate,
9:20
you know, like, and for skill, I always think about three things
9:24
that I, I look for just making sure every day, every week we
9:28
are churning value. That's not about a month.
9:30
It's almost every day. There's this notion my team
9:32
calls 16 hours. We have to come up with the with
9:35
the note and answer. Reason being because I tell the
9:38
team you have 16 hours that you're awake.
9:40
I know that. So do not tell me I'm going to
9:42
come up with an answer next week because in a weeks time we can
9:46
try three different teams fail and we can have a faster thing.
9:50
So the three component that we always remind ourselves is
9:53
creative abrasion, which is we got to have different point of
9:56
opinion and intellectual chemistry.
9:58
If you're not challenging each other, that doesn't work.
10:01
You can continue to have creative abrasion, but you also
10:03
need to have creative agility, make decisions fast and create a
10:08
resolution. So this is a really good
10:10
structure for us to think about it, but we just apply it daily
10:15
and weekly because world just changes for us every week.
10:18
So that's kind of how we think through it.
10:20
You know, once we we come together, we know the strategy,
10:23
we know the will, then it's all about execution and these three
10:26
components. You know, we just make sure it's
10:29
ingrained in our routines and culture.
10:33
What's like one feature that felt like risky to prioritize
10:36
early, but then turned out to be like a turning point for you
10:39
guys? I would say that one that comes
10:42
to mind is I used to run this product that does the classic
10:47
robotic process automation RPA. And then we got computer use
10:51
capabilities in large language model.
10:53
And we were really trying to see how do we apply that to RPA.
10:58
And we thought about one of the problems with RP.
11:03
I don't know if you know a little bit about RP.
11:04
RPA is all about you old, let's say SAP systems or stuff like
11:08
that. You create UI scripts and in the
11:10
cloud you have these program box that just do UI automation just
11:14
like humans would do. That's what RP is.
11:16
But RP is very fragile in the sense that if the application
11:19
changes, it says payments and blah blah blah, JavaScript error
11:24
or something like that, it just breaks.
11:26
And so we thought about computer use, how do you use both agentic
11:30
and deterministic automation, And we thought about self
11:33
healing. It was super risky for us to say
11:36
it can we combine computer use? It can self feel if it sees a
11:40
maintenance, it can click on it and and just just use computer
11:43
use to self feel. And it was something that we
11:46
said, wow, our business was making a lot of money RPA to put
11:51
something and these large organization do love, no matter
11:54
what we say about agentic, they do love deterministic outcome
11:57
because you're processing financial stuff, insurance
11:59
stuff. So initially it felt it was
12:02
risky. And what we learned was it
12:05
ultimately it it, it ended up being one of our key
12:08
differentiation. What we learned was if you give
12:11
enough governance and control and transparency around it, it
12:15
was actually actually helped the RPA developers to actually adopt
12:19
it. So initially it was like, you
12:21
know, we've got a type 1 mindset, which is very
12:24
protective mindset, protect the business.
12:26
And then you've got the prospect mindset.
12:28
Oh my God, LLM has this, I want to use it.
12:30
And and if you have a big business and now we are like one
12:33
of the top RPA leaders, the team over indexes to protect that.
12:37
So they're very nervous about this.
12:39
So that's the one that comes to mind.
12:41
But I'm. Glad just on this, on this kind
12:43
of thread as we, as we pull the prioritization challenges,
12:47
right, like how you kind of shift the, the thinking in the
12:50
team and drive that strategic prioritization inside of a large
12:53
organization like Microsoft. What are the mistakes that you
12:56
see teams make when they're rushing to ship AI, especially
12:59
when the tech is like, oh, this is really exciting, but maybe
13:01
the value is a little bit unclear.
13:03
Yeah, I would say I actually watched TR Top 20 clips or
13:08
something more like. Leverage sounds.
13:11
Like, oh, that one's a leverage. It's just like, anyway, I had
13:14
very little time, so I did 3, but one of them.
13:15
So I think you kind of know a little bit about anytime you
13:19
have something so cool, like every day something so cool
13:22
comes up and you have all sorts of early in career to senior
13:25
people. Your natural human tendency is
13:28
to quickly say, I support this model and I support computer use
13:33
technology for the sake of technology.
13:35
But how do you really to get durability?
13:37
A lot of the stuff is like, you know, pop singers, you know, you
13:40
can be 1111 song wonder or can you be truly durable?
