0:02
Hey, Shashir. We're talking to you at a pretty interesting week. Uh, you all
0:07
have some news to share. You want to share.
0:22
Hey folks, welcome to Agents of Scale. It's a show where I sit down with execs
0:26
and founders and CEOs uh who are using AI to transform their companies. I'm
0:31
Wade Foster and today I've got a really exciting guest. Uh I'm talking to
0:35
Shashir Morotra. He's the CEO of Superhuman uh which last week he was the
0:42
CEO of Grammarly and I think not that long ago he was the founder and CEO of
0:46
Kota. He probably has the most chaotic LinkedIn updates in the last year of any
0:51
uh founder at least I've talked to. Uh Shashir is uh someone I've known for a
0:55
long time. He's one of the clearest product thinkers. He's led product at
0:57
YouTube. He co-founded KOD. Uh he's on the board of Spotify. He loves bundles,
1:02
pricing, and packaging. Uh Shashir, I'm stoked to have you on the show. Welcome.
1:07
Thank you. Thank you. Yeah, the the LinkedIn thing was a little bit of
1:10
chaos. Although there was one person with a more chaotic LinkedIn than mine
1:13
was Rahul Vora. He had a he had a that was a pretty fun
1:19
twe uh tweet exchange with someone trying to reconstruct the the sequence
1:24
there for him. Uh well, hey Shashir, we're talking to
1:27
you at a pretty interesting week. Uh you all have some news to share. You want to
1:31
share like what what has changed at Superhum in the last week?
1:36
Yeah. So, uh maybe I'll I'll just give the the the highlights and we can dig in
1:41
the different parts. But uh as as you mentioned, I started KOD um back in
1:47
2014. Uh we uh sold the company to Grammarly uh late last year and we've
1:54
been working for the last 10 months on uh a set of really important
1:57
announcements and they all came out uh two days ago. Um and there are three big
2:02
things we did. Uh we rebranded the name of the company. We added a new uh
2:06
product we call Superhum Go. and we um added a a way to buy all the products
2:12
together as a suite in a in a bundle. So each one is interesting. We rebranded
2:16
the company. Uh the corporate name went from being Grammarly to being
2:20
superhuman. It's still um uh houses all the subbrands. So it's Grammarly, it's
2:25
KOD, uh mail, and now the new product go. Um and then Go is our new AI
2:32
assistant that really takes the best parts of Grammarly, the underlying
2:35
engine of Grammarly, and turns it into a platform. uh such that anyone can have
2:40
um assisted agents that work with them embedded in their experiences
2:45
everywhere. Uh and then we have a bundle where if you're a super fan of any of
2:52
the products, you can easily get access to the others as well. So that's uh
2:55
that's what we did. I love it. Um I'm gonna ask the fifth
2:59
grader question here. What does Grammarly, a writing assistant, Koda, a
3:06
document collaboration platform, and superhuman, an email platform, and go?
3:12
What do they all have to do with each other?
3:14
Yeah. Yeah. That's a it's a great question, and I it starts with a pretty
3:18
simple observation about the world of productivity. And so, in my view, we're
3:22
in this third era of productivity. Um, the first era was all about digit
3:27
digitization. We turned typewriters into word processors. The second area was
3:32
everything moved to the web and it was about collaboration and all of a sudden
3:35
every product was something where you could collaborate with other humans and
3:38
you knew you were in those products because like Google Docs and Slack you
3:41
knew you were in them because all the avatars showed up, you know, all the
3:44
people you were working with. And in this third era, it's where we all get to
3:47
work with AI or sometimes people use the term agents which is appropriate for
3:51
this podcast. And uh and in that world I almost mentally picture it as those
3:57
avatars are shifting from being my teammates to being my virtual teammates.
