March 7, 2025 • 51 mins
đź’ˇ What does it take to build 2 multi-billion-dollar companies?
In this episode of the Prime Venture Partners Podcast, we sit down with Dheeraj Pandey, Co-Founder of Nutanix & DevRev, to unpack the real challenges, pivotal decisions, and insights behind building and scaling a successful enterprise tech company.
00:00 - Introduction
00:52 - Dheeraj’s Early Life
04:15 - Leaving IIT & Taking a Second Shot
07:40 - Studying Computer Science at IIT Kanpur
10:55 - Early Career: Oracle, UT Austin & Trilogy Software
14:20 - The Entrepreneurial Itch & Starting Nutanix
19:30 - Scaling Challenges & Near-Failure Moments
26:35 - The “Tightrope Mentality” for Founders 🎯
33:05 - Going Public: Nutanix IPO & Key Lessons
39:40 - AI’s Role in Enterprise Software 🤖
43:10 - The Shift to DevRev & His New Vision
46:05 - Advice for Founders Scaling SaaS from India
49:00 - Final Thoughts & Key Takeaways
🎙 What’s Inside This Episode?
🔥 Why he left IIT after 2.5 months and came back stronger
🔥 How he built Nutanix into a multi-billion-dollar company
🔥 The risks, near-failure moments, and biggest lessons he learned
🔥 Why AI is reshaping enterprise software & what’s next
🔥 His take on scaling a startup, hiring, and finding product-market fit
🔥 How DevRev is redefining enterprise software with an AI-first approach
📢 This is not just a startup success story—it’s a masterclass in resilience, innovation, and leadership.
With DevRev, Experience conversational AI at work. Startups can get $10K in credits (12 months free) to accelerate their AI-first journey. Apply now - https://devrev.ai/startups/apply
đź”” Subscribe & Turn on Notifications for more game-changing insights!
#StartupJourney #SuccessMindset #BuildSomethingGreat #TechFounder #Startups #Nutanix #DevRev #Entrepreneurship #TechInnovation #SaaS #AI #VentureCapital #Leadership #FounderMindset #EnterpriseTech #PrimeVenturePartnersPodcast
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Speaker 1 (00:00):
Once you're on a tightrope and crossing a valley,
you can't turn back.
Probably one of the rare peoplein the history of IoT is to
have quit IoT after two and ahalf months.
I was a risk-taking person fromearly on, so we picked a pretty
contrarian use case and we wentagainst the mob.
In 2011, people had started tosay that Windows is dead.
We had three near-deathexperiences.
(00:21):
You know AI doesn't care aboutdepartmental boundaries.
If you have a Google workspaceand if you have a GitHub, all
you need is Debra.
Enterprise software at the coreis about reliability and
availability and security.
Extensibility Many of theseitties in the enterprise is what
you get paid for.
Speaker 3 (00:52):
Welcome to the Prime Venture Partners podcast.
I am delighted to have with ustoday Dheeraj Pandey.
He is founder and CEO of DevRev, an exciting new AI startup,
and then before that, of course,he's very famously known for
co-founding Nutanix, which isnow much more than a DecaCon, I
guess.
Welcome to the show, dheeraj.
Speaker 1 (01:12):
Thank you, Amit.
Speaker 3 (01:14):
Dheeraj, love to hear a little bit about your journey
of sort of growing up in Indiaand what led you to kind of
eventually the road ofentrepreneurship.
Speaker 1 (01:22):
What led you to kind of, eventually, the road of
entrepreneurship.
Yeah, in fact, you know, I wasin Patna, which is, you know, a
capital city, one of the easternstates, bihar, as many of you
might know, and I was there tillI was 16, 17.
(01:44):
And I went to a couple ofschools in Patna One was Don
Bosco and the other one was StMichael's and I just, you know,
tried to be in top three, topfour kind of class.
And I used to be there, I usedto enjoy math, and mostly math,
(02:07):
I would say, more than anythingelse.
And then, you know, I took theJ in 92 for the first time and I
hadn't really prepared at allfor it.
This was the year that Igraduated from high school and I
got a rank of 1,420.
And I really wanted to come toIIT, kanpur.
(02:32):
So I'm like, okay, whateverbranch I get there, I'll take it
.
I didn't want to go to anyother IIT.
So I went there and I got civilengineering and within a couple
of months I realized that theprobability of a brand change to
computer science is way lowerthan if I were to take the JEE
again and were to be in top 100and then get into computer
(02:54):
science.
So I was probably one of therare people in the history of
IITs to have quit IIT after twoand a half months of attending
classes and I came back and Iretook the exam next year and
this time I was in top 100 and Iwas ranked 84.
And I'm like, okay, this timeI'm definitely going to get
computer science, which I didback in the day Kanpur used to.
(03:16):
You know, if you had to getcomputer science, you have to be
in top 100.
And so this is 93, startedjourney by 94,.
As you remember, india was goingthrough this massive
liberalization process.
The economy was opening up andthere was something about that
(03:37):
new India that was opening up.
And Manmohan Singh as a financeminister Bless his soul, he
passed away, but he was such akind of North Star for me.
I'm like, wow, you know, blesshis soul, he passed away, but uh
, he was, uh, such a kind ofnorth star for me.
I'm like, wow, you know, oxfordeducated, economist and all
this, and, and then, uh, andhe's really transforming the
country.
Uh, so I really wanted to be aneconomist, I want to do a phd
(04:01):
in economics, uh, even though Iwas doing computer science in
undergrad.
And then I talked to my cousinwho had just come back from the
US and he explained to me verysimilar to what's happening
right now with AI.
Basically there's a bigrevolution underway with the
browser and HTML and the WorldWide Web and HTTP and everything
else and he's like you got toreally continue computer science
(04:24):
.
It's the best time to be incomputer science.
You know and really do well incomputer science.
So, lo and behold, you know Ijust kept enjoying computer
science and a lot of math aswell.
You know, did really well.
I used to love, love math.
You know, back in the day, ifnot for computer science,
probably would have done a lotof math and been a major in math
(04:44):
.
I was a good visual thinker andnow obviously I've slowed down
a lot in the last 30 years.
But in 96, I'm like OK, we havetwo forks in the road.
I could go and join a job or Icould just apply to grad school,
and that's what I did.
I ended up applying to a lot ofgrad schools.
(05:06):
I did get a couple of jobs, butI figured I'll come to the US
and really pursue a PhD incomputer science.
So I got a few offers from likeUT Austin, Urbana-Champaign,
usc, columbia, all these places,and I ended up choosing UT
Austin because of the fellowship.
So I landed in the US in 1997in Austin, texas, and you know,
(05:28):
after a couple of years I said Igot to be in the industry right
now.
So I went on leave of absenceand for a couple of years I was
on leave of absence but neverwent back to finish my PhD,
worked at Trilogy Software forabout a year, year and a half,
in Austin, texas, and thenworked for quite a few system
(05:49):
software companies building fileservers and databases at Oracle
and then distributed datawarehouse and then in 2009,.
Speaker 3 (06:01):
basically started Nutanix.
Fantastic, very, very inspiring.
And I was going to say a littlebit tongue in cheek that
leaving IIT in the first yearwith the 1400 rank to get back
in, that was like your firstexperience of a pivot, a hard
pivot.
Speaker 1 (06:14):
It was.
Indeed it was.
It was.
I think I was a risk takingperson from early on.
You know, I think and that wasone of the first big risks they
actually took and you know Ialways thought about the worst
is not bad enough.
Why wouldn't you?
And, uh, the worst situationthere was I had gotten admitted
(06:34):
to since defense of physics andlike, yeah, just gonna do
physics and we'll figure out therest from there next time,
because I've done fairly well inmy high school, you, you know,
secondary exams and stuff.
Speaker 3 (06:45):
Absolutely so.
That leads me to sort of thefact that you worked at a lot of
different companies, includingOracle, and on file system
database etc.
In the enterprise softwarearena.
What led to the idea forNutanix, number one and number
two?
Just this notion of you want tobe an entrepreneur because once
(07:05):
you're there, you're settled.
You're doing your PhD yes, youwent on a leave of absence or
sabbatical or whatever.
What made you think that youwanted to start a company back
in 2009?
Speaker 1 (07:17):
You know, there was always an itch and I think in
2000, I think 2009, february Ihad a really long discussion
with you know, a strategicdiscussion about with my
founders of the company that wewere working at.
There were three of us whoactually knew each other for
(07:39):
almost a decade or more and Ithink the core insight that
we're trying to bring was thatSQL and NoSQL are here to
coexist and we were kind ofpigeonholing ourselves into just
SQL and that to only datascientists and that to only ad
(08:00):
hoc querying and all that likeon-demand analytics, versus what
the world was doing with Hadoop, and the developers were
getting into analytics and theywere using Java and MapReduce
jobs back in the day to reallybuild a ton of the analytical
pipeline.
And we're like look, we'rebuilding a large-scale
distributed system, so it shouldactually come together.
(08:21):
We should not take sides on SQLversus NoSQL.
But it wasn't to be.
I think we couldn't convincethe it should actually come
together.
We should not take sides on SQLversus NoSQL.
But it wasn't to be.
I think we couldn't convincethe folks there that we should
actually build a larger business, a larger company that really
brought both of them together.
So I think we just kind ofspoke for six months and said,
look, we've got to do somethingon our own, and we didn't even
realize that this was the worsttime to start a company.
(08:43):
And we didn't even realize thatthis was the worst time to
start a company Because the firewithin it reminds me of the
fire within was calledAntaragini.
For us in our IT days, thecultural festival was called
Antaragini, but basically it wasthat Antaragini.
We said, look, what's the worstthat can happen?
You go and find a job andyou'll find a job in the Bay
(09:06):
Area.
It's the cornucopia in somesense, you know.
So data was kind of at the coreof what I had built in.
Both my co-founders, mohit andAjit had you know, especially
Mohit had actually been in thisspace for like 10 years and he
and I bumped into each other atZambil, which was my file server
(09:27):
company that we were alltogether at, and so data was
going to be the thing,distributed systems was going to
be the thing.
We're going to fight, not onthe hardware vendor's turf but
on our turf, and that turf wouldbe pure software running on
commodity hardware, becausethat's all I had done myself,
you know, starting with Zambiawas file servers and commodity
(09:49):
hardware.
You know Oracle had moved tocommodity hardware with
databases, oracle RAC, exadata,and then Astrodata was also
distributed data warehouses andcommodity hardware.
So the idea was to do somethingwith distributed systems, pure
software.
And so what's the killerapplication right now?
And the killer app back thenwas VMware.
(10:10):
Like, how do you run VMwareworkloads on a distributed
architecture?