13:44
And so I think that's a very natural tendency for someone to
13:49
write code and really apply and say, whoa, within a day of
13:53
something being announced to say like, Hey, we can do it.
13:55
We can get the PR. We can have, you know, Satya
13:59
tweeted, you know, so that's it's it's human nature to do
14:02
that. But we always, always go back to
14:05
our hit rate is not that high, which means you can try a lot of
14:08
things, but only a few things go.
14:10
So it's, it's, it's an art and a science is what I would say.
14:14
Like I talked about, agility is very important.
14:17
I, I want to make sure there's quality.
14:19
There's craftsmanship and quality, but not perfection.
14:22
I'd rather have 100 ideas go through and then find where the
14:25
hit rate is just because we're not good at hit rate, right?
14:29
Then wait too long and only do 4 ideas.
14:32
So I would say when prioritization comes through,
14:36
couple things that we talk about is, is kind of project thinking
14:40
versus product thinking, which is the why behind that.
14:43
There's a just a template when you come up with it, just say,
14:45
what is it, Why are you doing it?
14:47
And what is the outcome rather than just the output around it.
14:50
And it's no longer about being a multiplier effect.
14:53
Like what are we going to learn? Is there something exponentially
14:56
that we're going to learn? And it's data, data, data,
14:59
evals, evals, evals like don't come to the conversation has
15:02
changed, right? Try 100 experiments wherever we
15:05
see 5 to 10 points better increment.
15:08
Let's just go to 10% of the population.
15:11
So looking at leverage ideas is a skill, but not a lot of time.
15:17
Leverage means something that has an exponential impact rather
15:21
than multiplicative impact. I don't get it right all the
15:24
time. So sometimes you'll spot it as
15:26
leaders, but most of the time we don't.
15:28
So we go back to our rules of engagement, which is really
15:31
around eval and not about number of like we don't reward people
15:37
by saying number of blog post or launches that you do at a
15:40
conference that you do. We call those outputs, but we
15:43
really talk about outcomes. What have we learned eval data
15:46
and stuff like that. So it's it's a combination of a
15:48
lot of that, however, to get the right licensing and and stuff
15:51
like that. But what I've learned Kevin, at
15:53
Microsoft and it just blows me away so much as because we have
15:57
to go so fast. So we say there is no religion
16:00
on to use something that Microsoft only builds.
16:03
So, you know, Kevin's been there under Satya's leadership.
16:06
The whole mindset has changed under Satya where just adopt
16:10
what is needed to understand what do we need to do to build a
16:14
platform that our customers can develop on?
16:17
And so if that means we got to go to brain test, if that means
16:20
that are we just not going to do open AI, we're also going to
16:23
work with clod and Sonnet and you know, all of that stuff.
16:25
So that's always been our thing, like we've been pretty open to
16:29
adopt stuff that will help us accelerate.
16:31
So that's number one, brain trust initially to get started
16:35
with it. But then what I also see that's
16:37
amazing at Microsoft, Microsoft's like a big tanker
16:40
ship. It might seem.
16:41
It's moving slowly, but boy, when it enters, it enters really
16:44
well. What I also see, they invest a
16:46
lot in tooling, so you're not waiting for the perfect tool to
16:49
come. But you know, people from
16:51
Windows and Office and Cloud and you know, Azure, they all come
16:55
together and it's just amazing. Such a large organization.
16:59
They're so connected and they'll come with the tool tools
17:02
Hackathon is going on right now across the the company.
17:06
The main focus is tooling. How do you build tooling?
17:09
And then when something amazing comes out, it just gets adopted.
17:12
And so the agility of information, we do so much
17:15
learning. Oh my, this team did it on brain
17:18
trust. We like go, go, go.
17:19
You know, if you're pulling in another team, go with brain
17:22
trust right now. Don't wait for the perfect
17:24
because you haven't reached that scale.
17:25
A lot of the teams are starting to haven't reached the scale
17:28
office and stuff like that, and when you do reach the scale
17:30
there will be tools available for use.
17:32
Yeah, the scale thing is something I remember.
17:34
I remember when we were there, like Microsoft Teams was getting
17:37
started and I rolled out to their office and they said,
17:39
yeah, we're starting on this floor and build Bellevue.
17:41
And I'm like, that's kind of cool.
17:41
And I went up and met the guy and there was like 3 people
17:44
sitting on a floor that held, you know, probably 100 or 200.