4:00
Um so in that world there's an opportunity to rethink every part of the
4:05
product productivity experience. And so that's that's what we're working on
4:09
together. Um we think about ourselves as doing uh uh three main sta uh uh pieces
4:15
of that. So the fundamental one for us is is go. Um that's our platform for
4:20
agents. Uh next on top of that we build a set of agents. Grammarly is one of
4:25
those agents and we build many many more and then we build surfaces that those
4:28
agents work in and the two most prominent ones are the two places where
4:31
people spend the most time in their documents and their email. So that's how
4:34
they all connect together. So tell us more about superhuman go. I
4:38
think folks are familiar with superhuman with uh I guess now mail uh uh kod and
4:44
grammarly but go is the new piece of the puzzle. So that's right. What what does
4:47
go do? Yeah. And and maybe the the easiest way
4:50
to tell the history here is it it starts with a different way to think about
4:54
Grammarly. I I think Grammarly is one of the most misunderstood products on the
4:58
planet. I mean, f first off, it's way bigger than people think. You know, it's
5:01
a it's about 40 million daily active users. Um you know, hundreds of millions
5:05
of dollars in revenue. Um you know, a lot of people underestimate it because
5:09
it's a it's a product that feels is meant to feel incredibly simple and
5:13
almost invisible. It's a it's with you everywhere you work. Um, and most people
5:17
think it's about grammar, which is a totally reasonable assumption, but
5:20
actually most of the code, most of the technology of Grammarly is not about
5:24
grammar. Most of it is about bringing AI to where people work. Um, so we
5:29
sometimes refer to that as the AI superighway. We view Grammarly as the OG
5:33
agent. Uh, for the last 16 years, it's been bringing a writing assistant right
5:38
to where you work. So the key capability of Grammarly is we work in about a
5:41
million different applications, web applications, desktop applications and
5:45
mobile applications. Everything from Gmail to Google Docs to desktop apps
5:49
like Slack and Word um uh out to the you know thousands of different enterprise
5:53
applications and consumer applications you have and everything on your phone as
5:56
well. Um we can read what you're doing uh annotate it in a way that's
6:01
unobtrusive to you and to the application and we can make changes on
6:05
your behalf. Um, so that's what we call the AI superighway. There's one problem
6:10
with our AI superighway. We only run one car on that highway, and that's the one
6:14
that has your high school grammar teacher in it. And it turns out to be a
6:17
really valuable car that, you know, can drive tens of millions of users and and
6:21
hundreds of millions of dollars of revenue. Um, but it's just one car. And
6:25
our insight was what if we took that part of Grammarly and we sort of
6:29
separated it out and we said, this is can be a platform where we can run many
6:33
such agents. We can run many of these cars. So maybe to give you an easy
6:37
example, imagine you're a salesperson. You're writing an email to a client and
6:40
you know today Grammarly feels like you have your high school English teacher
6:44
sitting on your shoulder helping make sure you don't make an embarrassing
6:48
mistake. You communicate most effectively. Um you grammatically
6:51
correct, but now you also have all these other agents sitting on your shoulder.
6:55
So you have, you know, your your sales agent that's read your entire CRM and
6:59
tells you this is the product that that customer is interested and this is the
7:02
ones are not interested. There's another one that's your support agent that says,
7:05
"Hey, you know, this client had an outage last week. You should say
7:08
something in your email about that." Um, there's another one that knows your
7:10
calendar and says, "You said you can meet Tuesday at 7 p.m., but actually you
7:14
have your daughter's recital and you can't meet then." And all of them are
7:17
working together to help you communicate and collaborate better. So, we decided
7:21
to pull that part of the product out. We gave it a new name. So, we call that
7:25
superhuman go. Uh, it you can run many agents on it. There's a store with full
7:30
of agents. Um the most important agent that we think everyone will start with
7:34
is the Grammarly agent. So that's your grammar teacher. Uh but we built uh
7:38
hundreds of others. Actually, you know the history behind this, but we
7:41
basically took every Koda pack and we turned them all into agents. Um and so
7:46
there's now hundreds of those and we open it up so anybody can do it. And so
7:49
we had dozens of of agents launch Wednesday as well from you know all
7:53
sorts of different brands and ideas. you know, uh many different uh ways of
7:57
thinking about uh what it means to bring AI directly to where people are working.
8:01
So that's what we're up to with code. I love it. So you bootstrap it with
8:04
Grammarly, the agent, with the Kodapac agents. Um what's an example of one of
8:10
the newer agents that exist on this super highway?