None of us knew even how tospell virtualization, like even
the V of virtualization, so wehad to learn virtualization like
to the core and then go buildsystems on top of it and make
(10:30):
sure that we got the first usecase right.
And in 2011, people had startedto say that Windows is dead
because Apple is everywhere.
Who's going to care for Windows?
And we took a pretty contrarianview that look, long live
Windows.
You know, long live Windows.
And the way it would happen isthrough a digital Windows system
(10:52):
, like you'll have to streamWindows from somewhere you know,
from a data center, and Windowswill then come together as
clusters of systems that peopleare actually streaming, because
in business world, people willstill have a ton of Windows
because of Office and everybodywas like what are you talking
about?
This is never going to work,windows is dead.
So we picked a prettycontrarian use case and we went
(11:16):
against the mob.
The mob was saying that there'sno future for Windows, if
anything.
Three years later people startedsaying things about SQL.
The SQL is dead and here we arein our long-lived SQL.
So basically we built the firstuse case around that and we
went pretty deep to US FederalVery early on.
We went to US Federal.
So there was a kind of aduality there, like just being
(11:41):
really good with both contrarianviews.
There was a contrarian viewabout going early to federal
people like okay, no, startupsgo to us federal as one of the
early segments, and it was allabout great people.
We hired some really goodpeople who were miffed with
vmware.
They were vmware uh sort offolks and they came and built
something and, you know, westarted out being, you know, a
(12:05):
really good partner for VMware.
We used to run on top of VMwareand obviously over time we
started to get a lot of theirtalent and things became pretty
testy.
They became a frenemy, thenthey became an enemy, you know,
and then we just had to buildfor all that.
Speaker 3 (12:24):
Amazing, amazing journey.
A bit of a question for youknow entrepreneurs in India who
are trying to build.
Now, of course, we're going totalk a lot about you know AI and
SaaS and all that, but likegoing back five, seven, 10 years
ago.
Why is it that a lot ofenterprise software companies
necessarily haven't come out ofIndia?
And I know there is a few rightbut, and what does it take in
(12:47):
the early days, in the zero toone journey for an enterprise
software company?
There's, of course, the techstack you have to build, where
you probably have moreconviction as engineers or
whatever.
We have good quality engineers.
But then all there's also thecustomer validation, figuring
out the initial kind ofco-creators or the co-partners
to build with, and so on.
So maybe just a little bit ofbuilding a large scale, whether
(13:08):
it's a now SaaS company, AIcompany or enterprise software
company, in the zero to onephase.
Speaker 1 (13:12):
If you have any thoughts and suggestions for our
listeners, yeah, I think Iwould say that founders many of
the founders in India are notthe folks who actually have done
infrastructure work and a lotof business software, but also
enterprise software at large isa lot of infrastructure work.
(13:34):
Like you know what we're doingat DevRev now there's so much of
infrastructure work.
Devrev now there's so much ofinfrastructure work.
It's like you have to pay thedues for the object model, the
event model, the security model,the SQL model, all that stuff I
mean, and over time, obviouslythe AI model.
All this stuff isinfrastructure.
(13:55):
It's systems engineering andthe founders need to respect
systems engineering, becauseenterprise software at the core
is about reliability andavailability and security,
extensibility.
Many of these itties in theenterprise is what you get paid
for.
So I think being very good withinfrastructure is what I mean.
(14:20):
Think of Freshworks, for example.
I think why did they not goabove the lower end of the
mid-market?
It's a lot of SMB and, at most,the lower end of the mid-market
Because they also built lotsand lots and lots of apps, lots
and lots and lots of products,but they didn't have a shared
infrastructure.
(14:41):
There was no platformunderneath and the platform was
not supposed to be scalable.
I mean.
When I say scalable, I don'tmean millions of users, I mean
extensibility by an enterprise.
They could say look, I knowit's your platform, but I can go
and extend it, customize theheck out of it and so on.
So those are the kind of thingsthat I think the founders not
(15:06):
the people that you hireafterwards only it's the
founders who need to actuallyhave the appreciation for what
it means to build a platformcompany.
A lot of enterprise software isplatform.
Speaker 3 (15:20):
Fantastic.
And let's say you were tofigure that out, or you have a
co-founding team that iscross-border.
How would you do the earlycustomer validation in terms of,
okay, is there really a need?
I'm like a SaaS company orperhaps a consumer company or
whatever, where it's a littlebit easier to do so how would
you figure out, like, okay, isthis something that people are
(15:40):
going to be willing to pay for,or what is it going to take to
get to that level?
And perhaps also, I was verycurious to hear about this Maya
principle that I read about youwhile I was researching this
podcast.
Speaker 1 (15:52):
Yeah, I think on the first question I mean, at
Nutanix, the first 10, 15, 20probes we did, they said, don't
do it Because we were talking tothe wrong audience.
And that wrong audience wasmaybe enterprise and higher end
of the enterprise, where thingsare pretty calcified.
I mean we were trying to blurthe lines between teams.
(16:14):
We said, look, we don't needall these specialized teams.
Now obviously, lo and behold,what AWS is doing is very
similar, except that they hadtaken it out to a new location
which was about streaminginfrastructure and renting
infrastructure.
We were doing it on the samelocation, which is on-prem.
So it probably was harder forus, because sometimes when you
(16:38):
change locations, it's easier,because now you get the
self-selected people who arelike, yeah, I want to rent
infrastructure, I want to bypassall the people who build
infrastructure and go tosomebody who's willing to be a
vendor where I can stream stufffrom you know and swipe a credit
card, as opposed to wait fornine months to procure and plan
and build and rack and stack andmount something.
(16:59):
So, coming to early customervalidation, I think it had to be
brute force, like, look, welove distributed systems, we
love this data problem and theonly way to build a company in
this is to pick the rightapplication on top, which
happened to be an operatingsystem like VMware, and just go
(17:20):
deep with it.
And I think the first four yearswas not first five years, I
would say, were not easy on theproduct.
You know it was.
We had like three near-deathexperiences, you know.
So what I tell people and Itell myself and everybody within
the company is that once you'rein a tightrope and crossing a
valley on a tightrope, you can'tturn back.
(17:40):
There's no turning back.
At best you can adjust and youcan't lean too much to one or
the other Because physicallythink about it the imagery of
turning back on a tightrope isalmost impossible.
You will fall and a lot ofpeople actually think of turning
back.
Now we did have to actually do alot of micro pivots along the
way, a lot of micro pivots, uh.
(18:02):
But you know you also get paidfor seeing through things, maybe
two, three quarters in advance,so it doesn't look like it's a
hard pivot.
Uh, I mean, even at devrev, Ithink you know, initially we
thought we just start replacinga lot of apps, because what
we've really built is what themarket is coming towards.
I mean satya just talked aboutthis three, four weeks ago that
(18:24):
a lot of app boundaries willblur, you don't need all these
apps.
And uh, I think we were thefirst ones almost four years ago
said look, the app boundariesmust blur because we built all
these apps with thesedepartmental boundaries and
extremely calcified boundary tothat, and departments are
created to organize humans.
Speaker 3 (18:47):
But agents don't care yeah.
Speaker 1 (18:49):
Yeah, ai doesn't care about departmental boundaries.
If anything, the more you giveit, the more we call it a
knowledge graph.
The more connectedinterconnected the knowledge
graph is, the more it can reasonwith folks you know.
So I think you need to reallyhave that level of conviction,
(19:10):
not get flailed and, like youknow, start to dilly-dally on
the initial thing, but at thesame time you need to be humble
enough to know that you need totake two-degree turns sooner
than what a hard pivot wouldactually be.
I mean, so we started with oh,we can rip out Jira, we can rip
out Zendesk, all this stuff, andwe can rip out Service Cloud.
(19:33):
Now in the mid-market, we'redoing a lot of that.
We're ripping out a lot ofthese things because you do need
to start from, because youcan't retrofit AI into Zendesk
and Jira and you know, andIntercom and things like that.
But in the large enterprise wesaid let's go and coexist and
(19:54):
that was starting out with whatwe call AirDrop.
Airdrop is our data integrationplatform.
We basically do two-way syncswith all legacy systems and it's
a hard problem to do two-waysyncs, a really hard problem.
But it's also unlocked thisthing about how our agents are
not at the mercy of salesforceapis in real time, at runtime or
(20:14):
service now apis or atlassianjira apis, because we just use
our own data platform foreverything and then in the
background we sync it back towhere it needs to go and so on.
It also reduces a lot of SaaSlicenses we don't need.
Imagine for enterprise search,for example.
How do you even make searchwork if every link that the
(20:35):
search engine actually unearthsyou have to click on it and now
you need to log on to anothersystem.
Now you're proliferating SaaSlicenses, more so than shrinking
and consolidating SaaS licenses.
So I think at the core of theearly paths and early journey is
about having that level ofconviction.
You know just the fierceresolve and, at the same time,
(20:59):
humility to know that you willhave to keep micropivoting.
The use case is very important.
I mean, one of the things thatwe did well at Nutanix was
really picking the right usecase, which was virtual desktops
.
And here we said look, let'sstart with support, because at
(21:20):
the end of the day, you know wecome in when there's enough
complexity.
Otherwise people can keep usingNotion and Slack and just be
fine building a company, butthen it becomes really hard to
change that culture over time,to really be AI native, for
example, and to think of okay,can we stop disturbing people on
Slack if we can ask an LLM ofwhat happened and who works and
(21:44):
what and all sorts of enterprisequestions can get asked there.
So we come in when people havepaying customers and then be
like, hey, you can actually geta lot of this support stuff.
But even for those who arestarting out early, I think
they're like what does it meanfor you to really get the best
(22:05):
software development experienceand product management thinking
before you get to having paidcustomers?
I think at the core of PMF, Itell people that there's no end
to the journey of PMF.
You have a PMF problem at 1million, at 10 million, at 50,
at 100, at 250, at a billion.
Because if you've not startedto think about when you're at
(22:29):
100 million to say, okay, howwill I even get to 250 million?
And it can just be addingcapacity to the company.
You have to add capability andcapacity and really think about
capabilities as well as capacitytogether and they're orthogonal
things.
I mean capability is aboutpartnerships, but capability is
also about features and productsand multi-product thinking and
(22:53):
things like that, while capacityis about adding more
salespeople or more channels ormore regions and things of that
nature.
So there's no end to thejourney of PMF, you know, and as
long as people realize that,they'll probably be in good
stead.
Speaker 3 (23:09):
No, I love this notion of capability versus
capacity.
I think most people tend tothink more linearly in terms of
capacity.
Like you said, right, moresalespeople, more engineers,
next version of the roadmap.
But you always have to be kindof ready and open.
Like you said, right, moresales people, more engineers,
next version the roadmap.