17:48
And I was like, this is a big floor just like the three of you
17:50
in this cube. And he's like, Oh yeah, we have
17:51
the floor above too. Literally went back like a month
17:54
later, 400 people working on the thing.
17:56
Like they just, I mean, like the mat, the like the amount of like
17:58
people that they can assign to a project is like so scary in
18:02
terms of like how fast they can move, but also like a testament
18:07
to how quickly you can organize a really, really large body of
18:09
people to go after a problem that, you know, at the time,
18:12
Microsoft deeply wanted to care about solving Teams.
18:15
How do you think about Microsoft and the teams that you lead and
18:18
that you're talking about internal tools, but also what
18:20
you what you ship and bring to market in the context of like,
18:24
how do you compete in this, in this new world where there's
18:27
there's a new app spinning up everyday to kind of, you know,
18:30
it used to be what is it? Marc Andreessen said software is
18:32
going to eat the world. It feels like AI is eating
18:33
software and Microsoft is A at its core, a big, very, very big
18:37
software company. So I'd just be curious to get
18:40
your perspective on. Yeah.
18:41
Let me think through maybe a couple meta points here.
18:45
The first one is, I think, going back to the analogy around
18:48
speedboat startup P kind versus tanker ship, where you've got
18:53
lots of different levers that need to be coordinated, But when
18:56
you do it, you actually serve a much larger audience.
18:59
So there's this competency we talk about at Microsoft a lot. 1
19:03
of the competency for leadership is about awareness of context.
19:06
So as a strategist, the strategy that I define for Microsoft
19:11
versus if I join Replit or a startup or a stealth mode with
19:14
Kevin, you know, it's very different.
19:16
As Microsoft, what we define success is something that can
19:19
truly, truly get into billions. So that's number one.
19:23
So a, no matter what the idea is, it has to have such a big
19:28
unmet need or transformation and it has to be in Microsoft
19:32
ecosystem, enterprises and stuff like that, productivity and
19:36
stuff like that. And so if you come up with from
19:38
left field where you're not even playing in that field, it just
19:42
won't work. So awareness of context is
19:43
super, super important. Once you have that, then it is
19:47
really depends on the idea that you have.
19:50
Once it's in the context, Microsoft's, you know, the
19:55
brand, the, the area, because there's, there's not going to be
19:59
a lot of headwind around it then about how do you go fast?
20:03
That really comes from the top for out there, frankly all the
20:06
way where he's he's very hands on like he cannot enter any.
20:11
Review with him. He will just try it out right
20:14
there. Like you know, like Kevin knows
20:15
that, you know, like if you come in and say you don't have Sanga,
20:18
you don't have mobile, you know, outlook mobile calendar mobile,
20:21
not out like like no. So then it's all about it's no
20:25
longer 70 page specs that the PMS write anymore.
20:28
It's the agent PRD through like coding and stuff like that.
20:31
So it's truly, truly each one of us culturally transforming and
20:36
it goes down to how much of AI am I using?
20:38
What is the AI maturity that you're using?
20:40
And then, you know, reiterating that how important it is to feel
20:44
for it, which is something so common in your founders mode
20:47
thing, you know how you do so that's kind of what it is.
20:50
Number one thing is like having a Google satellite view.
20:54
I always say like the astronaut view we call racism, call it
20:57
wait a second are. Microsoft people allowed to say
20:59
the word Google Now? You didn't know this?
21:01
That's cool. Yes, they use Google Satellite.
21:04
I think that's so cool. I was caught off guard.
21:08
I liked it. Yes, yes, so really, really you
21:10
got to know where you're heading and is it in the right realm of
21:14
Microsoft? That's awareness account.
21:15
Then you got to like at the end of the day, you have to come up
21:18
to the map street where you have to define the paths to do that
21:21
and you got to go that really, really fast.
21:23
So then we go down to so it's it's always an end.
21:25
Once you've defined that, then you just see it just, you know,
21:28
from senior VP to Sati all the way, we just hone in.
21:32
If you believe there's a huge potential to it, you just hone
21:35
it. And then it's about how fast can
21:37
you do? How fast can you have this
21:38
capability and put it in M2 thread Co pilots or you know,
21:41
from vehicle that can go big and we can learn much, much faster.
21:45
How do you design new automations and apps and copilot
21:48
chat and you know all of that stuff.
21:49
Then it's all about every hundred days.