But you always have to be kindof ready and open and sensing
what the opportunity is right.
So, yeah, so I know you spokeabout maya.
Speaker 1 (23:29):
You talked about maya a little bit and I just want to
basically, uh, you know, whene-tronics was building apple,
the iphone was getting a lot oftraction as well.
So a lot of my design sort ofthinking and what it means to
actually make things simple andI'm still on the journey, I'm
(23:50):
probably 10% of the way on whatit means to really make things
simpler and I struggle, you know.
I just every day I think abouthave you made it simple enough
and simpler, and so on.
Uh, so there's a really gooddesigner, um uh, who actually
created some great brands, uh,in the U S, like uh Greyhound
(24:13):
and you know early Coke brandand all this stuff.
His name is Raymond Louie.
Uh, l O E WW-Y and he talkedabout this concept of most
advanced yet acceptable.
So how do you really cross thechasm with this M-A-Y-A
principle of most advanced yetacceptable?
And that's what a lot ofstartups actually have to go
(24:36):
through, because once they gothrough the innovators, there's
a chasm to cross to get to evenearly adopters and early
majority.
I mean, the innovators willlike the most advanced stuff
because they don't have theproblem of legacy or brownfield.
But the moment you go to peoplewho have some money and some
money, more money and even moremoney.
You know you have to thinkabout how to take the past into
(24:59):
the future.
You, you, can't ignore the past, and that's what Maya is about.
It's about being most advanced,yet acceptable.
Speaker 3 (25:09):
So, anyway, we started talking about DevRev
already, but I wanted to talkabout the transition for you
both from getting Nutanix to anIPO, to a DecaCorn and more
maybe then taking a step back.
So how was that transition likeat a personal level and then at
an intellectual level to startyet another company, which is
(25:31):
going to take a lot of time andenergy, and of course, I know
you're excited about it.
But maybe just a little bitabout that transition.
Speaker 1 (25:37):
Yeah, yeah, you know, I think, uh and I tell this to
myself, but I think it waspretty evident that, had it not
been for the public cloud, we'dprobably be bigger than vmware,
because we had gone through thetransition of subscription as a
public company, uh.
But then, you know, thebusiness model changed.
(25:57):
Uh, people wanted to actuallyrent more infrastructure and
stream more infrastructure, butthere was a pretty good path.
We had to actually be a $100billion business because we were
doing data, that VMware wasstruggling with data and, if
anything, the people whoactually own VMware wanted to
continue to do proprietaryhardware the old legacy world of
(26:19):
EMC and all that.
But then things changed.
Public cloud happened in around2016, when I was going for the
IPO non-deal roadshow, and eventhe IPO roadshow, there was a
lot of people who said, but whatabout that?
And I think it had started toreally hit my sort of stream of
(26:42):
consciousness that we need tofigure out how to change the
business model of this company,otherwise we won't survive.
And we started doing thatstarting in 2018 onwards, and it
took us a couple of years toeven get to the basics.
But then in 2020, I'm like,okay, I can continue to be
defensive about the public cloudand be ignorant about it, but
(27:04):
the developer in me would notstart a new company on-prem.
There was this realization thatmore and more things, people
want to actually stream and justbe able to use lightweight
stuff which was hosted in AWS orover time in GCP.
So I'm like now's the time toreally decouple my left brain
(27:25):
from the right brain.
There is an investor in me andthen there is a creator, slash
operator in me, and I need tolook at them as two different
things.
And it's the hardest thing forfounders to say, okay, you know
what, right now I'm an investorand an operator in one company.
What did it mean to be aninvestor in one and go and
create slash operate the other?
(27:45):
You know, and in many ways, youpick from the left and put it
to the right.
You know.
So the things that you do interms of, uh, diversifying and
all that.
I'm like I have another 20 years.
I turned 45 in 2020.
I'm like I have another 20years to give to the industry at
least.
And what are the problems thatI'd like to work on?
So Vinod and I were talking andhe talked about GPT.
(28:08):
Vinod was also the firstinvestor in OpenAI and also our
second investor, biginstitutional investor at
Nutanix in 2011.
And I'm like, wow, I need toreally refresh myself.
I used to be very good at math,but that was 20 years before
that, so I started to reallyread on it and I'm like, okay,
(28:30):
I'm passionate about businesssoftware, customer support.
We've done a really good job ofbusiness software at Nutanix,
which is the reason why ITdidn't own business software at
Nutanix.
We kept it in a separate teamand that's how we kept
transitioning our business model, because it was an engineering
(28:50):
problem, not an IT problem.
We said we've got to reallyhave a ton of developers
actually go and do things aroundeven things like configure,
price quote and things of thatnature that as we kept changing
their business model, we need tokeep changing, you know.
So I figured you know we needto start thinking about really
bringing a lot of these silostogether.
(29:12):
And uh, that's how the idea ofDevra really came about that,
look, we have tons of silos, youknow, not just in customer
support, but but productmanagement, software development
I mean, even sales is so siloedfrom the rest of product and
support and engineering.
So the idea of really bringingit all together, even though it
(29:32):
was not as AI native of thinking, but now looks like the AI
market is only bringing ittowards us, you know.
So it's been a good serendipity, you know.
Speaker 3 (29:45):
Absolutely.
I love what you said about thefact that departments and stuff
are organized more for humanbeings and organizing labor, not
software, let alone AI or workand departments do work with
each other.
So what is the greater visionfor DevRev and how are you doing
it this time around?
(30:06):
What is new between how youbuilt, perhaps, Nutanix I know
different company, different eraand what you're doing now
perhaps?
Speaker 1 (30:14):
One of the things was to really do this also for very
small companies.
We said what does it mean ifpeople only have GitHub and only
have Google Workspace, thenwhat else do they need to really
complete their business?
(30:35):
Of course they'll have HRsystems and payroll and all this
other stuff, but how do theycomplete their business and
really also start to be like aproduct manager?
Because a lot of founders, Imean there's no MBA for product
management.
If you realize and it's one ofthe core things in creating
anything is to really thinkabout what truly matters and how
(30:59):
do you really gain empathy forthe user, the end user, which is
so important.
So I think the core of DevRevwas customer and product are the
two entities that mostbusinesses have to understand
and most businesses struggle.
I mean, they all worship theirwork.
You know which is tickets andissues and incidents and
(31:21):
opportunities.
This is all work management.
You know, for every departmentthere's a work management tool.
Sometimes they call crm,sometimes they call software
development tools, yet othertimes they call support software
, but at the end of the day it'score work without really
understanding is it?
Do we understand the customerand do we understand the product
?
So we said, we're going tobuild a knowledge graph that are
(31:43):
rooted in these two things thecustomer and the product, and
everybody in the company needsto understand customers and
products.
So, rather than now beingforced to get a CRM license to
know about the customer, what ifyou brought customer in into
the back office, the mid office,and same thing with product,
rather than keep product asprojects within the back office?
(32:07):
How do you take product intothe front office so that they
are not just process people andaccount management people, but
they're also really knowingwhat's coming out, what's high
quality, what's low quality,what's usage, what's engagement?
How do we provide feedback?
So, really, customer andproduct became the two core
pillars of this knowledge graph.
And then we said look, then youget users, sessions, people work
(32:31):
, there's a lot of enterpriseactivity that you need to
capture, but also unstructureddata.
So now you have a knowledgegraph.
How do you even use a knowledgegraph?
Like well, you need to at leastthink about search, which just
never happened in the world ofSaaS Analytics, because they
punted it to IT say, hey, itsolves analytics for you and
(32:52):
brings all the data togetheracross different departments and
different software tools, andthen workflows like well, again,
that's punted to.
Either in the SMB market it'sZapier, workato, pipe, dream or
higher in the market, you needto go to service now because
(33:12):
there's nobody who doesworkflows better than some of
these guys.
So everything was punted SaaScompanies never solved for
search, never solved foranalytics, never solved for
workflows.
So we said any modern SaaS thatwe build has to have these
three big pillars.
But then how do you build theseengines without apps?
We're like well, if you don'tbuild our own apps, then it's
like, you know, there's noWindows without Excel,
(33:34):
powerpoint or Word.
There's no iOS without music,email, phone and some of the
native apps like Maps and so on.
So we said we're going to buildthree very good apps and then
also a chatbot.
You know we call that an agentwhich actually sits on the
customer's side.
So we built these three appsand a chatbot.
(33:56):
One was a support app, one wasa build app.
So support and build arecousins of each other.
They know about what'shappening in the customer side.
How do you build software?
How do you prioritize stuff?
And then a grow app, which isreally about CRM.
So we have these three apps,but they're on one platform, so
there's no struggle between likehey, are these silos again?
And so on.
(34:17):
And then the way we actually gosell is through solutions.
And so we have a solution forthe mid-market which is around
going and replacing Zendesk andService Cloud.
I mean, in fact, some of thelarge commerce companies in
India.
They've replaced likemillion-dollar displacements of
Jira because they want somethingwhich is AI-native.
(34:39):
But for the smallest of thesmall companies they're like
here's the cluster.
If you have a Google workspaceand if you have a GitHub, all
you need is DevRel.
If you have a Google workspaceand if you have a GitHub, all
you need is DevRel.
And you start with anintegrated company rather than a
completely siloed company.
And then for the high end of theenterprise, we are going with
(34:59):
enterprise search.
Let's go and solve, search forthem and be very differentiated.
So it's consumption-basedpricing.
You don't have to pay forshelfware that people are not
using.
We also reduced the number ofSaaS licenses so in search you
don't need to actually have asmany Salesforce and ServiceNow
and Jira and Zendesk licenses.
(35:21):
And these are the basic threesolutions we're doing.
So enterprise search, going andreplacing and modernizing
customer support.
And then for the startup, likea cluster of all three DevRev
apps plus the chatbot comestogether in one.
Speaker 3 (35:38):
No, very, very, very exciting, innumerable questions.
Maybe one that I will ask.
This, being representative ofstartups, is how do startups
work with you guys?
Right, either as partners orbuilding on your platform or
whatever One?
Is you selling to other clientsand customers, and whether
you're placing Jira or Zendeskor what have you, but are there
(36:00):
ways in which startups can workwith you guys?
Speaker 1 (36:02):
Yeah, I mean, you know we have a again going back
to the idea of extensibility andcustomization, we built a
marketplace very early on in thejourney of this company and the
idea of marketplace was that,just like Windows, you remember,
most things in Windows weredone in the user space as an
application, and this is 30years ago.
(36:24):
The architecture wasmicrokernel-based.
They got a lot of people fromDEC, and DEC people were
microkernel people and they hadcome up with Mark, the Mark
kernel, and I think, if anything, linux copied a lot from that
as well from Windows.