21:52
We say if you cannot release something in 100 days or
21:55
something of great value in 100 days and it's too slow because
21:58
you just want to not be able to, to to learn because the world
22:02
would have changed in 100 days. And so it's it's about.
22:05
How does that help you with your product sense and make better
22:07
product decisions? My God, Jason, if somebody could
22:10
just pay me a lot. And I keep going back to school.
22:13
I keep going back to school. I keep.
22:14
I'm pretty. Sure that Microsoft would be
22:16
happy to help you go back to school anytime you want to learn
22:19
anything you want, yes. Yes, as long as I can pay them
22:22
bills. Also I would.
22:24
You know, Social Security tells me that I would live till 86,
22:27
but somehow in my mind I think it's 91.
22:29
I don't know where. That I think you will live the
22:31
91. And I think, you know, Kevin, I
22:33
just talked about this. He's only going to make it with
22:34
the 83, Kevin. So yeah.
22:36
Yeah, I ran the chat based on where you're I'm 48 and like,
22:40
what's it gonna take? And say yeah, yeah.
22:41
And so I want I wanna be in a position where I can have people
22:45
say, wow, at 90, she's going to school.
22:48
So I love, I love, love, love. One of my biggest thing is
22:51
learning is because I get exposed to these amazing and I
22:56
look very structured. Just learning is very different
22:59
things for different, different people.
23:00
For me, I keep going back to great institutions because
23:04
that's where the professors are researching what's happening.
23:08
And these are the professors that the CEOs pay per hour or
23:13
per sitting, I don't know, $100,000 and I get to be with
23:16
them for 18 months live. It's like, Oh my God, I don't
23:20
care if it takes $18,000 a year. It's like it's such small money
23:24
if you think like that with what I get, you know, so I always
23:28
follow like I read voraciously, you know, the four years, like,
23:32
you know, I would treat like 250.
23:34
And and so I've written a big medium article on how learning
23:37
works because I know how learning our own LLM who you
23:39
exposed to the types of people you exposed to the types of
23:42
books that you read too, because at the end of the day, via IWS.
23:46
So the way I keep relevant and reinventing myself is the what
23:52
and The Who I spend time with. And the bar is really high.
23:55
I used to think good and great. Now it's great.
23:57
And next you have to say no to great, to say yes to
24:00
extraordinary. So I'm like, holy moly, you
24:02
know? That's a good one.
24:03
Is that your line? You have to say no to great, to
24:05
say yes to extraordinary. Did you?
24:07
It's a good line. Yeah, it's like it's the bar is
24:10
so high. Like I'm, I'm teaching.
24:12
Even my kid, my my older 1 is in college now.
24:14
He's like, wow, I've been asked to mentor this one.
24:16
I was like, OK, you only have limited energy.
24:19
You have to what is your leverage task exponentially.
24:22
Who do you want to be with the great?
24:24
I was like, I used to think, oh, good to great.
24:27
Then vocabulary is so important. That's why I love reading and
24:29
being with people. So I did not invent this.
24:31
That's why I like learning because my LLM has picked this
24:34
up from some point in my neuron, you know, and to say like, Oh my
24:38
God, I have to say no to great. So I can say yes to
24:40
extraordinary because the time I spend with extraordinary
24:44
managers and leaders and professors and, and, and, and
24:48
read about someone great from back history, like Abraham
24:52
Lincoln, what he did, I was like, Oh my God, you know, it's
24:55
like, it's like, wow, because I want to solve world.
25:00
I'm just kidding. So learning is super important.
25:03
And I also know that idle mind is not good because then brain
25:08
just just, you know, anxiety. Oh my God, LinkedIn post
25:11
everyone's getting promoted and Instagram.
25:13
Every marriage is happy. So the best thing is you study
25:16
and I love study and I love learning a new topic and looking
25:20
it from a different angle, getting your vocabulary and
25:22
stuff like that. So I could go back to school.
25:25
I always keep going back to school.
25:26
I love it, Sangia. Last question, you've hinted it
25:29
kind of what's next in the space?
25:30
Like what are you excited to see teams shipping in 26 and beyond?
25:35
Yeah, so I think there's so much focus right now, Kevin, on how
25:39
to build agents because a gentic thing is superpower right now.
25:43
It can do stuff on a behalf because I'm a futurist.
25:46
I always think what's ahead next?
25:48
And I think we have to get to a we have to go upstream a little
25:53
bit, I imagine, not just a chat interface in AI surface, which
25:59
truly, truly can actually help you to even find what to
26:04
automate. Yeah, understand.
26:06
Like, you know, imagine AI don't know a voice based interviewing
26:09
persona in this AI surface that talks to all the leaders, talks
26:13
about tell me about your business process and stuff like
26:16
that and automatically extracts user needs, shows you
26:19
visualization of user stories and so truly, truly helps you
26:23
elevate and see what do I even need to automate and how do I do
26:28
it transformatively? So I think there's this huge
26:30
power in LLM that I want to extract and create innovation
26:34
from and, and all these books that I have of all the leaders
26:38
who did Bob Modesto, who did jobs to be done, all those
26:42
insights. What if it's the LLMS are
26:44
fine-tuned with the best of the minds democratizing that.
26:49
So it's all about how do you democratize this power?
26:52
So right now we're dealing a lot with how to build the agents and
26:56
how to govern it and how to know.
26:58
So that's one one area, as I would say is I want to reinvent
27:02
on that part. And then the other part of them
27:04
super, super excited and curious about, I don't have all the
27:08
answers is mech interrupt, which is AI safety.
27:11
You know, you don't know if a model is misaligned or not.
27:14
And we know by 20-30 it's going to be running all the process.
27:18
So how do you know internals, not black box thinking, not just
27:22
chain of thought, you know, because chain of thoughts, it
27:25
could maybe misaligned with the goals that it has.
27:27
So how do you do AI safety and Mac and trip?
27:30
And how do you create tools for visualizations for organizations
27:34
to actually see what's happening?
27:36
And then how do you fix it? So those are the two, I would
27:38
say, problem areas. I don't have the answers, but I
27:41
know we have to invent. So where can people find you,
27:43
Sonya? We want folks to be able to
27:45
reach out, connect and find you online.
27:47
Where can they find you? Maybe attend upcoming talks or
27:51
hear more from your thinking. So I'm on LinkedIn, started
27:55
writing a lot and sub stack and stuff like that.
27:57
But you can find, I'm a big learner.
27:59
You can find a big article on Medium that is very popular, how
28:03
learning works, AKMS daily. Again, I can send you the link
28:06
for your readers. If you are going to be in Vegas
28:09
in October. I am interviewing Michael Phelps
28:12
at a big conference, power platform conference.
28:15
I get to talk to him about leadership in and the audience
28:19
is all AI related agentic stuff. So I'm very excited about it.
28:24
Two years ago I got to do Adam Grant.
28:26
So I do speak at conferences and I get to do fire chat with some
28:30
amazing top leaders. So mainly on LinkedIn is kind of
28:33
where you see, but I'm not a big, big.
28:36
All good, all good Sangha. That was clear, grounded, like
28:39
deeply thoughtful. And thank you so much for
28:41
joining us. No, thank you so much.
28:43
It's been a blast. It's been an honor.
28:45
Dude, that was one of those conversations where you want to
28:47
just pause and take notes. Sangha gave us a real framework
28:50
for thinking about what matters and you know, really what
28:53
building with AI looks like at Microsoft scale and and really
28:57
some cool insights into kind of how things are are working there
29:00
today and quite frankly, vastly different than when I was there,
29:03
which is really cool to see. Yeah, it was.
29:05
It was a fun one and she reminded us that velocity
29:08
without clarity is just noise. I loved Say no to Great so you
29:13
can say yes to extraordinary. Kevin, what were your top five?
29:16
Dude, that quote was good. That was like a money one for
29:18
sure. And so yeah, the key takeaways,
29:20
I mean, kind of this strategy first then the org follows
29:23
customers define that real focus ship fast and then that learn
29:27
daily, right? Like how do you kind of iterate?
29:29
And then really it's about outcomes over like the outputs.
29:31
It's like, what are you actually achieving versus like what's
29:34
coming out of the factory? And then this notion of like
29:36
self healing automation really wins, right?
29:38
It's like how can it sort of like use automation to help do
29:41
things? Better, Yeah, If this helped
29:43
clarify how to think about AI and automation, share it with
29:46
someone launching their next product, Sprint.
29:48
And if you're building systems that serve real people, don't
29:51
forget to subscribe, leave a review, and send this to your
29:55
team. That's a wrap because it's not
29:57
just about what you build, it's about what you build first that
30:00
defines everything that comes after.