But the idea that has keptgoing on ever since is that you
don't shove everything insidethe platform, you put things
(36:46):
above the platform through amarketplace.
People call it app store overtime.
So we have a really, reallyflourishing marketplace
architecture.
People go build all sorts ofconnectors we call you know
AirDrop is actually a way to doconnectors for data.
Then we have workflows and wehave analytics.
(37:08):
So basically it's a great placefor startups to actually go
build all these plugins with us.
We call them snap-ins and Ithink at some level, using the
product itself will give them aton of ideas.
We have a freemium model.
Gaurav is passionate aboutfreemium, and we've done a very
good job of saying don't worry,if you've not raised meaningful
(37:32):
dollars, we'll actually do thiswith you, and that was the other
thing that we did verydifferently this time that look
the long tail of companies thatare still not raised enough
money.
You know how do we help them.
You know how do we build acommunity around them.
You know, and how do we stillgive them support rather than
leave them at the mercy of just,you know, being on their own.
(37:56):
So we've done a lot to reallybuild this PLG muscle and Gaurav
, having spent 12 years at AWS,he's bringing a lot of that PLG
muscle to us.
So I think at the core, we alsowant to probably share notes
with startups on how we'rethinking about AI, because AI is
(38:16):
actually quite a spectrum.
The more I dig into it, amit,the more I realize that it's
actually quite a spectrum.
And one end of the spectrum ispeople who just think that
prompt engineering andfoundation models is done, it's
a done deal, that's all you needto do.
But then quickly you realizethat prompts get to become too
big and it starts to confuse theheck out of foundation models
(38:38):
because they don't understandattention with such a context, a
large context, and then they'relike hey, we need to do RAG.
So now you need to do RAG,which is semantic search, and
now you need to understandembeddings and vectors and
vector databases and that's thekind of thing in the middle.
But then RAG for everything isa harder problem.
Like, you need to do RAG fornot just documents but RAG for
(39:03):
every workflow asset, which isevery automation you build.
You want to do search on thosethings.
Similarly, you want to do RAGfor analytics widgets.
There's so many widgets thatpeople build in the enterprise
that need to be searchable.
So you went from promptengineering to RAG, to RAG for
everything.
And then you realize that peopleare looking for reasoning.
(39:24):
Like, hey, I want to reason,because now you're going from a
lot of peripheral agentic work,which is what happened in
customer support, front office,l1, l2 support.
Now you're getting to L3, l4.
And these are getting closer tothe mid office and then back
office.
There you need to havereasoning.
(39:44):
You need to spring signals fromfive different sources.
Look at the history of the lastthree years of how we did things
.
So then you need to do signalsfrom five different sources.
Look at the history of the lastthree years of how we did
things.
So then you need to dosupervised fine-tuning, and I
think one of the things I justheard recently, controversially,
when Nandan was saying thatIndia nobody needs to learn how
to fine-tune a model and I sawArvind talking about this from
(40:06):
perplexity that he's so wrongand I agree.
I think, going back to yourquestion on enterprise software,
and why would India not producesuch companies in the future,
it's because we're not going tobe deep enough in AI, the new
systems and systems engineeringand building infrastructure is
(40:27):
to go beyond prompt engineeringto not just RAG, also to
supervise fine tuning, smallmodels running on the edge
hosted by you and Kubernetes,maybe bedrock, but I think
understanding that spectrum issomething that we'd love to
share with startups and havethem really build deeper
(40:47):
businesses, not just, you know,gpt wrappers, which is what a
lot of prompt engineering is,but to build deeper businesses.
Speaker 3 (40:56):
No, very, very, very exciting, very, very exciting
thoughts.
Just one more thing about justmore broadly, beyond Evrev how
do you think about the SaaScompanies of yore and by yore I
mean the last five, seven years,I don't mean 10, 15 years ago?
How are you seeing theirevolution into the AI world?
(41:19):
Because obviously now they'rethe incumbents and they don't
have the luxury of starting AIfirst.
They have to adapt to AI.
So how do you think about anexisting SaaS company founder
who's already three, four, fiveyears in, has products, has
customers and has maybe two,three, five, 10, 20 million ARR
for them to adapt to the AIworld?
Speaker 1 (41:38):
So if you look like 10 years back, a lot of the
modern PLG SaaS companies I meansome of them got acquired, like
Slack got acquired, figma,almost so.
Notion is trying to reallyfigure out its place in the
world of AI.
Canva has been another PLGdarling, but they will need to.
(42:06):
I mean, maybe they need to findmore time to embrace AI,
because the DNA of thosecompanies was probably more
design and they haven't figuredout AI because the DNA at the
very top is not beyond designand we look at AI and design as
kind of two sides of the samecoin.
In fact, ai and UI are kind ofthe yin and yang that you need.
(42:32):
So I think that a lot of themodern companies that are
building PLG era, they will findit hard to make any more AI
than what ChatGPT alreadyprovides you.
If you look at verbs likesummarize or generate, based on
some points and things like that.
(42:53):
Now, maybe the win for them isthat they keep it integrated and
maybe the simplest and mostpowerful thing that they can do
is like, hey, you don't need tobuy a license for ChatGPT if I
can provide that to you andmaybe that's good enough.
You don't need to do anythingmore than that, that they become
chat GPT wrappers for the lowend of the market and maybe low
(43:15):
end of the mid market orsomething.
But I think the last five,seven years, most companies that
have been formed they didn'thave enough of, I would say,
systems chops, infrastructurechops, because AI is now an
infrastructure problem.
You know, anything that couldhave been GPT-wrapped has been,
will probably get done and maybethose people will bring a lot
(43:35):
of go-to-market skills on theother side and maybe that's one
formula that I'm seeing succeed.
It's okay, I'll actually pick adepartment or a function in a
company and just go deep intobuilding AI for them and that
probably is a good I would say$100 million to $100 million
(43:56):
story in terms of revenue, andthen you just be very good at
inside sales and thego-to-market machine and
probably building some reallygood workflows.
But I think where it starts tobecome a problem, because if you
can't make a platform, youcan't make half a billion, a
billion dollars in revenueperiod, because to go upmarket
(44:17):
if you only need to make abillion dollars, and how do you
go upmarket if you don't have anextensible platform where the
partners, the SIs, theAccentures, the Infosys and even
the customers.
Developers can actually buildstuff on top of, and the most
important stuff they'll build istwo things workflows and
analytics widgets.
So I think the company's lastfive, seven years either they'll
(44:41):
be very good like Gong was.
Gong was very good in 2020,2021 era and all like oh, oh
fast to 300 million.
And the question is now what?
You know, uh, basically therewas so much churn, uh, because
there's so much zarp money, thatuh went away with interest
rates being high, that the smbsuffered.
So I feel like going to the midmarket in enterprise is
(45:04):
probably the real challenge andopportunity for a lot of
companies right now, because theinterest rates are where they
are, whether we like it or not.
Until the wars end, untilUkraine normalizes, I think it's
really hard.
I mean Europe has to actuallyget back to normal.
I mean Russia has to start toprovide more things to Europe.
(45:24):
Those things start to happen.
Then interest rates come down.
Until then, there's a reallyhard thing, especially
non-digital startup companies.
They're not coming back anytimesoon.
Startups in software areprobably one good thing.
That is still okay in the SMB,but everybody else has to really
think about the mid-market andthe enterprise.
(45:45):
Rajat.
Speaker 3 (45:47):
Mittal Fascinating, dheeraj.
So, as we bring this to a wrap,you mentioned one very
interesting thing about PMF at 1million, 100 million, 250
million, a billion I'msimplifying a little bit, but
there's also I'm just going tomake up a new term like
organization or founder marketfit at each stage.
(46:08):
So how do you keep evolving asthe company is scaling, and
whether it's you or yourco-founding team or your
leadership team and any thoughts, and obviously, since our
audience is a lot more earlystage entrepreneurs even, let's
say, do the exercise at 1, 10,50, 100, you know, like at the
early stage of that ladder, andwhat are some kind of tips,
suggestions, maybe even amber oryellow flags to watch out for
(46:32):
so that you know you're yourselfscaling with the company?
Speaker 1 (46:36):
Yeah, I think at the core, it's about how do you
really get to relate with peopleover time, and this includes
not just investors and yourboard members, but also your
executives.
You hire some executives whoare very good at what they do,
and the only way you can retainthem is if you know that they're
good at things.
Then you are, and buildingpeople relationships is probably
(47:02):
another way to really build agreat mid-market enterprise
business too, because people buyfrom people in mid-market and
enterprise.
I think process is anotherpiece that people have to start
to respect, because processmatters, efficiency matters,
sustainable stuff matters.
Growth hacks don't work whenyou're starting to really grow
(47:23):
bigger.
So a lot of founders who areproduct people, how they really
begin to embrace people andprocess is probably the way to
do this, cause then more peoplewant to come and work for you,
work with you.
Uh, and in this, in this wholejourney, you need to really know
what it means to let go ofthings that were very near and
(47:43):
dear to you and also be verygood at negotiation, because
your word and your sort of whatused to be the diktat is not
going to be that you know sortof impactful anymore, because
then you're only hiring doers,as opposed to people who are not
(48:05):
just doers but thinkers.
I think how you continue tohire thinkers and of course not
every thinker will align withyou.
They'll say, let's do this andlet's do that and this is not
going to work.
So it's such a fine balancebetween the things that brought
you here and what percentage ofthat you continue to keep and
(48:25):
what percentage you continue togo and evolve is basically at
the core of this.
So learning to negotiate,learning to actually really
build those bridges with peopleyou're hiring and with customers
, respecting people and process,is the only way you can really
scale a company.
I mean, because not everybodycan be Elon Musk or Bill Gates,
(48:47):
you know.
I mean even Jeff Bezos.
I mean I'm a big fan of Jeffand you know people like Steve
Jobs.
You know they said, look, andJeff himself had gotten so good
at framework thinking.
You know he probably was verygood at it from day one.
But this idea that, look, I needto bring framework so that now
(49:08):
I can leave that behind in ameeting room so I don't need to
be in every meeting.
So frameworks also become agood way to scale the company,
because then process thinkingcomes, scalable thinking comes.
But someone like Steve realizedthat if he can't deal with
people, you need to get a TimCook and you need to complement
(49:28):
yourself with that kind of aperson who deal with people and
process.
There was a lot of process inApple dealing with China and
manufacturing and Foxconn andall this stuff, and then knowing
that the people within Applehad to respect that process to
go and sit in China for sixmonths every year if that's what
it took to really build Applein.
(49:49):
So really at the core is peoplein process.
Speaker 3 (49:53):
Yeah, I think people process and delegation.
That's a wonderful place to end, and I remember at Google,
there was Eric Schmidt, therewas Sheryl Sandberg, there was
Nikesh Arora.
There was a whole bunch ofpeople that were brought in
because the founders had thecreative mojo, but they were
like all this other stuff I need.
We need people to do this right.
And Dheeraj fascinatingconversation.
(50:14):
You could go on and on, but Iwant to be respectful of your
time.
Thank you so much for being onthe Prime Venture Partners
podcast.
Speaker 1 (50:22):
Yeah, it's a pleasure myself.
Thank you again and hope youraudience finds it's meaningful.
Speaker 3 (50:27):
Thanks, Dheeraj.
Speaker 2 (50:31):
Dear listeners, thank you for listening to this
episode of the podcast.
Subscribe now on your favoritepodcast app for free and you'll
be the first one to know whennew episodes are available.
Just search for Prime VenturePartners Podcast in Apple
Podcast, spotify, castbox orhowever.
You get your podcasts, then hitsubscribe and if you have
(50:52):
enjoyed the show, we would bereally grateful if you leave us
a review on Apple podcast.
To read the full transcript,find the link in the show notes.
Once you're on a tightrope and crossing a valley,
you can't turn back.
Probably one of the rare peoplein the history of IoT is to
have quit IoT after two and ahalf months.
I was a risk-taking person fromearly on, so we picked a pretty
contrarian use case and we wentagainst the mob.
In 2011, people had started tosay that Windows is dead.
We had three near-deathexperiences.
(00:21):
You know AI doesn't care aboutdepartmental boundaries.
If you have a Google workspaceand if you have a GitHub, all
you need is Debra.
Enterprise software at the coreis about reliability and
availability and security.
Extensibility Many of theseitties in the enterprise is what
you get paid for.
Speaker 3 (00:52):
Welcome to the Prime Venture Partners podcast.
I am delighted to have with ustoday Dheeraj Pandey.
He is founder and CEO of DevRev, an exciting new AI startup,
and then before that, of course,he's very famously known for
co-founding Nutanix, which isnow much more than a DecaCon, I
guess.
Welcome to the show, dheeraj.
Speaker 1 (01:12):
Thank you, Amit.
Speaker 3 (01:14):
Dheeraj, love to hear a little bit about your journey
of sort of growing up in Indiaand what led you to kind of
eventually the road ofentrepreneurship.
Speaker 1 (01:22):
What led you to kind of, eventually, the road of
entrepreneurship.
Yeah, in fact, you know, I wasin Patna, which is, you know, a
capital city, one of the easternstates, bihar, as many of you
might know, and I was there tillI was 16, 17.
(01:44):
And I went to a couple ofschools in Patna One was Don
Bosco and the other one was StMichael's and I just, you know,
tried to be in top three, topfour kind of class.
And I used to be there, I usedto enjoy math, and mostly math,
(02:07):
I would say, more than anythingelse.
And then, you know, I took theJ in 92 for the first time and I
hadn't really prepared at allfor it.
This was the year that Igraduated from high school and I
got a rank of 1,420.
And I really wanted to come toIIT, kanpur.
(02:32):
So I'm like, okay, whateverbranch I get there, I'll take it
.
I didn't want to go to anyother IIT.
So I went there and I got civilengineering and within a couple
of months I realized that theprobability of a brand change to
computer science is way lowerthan if I were to take the JEE
again and were to be in top 100and then get into computer
(02:54):
science.
So I was probably one of therare people in the history of
IITs to have quit IIT after twoand a half months of attending
classes and I came back and Iretook the exam next year and
this time I was in top 100 and Iwas ranked 84.
And I'm like, okay, this timeI'm definitely going to get
computer science, which I didback in the day Kanpur used to.
(03:16):
You know, if you had to getcomputer science, you have to be
in top 100.
And so this is 93, startedjourney by 94,.
As you remember, india was goingthrough this massive
liberalization process.
The economy was opening up andthere was something about that
(03:37):
new India that was opening up.
And Manmohan Singh as a financeminister Bless his soul, he
passed away, but he was such akind of North Star for me.
I'm like, wow, you know, blesshis soul, he passed away, but uh
, he was, uh, such a kind ofnorth star for me.
I'm like, wow, you know, oxfordeducated, economist and all
this, and, and then, uh, andhe's really transforming the
country.
Uh, so I really wanted to be aneconomist, I want to do a phd
(04:01):
in economics, uh, even though Iwas doing computer science in
undergrad.
And then I talked to my cousinwho had just come back from the
US and he explained to me verysimilar to what's happening
right now with AI.
Basically there's a bigrevolution underway with the
browser and HTML and the WorldWide Web and HTTP and everything
else and he's like you got toreally continue computer science
(04:24):
.
It's the best time to be incomputer science.
You know and really do well incomputer science.
So, lo and behold, you know Ijust kept enjoying computer
science and a lot of math aswell.
You know, did really well.
I used to love, love math.
You know, back in the day, ifnot for computer science,
probably would have done a lotof math and been a major in math
(04:44):
.
I was a good visual thinker andnow obviously I've slowed down
a lot in the last 30 years.
But in 96, I'm like OK, we havetwo forks in the road.
I could go and join a job or Icould just apply to grad school,
and that's what I did.
I ended up applying to a lot ofgrad schools.
(05:06):
I did get a couple of jobs, butI figured I'll come to the US
and really pursue a PhD incomputer science.
So I got a few offers from likeUT Austin, Urbana-Champaign,
usc, columbia, all these places,and I ended up choosing UT
Austin because of the fellowship.
So I landed in the US in 1997in Austin, texas, and you know,
(05:28):
after a couple of years I said Igot to be in the industry right
now.
So I went on leave of absenceand for a couple of years I was
on leave of absence but neverwent back to finish my PhD,
worked at Trilogy Software forabout a year, year and a half,
in Austin, texas, and thenworked for quite a few system
(05:49):
software companies building fileservers and databases at Oracle
and then distributed datawarehouse and then in 2009,.
Speaker 3 (06:01):
basically started Nutanix.
Fantastic, very, very inspiring.
And I was going to say a littlebit tongue in cheek that
leaving IIT in the first yearwith the 1400 rank to get back
in, that was like your firstexperience of a pivot, a hard
pivot.
Speaker 1 (06:14):
It was.
Indeed it was.
It was.
I think I was a risk takingperson from early on.
You know, I think and that wasone of the first big risks they
actually took and you know Ialways thought about the worst
is not bad enough.
Why wouldn't you?
And, uh, the worst situationthere was I had gotten admitted
(06:34):
to since defense of physics andlike, yeah, just gonna do
physics and we'll figure out therest from there next time,
because I've done fairly well inmy high school, you, you know,
secondary exams and stuff.
Speaker 3 (06:45):
Absolutely so.
That leads me to sort of thefact that you worked at a lot of
different companies, includingOracle, and on file system
database etc.
In the enterprise softwarearena.
What led to the idea forNutanix, number one and number
two?
Just this notion of you want tobe an entrepreneur because once
(07:05):
you're there, you're settled.
You're doing your PhD yes, youwent on a leave of absence or
sabbatical or whatever.
What made you think that youwanted to start a company back
in 2009?
Speaker 1 (07:17):
You know, there was always an itch and I think in
2000, I think 2009, february Ihad a really long discussion
with you know, a strategicdiscussion about with my
founders of the company that wewere working at.
There were three of us whoactually knew each other for
(07:39):
almost a decade or more and Ithink the core insight that
we're trying to bring was thatSQL and NoSQL are here to
coexist and we were kind ofpigeonholing ourselves into just
SQL and that to only datascientists and that to only ad
(08:00):
hoc querying and all that likeon-demand analytics, versus what
the world was doing with Hadoop, and the developers were
getting into analytics and theywere using Java and MapReduce
jobs back in the day to reallybuild a ton of the analytical
pipeline.
And we're like look, we'rebuilding a large-scale
distributed system, so it shouldactually come together.
(08:21):
We should not take sides on SQLversus NoSQL.
But it wasn't to be.
I think we couldn't convincethe it should actually come
together.
We should not take sides on SQLversus NoSQL.
But it wasn't to be.
I think we couldn't convincethe folks there that we should
actually build a larger business, a larger company that really
brought both of them together.
So I think we just kind ofspoke for six months and said,
look, we've got to do somethingon our own, and we didn't even
realize that this was the worsttime to start a company.
(08:43):
And we didn't even realize thatthis was the worst time to
start a company Because the firewithin it reminds me of the
fire within was calledAntaragini.
For us in our IT days, thecultural festival was called
Antaragini, but basically it wasthat Antaragini.
We said, look, what's the worstthat can happen?
You go and find a job andyou'll find a job in the Bay
(09:06):
Area.
It's the cornucopia in somesense, you know.
So data was kind of at the coreof what I had built in.
Both my co-founders, mohit andAjit had you know, especially
Mohit had actually been in thisspace for like 10 years and he
and I bumped into each other atZambil, which was my file server
(09:27):
company that we were alltogether at, and so data was
going to be the thing,distributed systems was going to
be the thing.
We're going to fight, not onthe hardware vendor's turf but
on our turf, and that turf wouldbe pure software running on
commodity hardware, becausethat's all I had done myself,
you know, starting with Zambiawas file servers and commodity
(09:49):
hardware.
You know Oracle had moved tocommodity hardware with
databases, oracle RAC, exadata,and then Astrodata was also
distributed data warehouses andcommodity hardware.
So the idea was to do somethingwith distributed systems, pure
software.
And so what's the killerapplication right now?
And the killer app back thenwas VMware.
(10:10):
Like, how do you run VMwareworkloads on a distributed
architecture?
None of us knew even how tospell virtualization, like even
the V of virtualization, so wehad to learn virtualization like
to the core and then go buildsystems on top of it and make
(10:30):
sure that we got the first usecase right.
And in 2011, people had startedto say that Windows is dead
because Apple is everywhere.
Who's going to care for Windows?
And we took a pretty contrarianview that look, long live
Windows.
You know, long live Windows.
And the way it would happen isthrough a digital Windows system
(10:52):
, like you'll have to streamWindows from somewhere you know,
from a data center, and Windowswill then come together as
clusters of systems that peopleare actually streaming, because
in business world, people willstill have a ton of Windows
because of Office and everybodywas like what are you talking
about?
This is never going to work,windows is dead.
So we picked a prettycontrarian use case and we went
(11:16):
against the mob.
The mob was saying that there'sno future for Windows, if
anything.
Three years later people startedsaying things about SQL.
The SQL is dead and here we arein our long-lived SQL.
So basically we built the firstuse case around that and we
went pretty deep to US FederalVery early on.
We went to US Federal.
So there was a kind of aduality there, like just being
(11:41):
really good with both contrarianviews.
There was a contrarian viewabout going early to federal
people like okay, no, startupsgo to us federal as one of the
early segments, and it was allabout great people.
We hired some really goodpeople who were miffed with
vmware.
They were vmware uh sort offolks and they came and built
something and, you know, westarted out being, you know, a
(12:05):
really good partner for VMware.
We used to run on top of VMwareand obviously over time we
started to get a lot of theirtalent and things became pretty
testy.
They became a frenemy, thenthey became an enemy, you know,
and then we just had to buildfor all that.
Speaker 3 (12:24):
Amazing, amazing journey.
A bit of a question for youknow entrepreneurs in India who
are trying to build.
Now, of course, we're going totalk a lot about you know AI and
SaaS and all that, but likegoing back five, seven, 10 years
ago.
Why is it that a lot ofenterprise software companies
necessarily haven't come out ofIndia?
And I know there is a few rightbut, and what does it take in
(12:47):
the early days, in the zero toone journey for an enterprise
software company?
There's, of course, the techstack you have to build, where
you probably have moreconviction as engineers or
whatever.
We have good quality engineers.
But then all there's also thecustomer validation, figuring
out the initial kind ofco-creators or the co-partners
to build with, and so on.
So maybe just a little bit ofbuilding a large scale, whether
(13:08):
it's a now SaaS company, AIcompany or enterprise software
company, in the zero to onephase.
Speaker 1 (13:12):
If you have any thoughts and suggestions for our
listeners, yeah, I think Iwould say that founders many of
the founders in India are notthe folks who actually have done
infrastructure work and a lotof business software, but also
enterprise software at large isa lot of infrastructure work.
(13:34):
Like you know what we're doingat DevRev now there's so much of
infrastructure work.
Devrev now there's so much ofinfrastructure work.
It's like you have to pay thedues for the object model, the
event model, the security model,the SQL model, all that stuff I
mean, and over time, obviouslythe AI model.
All this stuff isinfrastructure.
(13:55):
It's systems engineering andthe founders need to respect
systems engineering, becauseenterprise software at the core
is about reliability andavailability and security,
extensibility.
Many of these itties in theenterprise is what you get paid
for.
So I think being very good withinfrastructure is what I mean.
(14:20):
Think of Freshworks, for example.
I think why did they not goabove the lower end of the
mid-market?
It's a lot of SMB and, at most,the lower end of the mid-market
Because they also built lotsand lots and lots of apps, lots
and lots and lots of products,but they didn't have a shared
infrastructure.
(14:41):
There was no platformunderneath and the platform was
not supposed to be scalable.
I mean.
When I say scalable, I don'tmean millions of users, I mean
extensibility by an enterprise.
They could say look, I knowit's your platform, but I can go
and extend it, customize theheck out of it and so on.
So those are the kind of thingsthat I think the founders not
(15:06):
the people that you hireafterwards only it's the
founders who need to actuallyhave the appreciation for what
it means to build a platformcompany.
A lot of enterprise software isplatform.
Speaker 3 (15:20):
Fantastic.
And let's say you were tofigure that out, or you have a
co-founding team that iscross-border.
How would you do the earlycustomer validation in terms of,
okay, is there really a need?
I'm like a SaaS company orperhaps a consumer company or
whatever, where it's a littlebit easier to do so how would
you figure out, like, okay, isthis something that people are
(15:40):
going to be willing to pay for,or what is it going to take to
get to that level?
And perhaps also, I was verycurious to hear about this Maya
principle that I read about youwhile I was researching this
podcast.
Speaker 1 (15:52):
Yeah, I think on the first question I mean, at
Nutanix, the first 10, 15, 20probes we did, they said, don't
do it Because we were talking tothe wrong audience.
And that wrong audience wasmaybe enterprise and higher end
of the enterprise, where thingsare pretty calcified.
I mean we were trying to blurthe lines between teams.
(16:14):
We said, look, we don't needall these specialized teams.
Now obviously, lo and behold,what AWS is doing is very
similar, except that they hadtaken it out to a new location
which was about streaminginfrastructure and renting
infrastructure.
We were doing it on the samelocation, which is on-prem.
So it probably was harder forus, because sometimes when you
(16:38):
change locations, it's easier,because now you get the
self-selected people who arelike, yeah, I want to rent
infrastructure, I want to bypassall the people who build
infrastructure and go tosomebody who's willing to be a
vendor where I can stream stufffrom you know and swipe a credit
card, as opposed to wait fornine months to procure and plan
and build and rack and stack andmount something.
(16:59):
So, coming to early customervalidation, I think it had to be
brute force, like, look, welove distributed systems, we
love this data problem and theonly way to build a company in
this is to pick the rightapplication on top, which
happened to be an operatingsystem like VMware, and just go
(17:20):
deep with it.
And I think the first four yearswas not first five years, I
would say, were not easy on theproduct.
You know it was.
We had like three near-deathexperiences, you know.
So what I tell people and Itell myself and everybody within
the company is that once you'rein a tightrope and crossing a
valley on a tightrope, you can'tturn back.
(17:40):
There's no turning back.
At best you can adjust and youcan't lean too much to one or
the other Because physicallythink about it the imagery of
turning back on a tightrope isalmost impossible.
You will fall and a lot ofpeople actually think of turning
back.
Now we did have to actually do alot of micro pivots along the
way, a lot of micro pivots, uh.
(18:02):
But you know you also get paidfor seeing through things, maybe
two, three quarters in advance,so it doesn't look like it's a
hard pivot.
Uh, I mean, even at devrev, Ithink you know, initially we
thought we just start replacinga lot of apps, because what
we've really built is what themarket is coming towards.
I mean satya just talked aboutthis three, four weeks ago that
(18:24):
a lot of app boundaries willblur, you don't need all these
apps.
And uh, I think we were thefirst ones almost four years ago
said look, the app boundariesmust blur because we built all
these apps with thesedepartmental boundaries and
extremely calcified boundary tothat, and departments are
created to organize humans.
Speaker 3 (18:47):
But agents don't care yeah.
Speaker 1 (18:49):
Yeah, ai doesn't care about departmental boundaries.
If anything, the more you giveit, the more we call it a
knowledge graph.
The more connectedinterconnected the knowledge
graph is, the more it can reasonwith folks you know.
So I think you need to reallyhave that level of conviction,
(19:10):
not get flailed and, like youknow, start to dilly-dally on
the initial thing, but at thesame time you need to be humble
enough to know that you need totake two-degree turns sooner
than what a hard pivot wouldactually be.
I mean, so we started with oh,we can rip out Jira, we can rip
out Zendesk, all this stuff, andwe can rip out Service Cloud.
(19:33):
Now in the mid-market, we'redoing a lot of that.
We're ripping out a lot ofthese things because you do need
to start from, because youcan't retrofit AI into Zendesk
and Jira and you know, andIntercom and things like that.
But in the large enterprise wesaid let's go and coexist and
(19:54):
that was starting out with whatwe call AirDrop.
Airdrop is our data integrationplatform.
We basically do two-way syncswith all legacy systems and it's
a hard problem to do two-waysyncs, a really hard problem.
But it's also unlocked thisthing about how our agents are
not at the mercy of salesforceapis in real time, at runtime or
(20:14):
service now apis or atlassianjira apis, because we just use
our own data platform foreverything and then in the
background we sync it back towhere it needs to go and so on.
It also reduces a lot of SaaSlicenses we don't need.
Imagine for enterprise search,for example.
How do you even make searchwork if every link that the
(20:35):
search engine actually unearthsyou have to click on it and now
you need to log on to anothersystem.
Now you're proliferating SaaSlicenses, more so than shrinking
and consolidating SaaS licenses.
So I think at the core of theearly paths and early journey is
about having that level ofconviction.
You know just the fierceresolve and, at the same time,
(20:59):
humility to know that you willhave to keep micropivoting.
The use case is very important.
I mean, one of the things thatwe did well at Nutanix was
really picking the right usecase, which was virtual desktops
.
And here we said look, let'sstart with support, because at
(21:20):
the end of the day, you know wecome in when there's enough
complexity.
Otherwise people can keep usingNotion and Slack and just be
fine building a company, butthen it becomes really hard to
change that culture over time,to really be AI native, for
example, and to think of okay,can we stop disturbing people on
Slack if we can ask an LLM ofwhat happened and who works and
(21:44):
what and all sorts of enterprisequestions can get asked there.
So we come in when people havepaying customers and then be
like, hey, you can actually geta lot of this support stuff.
But even for those who arestarting out early, I think
they're like what does it meanfor you to really get the best
(22:05):
software development experienceand product management thinking
before you get to having paidcustomers?
I think at the core of PMF, Itell people that there's no end
to the journey of PMF.
You have a PMF problem at 1million, at 10 million, at 50,
at 100, at 250, at a billion.
Because if you've not startedto think about when you're at
(22:29):
100 million to say, okay, howwill I even get to 250 million?
And it can just be addingcapacity to the company.
You have to add capability andcapacity and really think about
capabilities as well as capacitytogether and they're orthogonal
things.
I mean capability is aboutpartnerships, but capability is
also about features and productsand multi-product thinking and
(22:53):
things like that, while capacityis about adding more
salespeople or more channels ormore regions and things of that
nature.
So there's no end to thejourney of PMF, you know, and as
long as people realize that,they'll probably be in good
stead.
Speaker 3 (23:09):
No, I love this notion of capability versus
capacity.
I think most people tend tothink more linearly in terms of
capacity.
Like you said, right, moresalespeople, more engineers,
next version of the roadmap.
But you always have to be kindof ready and open.
Like you said, right, moresales people, more engineers,
next version the roadmap.
But you always have to be kindof ready and open and sensing
what the opportunity is right.
So, yeah, so I know you spokeabout maya.
Speaker 1 (23:29):
You talked about maya a little bit and I just want to
basically, uh, you know, whene-tronics was building apple,
the iphone was getting a lot oftraction as well.
So a lot of my design sort ofthinking and what it means to
actually make things simple andI'm still on the journey, I'm
(23:50):
probably 10% of the way on whatit means to really make things
simpler and I struggle, you know.
I just every day I think abouthave you made it simple enough
and simpler, and so on.
Uh, so there's a really gooddesigner, um uh, who actually
created some great brands, uh,in the U S, like uh Greyhound
(24:13):
and you know early Coke brandand all this stuff.
His name is Raymond Louie.
Uh, l O E WW-Y and he talkedabout this concept of most
advanced yet acceptable.
So how do you really cross thechasm with this M-A-Y-A
principle of most advanced yetacceptable?
And that's what a lot ofstartups actually have to go
(24:36):
through, because once they gothrough the innovators, there's
a chasm to cross to get to evenearly adopters and early
majority.
I mean, the innovators willlike the most advanced stuff
because they don't have theproblem of legacy or brownfield.
But the moment you go to peoplewho have some money and some
money, more money and even moremoney.
You know you have to thinkabout how to take the past into
(24:59):
the future.
You, you, can't ignore the past, and that's what Maya is about.
It's about being most advanced,yet acceptable.
Speaker 3 (25:09):
So, anyway, we started talking about DevRev
already, but I wanted to talkabout the transition for you
both from getting Nutanix to anIPO, to a DecaCorn and more
maybe then taking a step back.
So how was that transition likeat a personal level and then at
an intellectual level to startyet another company, which is
(25:31):
going to take a lot of time andenergy, and of course, I know
you're excited about it.
But maybe just a little bitabout that transition.
Speaker 1 (25:37):
Yeah, yeah, you know, I think, uh and I tell this to
myself, but I think it waspretty evident that, had it not
been for the public cloud, we'dprobably be bigger than vmware,
because we had gone through thetransition of subscription as a
public company, uh.
But then, you know, thebusiness model changed.
(25:57):
Uh, people wanted to actuallyrent more infrastructure and
stream more infrastructure, butthere was a pretty good path.
We had to actually be a $100billion business because we were
doing data, that VMware wasstruggling with data and, if
anything, the people whoactually own VMware wanted to
continue to do proprietaryhardware the old legacy world of
(26:19):
EMC and all that.
But then things changed.
Public cloud happened in around2016, when I was going for the
IPO non-deal roadshow, and eventhe IPO roadshow, there was a
lot of people who said, but whatabout that?
And I think it had started toreally hit my sort of stream of
(26:42):
consciousness that we need tofigure out how to change the
business model of this company,otherwise we won't survive.
And we started doing thatstarting in 2018 onwards, and it
took us a couple of years toeven get to the basics.
But then in 2020, I'm like,okay, I can continue to be
defensive about the public cloudand be ignorant about it, but
(27:04):
the developer in me would notstart a new company on-prem.
There was this realization thatmore and more things, people
want to actually stream and justbe able to use lightweight
stuff which was hosted in AWS orover time in GCP.
So I'm like now's the time toreally decouple my left brain
(27:25):
from the right brain.
There is an investor in me andthen there is a creator, slash
operator in me, and I need tolook at them as two different
things.
And it's the hardest thing forfounders to say, okay, you know
what, right now I'm an investorand an operator in one company.
What did it mean to be aninvestor in one and go and
create slash operate the other?
(27:45):
You know, and in many ways, youpick from the left and put it
to the right.
You know.
So the things that you do interms of, uh, diversifying and
all that.
I'm like I have another 20 years.
I turned 45 in 2020.
I'm like I have another 20years to give to the industry at
least.
And what are the problems thatI'd like to work on?
So Vinod and I were talking andhe talked about GPT.
(28:08):
Vinod was also the firstinvestor in OpenAI and also our
second investor, biginstitutional investor at
Nutanix in 2011.
And I'm like, wow, I need toreally refresh myself.
I used to be very good at math,but that was 20 years before
that, so I started to reallyread on it and I'm like, okay,
(28:30):
I'm passionate about businesssoftware, customer support.
We've done a really good job ofbusiness software at Nutanix,
which is the reason why ITdidn't own business software at
Nutanix.
We kept it in a separate teamand that's how we kept
transitioning our business model, because it was an engineering
(28:50):
problem, not an IT problem.
We said we've got to reallyhave a ton of developers
actually go and do things aroundeven things like configure,
price quote and things of thatnature that as we kept changing
their business model, we need tokeep changing, you know.
So I figured you know we needto start thinking about really
bringing a lot of these silostogether.
(29:12):
And uh, that's how the idea ofDevra really came about that,
look, we have tons of silos, youknow, not just in customer
support, but but productmanagement, software development
I mean, even sales is so siloedfrom the rest of product and
support and engineering.
So the idea of really bringingit all together, even though it
(29:32):
was not as AI native of thinking, but now looks like the AI
market is only bringing ittowards us, you know.
So it's been a good serendipity, you know.
Speaker 3 (29:45):
Absolutely.
I love what you said about thefact that departments and stuff
are organized more for humanbeings and organizing labor, not
software, let alone AI or workand departments do work with
each other.
So what is the greater visionfor DevRev and how are you doing
it this time around?
(30:06):
What is new between how youbuilt, perhaps, Nutanix I know
different company, different eraand what you're doing now
perhaps?
Speaker 1 (30:14):
One of the things was to really do this also for very
small companies.
We said what does it mean ifpeople only have GitHub and only
have Google Workspace, thenwhat else do they need to really
complete their business?
(30:35):
Of course they'll have HRsystems and payroll and all this
other stuff, but how do theycomplete their business and
really also start to be like aproduct manager?
Because a lot of founders, Imean there's no MBA for product
management.
If you realize and it's one ofthe core things in creating
anything is to really thinkabout what truly matters and how
(30:59):
do you really gain empathy forthe user, the end user, which is
so important.
So I think the core of DevRevwas customer and product are the
two entities that mostbusinesses have to understand
and most businesses struggle.
I mean, they all worship theirwork.
You know which is tickets andissues and incidents and
(31:21):
opportunities.
This is all work management.
You know, for every departmentthere's a work management tool.
Sometimes they call crm,sometimes they call software
development tools, yet othertimes they call support software
, but at the end of the day it'score work without really
understanding is it?
Do we understand the customerand do we understand the product
?
So we said, we're going tobuild a knowledge graph that are
(31:43):
rooted in these two things thecustomer and the product, and
everybody in the company needsto understand customers and
products.
So, rather than now beingforced to get a CRM license to
know about the customer, what ifyou brought customer in into
the back office, the mid office,and same thing with product,
rather than keep product asprojects within the back office?
(32:07):
How do you take product intothe front office so that they
are not just process people andaccount management people, but
they're also really knowingwhat's coming out, what's high
quality, what's low quality,what's usage, what's engagement?
How do we provide feedback?
So, really, customer andproduct became the two core
pillars of this knowledge graph.
And then we said look, then youget users, sessions, people work
(32:31):
, there's a lot of enterpriseactivity that you need to
capture, but also unstructureddata.
So now you have a knowledgegraph.
How do you even use a knowledgegraph?
Like well, you need to at leastthink about search, which just
never happened in the world ofSaaS Analytics, because they
punted it to IT say, hey, itsolves analytics for you and
(32:52):
brings all the data togetheracross different departments and
different software tools, andthen workflows like well, again,
that's punted to.
Either in the SMB market it'sZapier, workato, pipe, dream or
higher in the market, you needto go to service now because
(33:12):
there's nobody who doesworkflows better than some of
these guys.
So everything was punted SaaScompanies never solved for
search, never solved foranalytics, never solved for
workflows.
So we said any modern SaaS thatwe build has to have these
three big pillars.
But then how do you build theseengines without apps?
We're like well, if you don'tbuild our own apps, then it's
like, you know, there's noWindows without Excel,
(33:34):
powerpoint or Word.
There's no iOS without music,email, phone and some of the
native apps like Maps and so on.
So we said we're going to buildthree very good apps and then
also a chatbot.
You know we call that an agentwhich actually sits on the
customer's side.
So we built these three appsand a chatbot.
(33:56):
One was a support app, one wasa build app.
So support and build arecousins of each other.
They know about what'shappening in the customer side.
How do you build software?
How do you prioritize stuff?
And then a grow app, which isreally about CRM.
So we have these three apps,but they're on one platform, so
there's no struggle between likehey, are these silos again?
And so on.
(34:17):
And then the way we actually gosell is through solutions.
And so we have a solution forthe mid-market which is around
going and replacing Zendesk andService Cloud.
I mean, in fact, some of thelarge commerce companies in
India.
They've replaced likemillion-dollar displacements of
Jira because they want somethingwhich is AI-native.
(34:39):
But for the smallest of thesmall companies they're like
here's the cluster.
If you have a Google workspaceand if you have a GitHub, all
you need is DevRel.
If you have a Google workspaceand if you have a GitHub, all
you need is DevRel.
And you start with anintegrated company rather than a
completely siloed company.
And then for the high end of theenterprise, we are going with
(34:59):
enterprise search.
Let's go and solve, search forthem and be very differentiated.
So it's consumption-basedpricing.
You don't have to pay forshelfware that people are not
using.
We also reduced the number ofSaaS licenses so in search you
don't need to actually have asmany Salesforce and ServiceNow
and Jira and Zendesk licenses.
(35:21):
And these are the basic threesolutions we're doing.
So enterprise search, going andreplacing and modernizing
customer support.
And then for the startup, likea cluster of all three DevRev
apps plus the chatbot comestogether in one.
Speaker 3 (35:38):
No, very, very, very exciting, innumerable questions.
Maybe one that I will ask.
This, being representative ofstartups, is how do startups
work with you guys?
Right, either as partners orbuilding on your platform or
whatever One?
Is you selling to other clientsand customers, and whether
you're placing Jira or Zendeskor what have you, but are there
(36:00):
ways in which startups can workwith you guys?
Speaker 1 (36:02):
Yeah, I mean, you know we have a again going back
to the idea of extensibility andcustomization, we built a
marketplace very early on in thejourney of this company and the
idea of marketplace was that,just like Windows, you remember,
most things in Windows weredone in the user space as an
application, and this is 30years ago.
(36:24):
The architecture wasmicrokernel-based.
They got a lot of people fromDEC, and DEC people were
microkernel people and they hadcome up with Mark, the Mark
kernel, and I think, if anything, linux copied a lot from that
as well from Windows.
But the idea that has keptgoing on ever since is that you
don't shove everything insidethe platform, you put things
(36:46):
above the platform through amarketplace.
People call it app store overtime.
So we have a really, reallyflourishing marketplace
architecture.
People go build all sorts ofconnectors we call you know
AirDrop is actually a way to doconnectors for data.
Then we have workflows and wehave analytics.
(37:08):
So basically it's a great placefor startups to actually go
build all these plugins with us.
We call them snap-ins and Ithink at some level, using the
product itself will give them aton of ideas.
We have a freemium model.
Gaurav is passionate aboutfreemium, and we've done a very
good job of saying don't worry,if you've not raised meaningful
(37:32):
dollars, we'll actually do thiswith you, and that was the other
thing that we did verydifferently this time that look
the long tail of companies thatare still not raised enough
money.
You know how do we help them.
You know how do we build acommunity around them.
You know, and how do we stillgive them support rather than
leave them at the mercy of just,you know, being on their own.
(37:56):
So we've done a lot to reallybuild this PLG muscle and Gaurav
, having spent 12 years at AWS,he's bringing a lot of that PLG
muscle to us.
So I think at the core, we alsowant to probably share notes
with startups on how we'rethinking about AI, because AI is
(38:16):
actually quite a spectrum.
The more I dig into it, amit,the more I realize that it's
actually quite a spectrum.
And one end of the spectrum ispeople who just think that
prompt engineering andfoundation models is done, it's
a done deal, that's all you needto do.
But then quickly you realizethat prompts get to become too
big and it starts to confuse theheck out of foundation models
(38:38):
because they don't understandattention with such a context, a
large context, and then they'relike hey, we need to do RAG.
So now you need to do RAG,which is semantic search, and
now you need to understandembeddings and vectors and
vector databases and that's thekind of thing in the middle.
But then RAG for everything isa harder problem.
Like, you need to do RAG fornot just documents but RAG for
(39:03):
every workflow asset, which isevery automation you build.
You want to do search on thosethings.
Similarly, you want to do RAGfor analytics widgets.
There's so many widgets thatpeople build in the enterprise
that need to be searchable.
So you went from promptengineering to RAG, to RAG for
everything.
And then you realize that peopleare looking for reasoning.
(39:24):
Like, hey, I want to reason,because now you're going from a
lot of peripheral agentic work,which is what happened in
customer support, front office,l1, l2 support.
Now you're getting to L3, l4.
And these are getting closer tothe mid office and then back
office.
There you need to havereasoning.
(39:44):
You need to spring signals fromfive different sources.
Look at the history of the lastthree years of how we did things
.
So then you need to do signalsfrom five different sources.
Look at the history of the lastthree years of how we did
things.
So then you need to dosupervised fine-tuning, and I
think one of the things I justheard recently, controversially,
when Nandan was saying thatIndia nobody needs to learn how
to fine-tune a model and I sawArvind talking about this from
(40:06):
perplexity that he's so wrongand I agree.
I think, going back to yourquestion on enterprise software,
and why would India not producesuch companies in the future,
it's because we're not going tobe deep enough in AI, the new
systems and systems engineeringand building infrastructure is
(40:27):
to go beyond prompt engineeringto not just RAG, also to
supervise fine tuning, smallmodels running on the edge
hosted by you and Kubernetes,maybe bedrock, but I think
understanding that spectrum issomething that we'd love to
share with startups and havethem really build deeper
(40:47):
businesses, not just, you know,gpt wrappers, which is what a
lot of prompt engineering is,but to build deeper businesses.
Speaker 3 (40:56):
No, very, very, very exciting, very, very exciting
thoughts.
Just one more thing about justmore broadly, beyond Evrev how
do you think about the SaaScompanies of yore and by yore I
mean the last five, seven years,I don't mean 10, 15 years ago?
How are you seeing theirevolution into the AI world?
(41:19):
Because obviously now they'rethe incumbents and they don't
have the luxury of starting AIfirst.
They have to adapt to AI.
So how do you think about anexisting SaaS company founder
who's already three, four, fiveyears in, has products, has
customers and has maybe two,three, five, 10, 20 million ARR
for them to adapt to the AIworld?
Speaker 1 (41:38):
So if you look like 10 years back, a lot of the
modern PLG SaaS companies I meansome of them got acquired, like
Slack got acquired, figma,almost so.
Notion is trying to reallyfigure out its place in the
world of AI.
Canva has been another PLGdarling, but they will need to.
(42:06):
I mean, maybe they need to findmore time to embrace AI,
because the DNA of thosecompanies was probably more
design and they haven't figuredout AI because the DNA at the
very top is not beyond designand we look at AI and design as
kind of two sides of the samecoin.
In fact, ai and UI are kind ofthe yin and yang that you need.
(42:32):
So I think that a lot of themodern companies that are
building PLG era, they will findit hard to make any more AI
than what ChatGPT alreadyprovides you.
If you look at verbs likesummarize or generate, based on
some points and things like that.
(42:53):
Now, maybe the win for them isthat they keep it integrated and
maybe the simplest and mostpowerful thing that they can do
is like, hey, you don't need tobuy a license for ChatGPT if I
can provide that to you andmaybe that's good enough.
You don't need to do anythingmore than that, that they become
chat GPT wrappers for the lowend of the market and maybe low
(43:15):
end of the mid market orsomething.
But I think the last five,seven years, most companies that
have been formed they didn'thave enough of, I would say,
systems chops, infrastructurechops, because AI is now an
infrastructure problem.
You know, anything that couldhave been GPT-wrapped has been,
will probably get done and maybethose people will bring a lot
(43:35):
of go-to-market skills on theother side and maybe that's one
formula that I'm seeing succeed.
It's okay, I'll actually pick adepartment or a function in a
company and just go deep intobuilding AI for them and that
probably is a good I would say$100 million to $100 million
(43:56):
story in terms of revenue, andthen you just be very good at
inside sales and thego-to-market machine and
probably building some reallygood workflows.
But I think where it starts tobecome a problem, because if you
can't make a platform, youcan't make half a billion, a
billion dollars in revenueperiod, because to go upmarket
(44:17):
if you only need to make abillion dollars, and how do you
go upmarket if you don't have anextensible platform where the
partners, the SIs, theAccentures, the Infosys and even
the customers.
Developers can actually buildstuff on top of, and the most
important stuff they'll build istwo things workflows and
analytics widgets.
So I think the company's lastfive, seven years either they'll
(44:41):
be very good like Gong was.
Gong was very good in 2020,2021 era and all like oh, oh
fast to 300 million.
And the question is now what?
You know, uh, basically therewas so much churn, uh, because
there's so much zarp money, thatuh went away with interest
rates being high, that the smbsuffered.
So I feel like going to the midmarket in enterprise is
(45:04):
probably the real challenge andopportunity for a lot of
companies right now, because theinterest rates are where they
are, whether we like it or not.
Until the wars end, untilUkraine normalizes, I think it's
really hard.
I mean Europe has to actuallyget back to normal.
I mean Russia has to start toprovide more things to Europe.
(45:24):
Those things start to happen.
Then interest rates come down.
Until then, there's a reallyhard thing, especially
non-digital startup companies.
They're not coming back anytimesoon.
Startups in software areprobably one good thing.
That is still okay in the SMB,but everybody else has to really
think about the mid-market andthe enterprise.
(45:45):
Rajat.
Speaker 3 (45:47):
Mittal Fascinating, dheeraj.
So, as we bring this to a wrap,you mentioned one very
interesting thing about PMF at 1million, 100 million, 250
million, a billion I'msimplifying a little bit, but
there's also I'm just going tomake up a new term like
organization or founder marketfit at each stage.
(46:08):
So how do you keep evolving asthe company is scaling, and
whether it's you or yourco-founding team or your
leadership team and any thoughts, and obviously, since our
audience is a lot more earlystage entrepreneurs even, let's
say, do the exercise at 1, 10,50, 100, you know, like at the
early stage of that ladder, andwhat are some kind of tips,
suggestions, maybe even amber oryellow flags to watch out for
(46:32):
so that you know you're yourselfscaling with the company?
Speaker 1 (46:36):
Yeah, I think at the core, it's about how do you
really get to relate with peopleover time, and this includes
not just investors and yourboard members, but also your
executives.
You hire some executives whoare very good at what they do,
and the only way you can retainthem is if you know that they're
good at things.
Then you are, and buildingpeople relationships is probably
(47:02):
another way to really build agreat mid-market enterprise
business too, because people buyfrom people in mid-market and
enterprise.
I think process is anotherpiece that people have to start
to respect, because processmatters, efficiency matters,
sustainable stuff matters.
Growth hacks don't work whenyou're starting to really grow
(47:23):
bigger.
So a lot of founders who areproduct people, how they really
begin to embrace people andprocess is probably the way to
do this, cause then more peoplewant to come and work for you,
work with you.
Uh, and in this, in this wholejourney, you need to really know
what it means to let go ofthings that were very near and
(47:43):
dear to you and also be verygood at negotiation, because
your word and your sort of whatused to be the diktat is not
going to be that you know sortof impactful anymore, because
then you're only hiring doers,as opposed to people who are not
(48:05):
just doers but thinkers.
I think how you continue tohire thinkers and of course not
every thinker will align withyou.
They'll say, let's do this andlet's do that and this is not
going to work.
So it's such a fine balancebetween the things that brought
you here and what percentage ofthat you continue to keep and
(48:25):
what percentage you continue togo and evolve is basically at
the core of this.
So learning to negotiate,learning to actually really
build those bridges with peopleyou're hiring and with customers
, respecting people and process,is the only way you can really
scale a company.
I mean, because not everybodycan be Elon Musk or Bill Gates,
(48:47):
you know.
I mean even Jeff Bezos.
I mean I'm a big fan of Jeffand you know people like Steve
Jobs.
You know they said, look, andJeff himself had gotten so good
at framework thinking.
You know he probably was verygood at it from day one.
But this idea that, look, I needto bring framework so that now
(49:08):
I can leave that behind in ameeting room so I don't need to
be in every meeting.
So frameworks also become agood way to scale the company,
because then process thinkingcomes, scalable thinking comes.
But someone like Steve realizedthat if he can't deal with
people, you need to get a TimCook and you need to complement
(49:28):
yourself with that kind of aperson who deal with people and
process.
There was a lot of process inApple dealing with China and
manufacturing and Foxconn andall this stuff, and then knowing
that the people within Applehad to respect that process to
go and sit in China for sixmonths every year if that's what
it took to really build Applein.
(49:49):
So really at the core is peoplein process.
Speaker 3 (49:53):
Yeah, I think people process and delegation.
That's a wonderful place to end, and I remember at Google,
there was Eric Schmidt, therewas Sheryl Sandberg, there was
Nikesh Arora.
There was a whole bunch ofpeople that were brought in
because the founders had thecreative mojo, but they were
like all this other stuff I need.
We need people to do this right.
And Dheeraj fascinatingconversation.
(50:14):
You could go on and on, but Iwant to be respectful of your
time.
Thank you so much for being onthe Prime Venture Partners
podcast.
Speaker 1 (50:22):
Yeah, it's a pleasure myself.
Thank you again and hope youraudience finds it's meaningful.
Speaker 3 (50:27):
Thanks, Dheeraj.
Speaker 2 (50:31):
Dear listeners, thank you for listening to this
episode of the podcast.
Subscribe now on your favoritepodcast app for free and you'll
be the first one to know whennew episodes are available.
Just search for Prime VenturePartners Podcast in Apple
Podcast, spotify, castbox orhowever.
You get your podcasts, then hitsubscribe and if you have
(50:52):
enjoyed the show, we would bereally grateful if you leave us
a review on Apple podcast.
To read the full transcript,find the link in the show notes.
Popular Podcasts
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
24/7 News: The Latest
The latest news in 4 minutes updated every hour, every day.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.