8:13
Yeah. So there's a lot of interesting ones. I mean, I I'll pick uh one I was
8:17
really excited about was one of the first people I pitched when we were
8:19
talking about the superighway and really, you know, what what would you do
8:22
with it? Uh, and I was talking to uh a book author actually, Kim Scott. She
8:26
wrote a book called Radical Cander. And uh, Radical Cander is a great book. It's
8:30
all about how to communicate effectively. Most people who read it,
8:33
it's Kim Woodstall and me, they take it, they stick it on their shelf, and they
8:36
kind of forget about it. And it's a that's the nature of all the books we
8:39
read. Um, and she said, "What would it feel like if we actually could take that
8:44
book and have it work with you everywhere you are?" So, they shipped a
8:48
radical cander agent that just like Grammarly helps make sure you're
8:51
grammatically correct. This helps make sure you're following the principles of
8:54
radical cander and it just takes anywhere you're writing. You can be in
8:57
Slack, you can be in Google Docs, you can be trying to take an interview note
9:01
in Greenhouse or you can be anywhere and uh you know it feels like Kim Scott is
9:06
sitting on your shoulder ready to help you be a better communicator. So it's a
9:09
I thought that was a really fun example. You know I love that use case. Uh we
9:13
have an agent Zap that does something similar where uh we're fans of the book
9:18
uh five dysfunctions of a team. Yeah. Yeah.
9:20
And you know, same phenomena, right? You read this book and you're like, oh,
9:23
these concepts are great. I'd love to, you know,
9:25
right? Yeah. Lion. And then you put the book on
9:27
a shelf and you're just like, well, you know, hopefully hopefully we remember
9:31
what's in it. And so we built out a little agent that takes uh our
9:36
transcripts from our exec team meeting and it runs it through a prompt that's
9:39
got all the concepts from the book and then it DMs people feedback after
9:43
meetings uh in Slack. So like, you know, here's what you here's what you did
9:46
well, here's what you didn't. And it's great to have this like ambient feedback
9:49
just always because I've never come across somebody who's like, "No, I don't
9:53
want to, you know, I I want to be a bad teammate or, you know, I want to I want
9:57
to be on a dis dysfunctional team." Like I think we all want to be on good teams,
10:00
but the reality is like in the moment of shipping products and, you know, trying
10:04
to make customers like you don't always show up your best and it's nice to just
10:07
have little a little coach like nudging and I I so I'm I'm pretty bullish on
10:11
this concept of AI as coach like I think it's one of the the best early use cases
10:15
so far. Right. Right. Yeah. Yeah, and I think
10:17
and I think it, you know, for our our approach to it is you just want that
10:20
coach kind of with you everywhere you are. That that's and that's that's the
10:23
heart of what we're doing with uh with Go.
10:25
I love it. Um so what a second thing you mentioned is
10:30
you have the the name change for the corporate entity. You adopted superhuman
10:34
as the name and one of the things I heard you share on X was that this
10:38
reflects this bigger vision for the company. Can you talk a little bit about
10:42
where is this going beyond sort of the like uh you know information superighway
10:46
this AI superighway right and like why why superhuman I mean it's
10:50
clearly a phenomenal name but you know Grammarly's got 40 million users so
10:55
yeah I I mean I think first off I think the it's really important to clarify for
10:59
everybody we're renaming the corporate name not the not the brand name so those
11:02
those all continue uh it's a little bit similar to um what Google is to Alphabet
11:08
what Facebook is to Meta what Square is to block. I think I I think there's a
11:12
good history of companies as they broaden their ambition needing a
11:15
corporate name that allows them to to sort of spread their wings across
11:18
different areas without confusing people. Uh because otherwise you take
11:21
the Grammarly name and you stretch it too far and and you say, "Well,
11:24
Grammarly now stands for all these other things." And that's actually the the
11:28
beauty of the Grammarly name is that it's incredibly clear what it does. Um
11:32
you know, it's a downside, but it's also it's also a a uh uh a clear uh positive
11:39
as well. Um, and so we knew we wanted to rebrand the company. Um, especially as
11:44
we started adding more uh products and capabilities to our portfolio and we
11:48
added Kod and we added the Supreme Mail team and and we're starting to work on
11:52
Go and we said okay we need a different corporate entity. uh we brainstormed I
11:57
mean you've been through this process but the we brainstormed hundreds of
12:00
names uh went through many many different examples and interestingly the
12:04
name superhuman before we even bought the company kept coming up like it it
12:08
would it would show up in these you these brainstorms and people say you
12:11
know this is actually a really good name our design team even did a riff on a
12:15
brand system and they called it the superhuman design system and they said
12:19
we're using this name because it's not possibly that we're going to use it as a
12:22
company name it's another company but it really reflects the spirit of what we're
12:26
doing. Um, and so it was kind of, you know, running around in our heads. And
12:30
then when I ended up talking to Rahul and deciding to to pull our companies
12:33
together, um, you know, after that closed, we we sat down and talked about,
12:38
you know, what would it what would it feel like to to to swap the names by by
12:41
the way is it's pretty atypical, right? So that's mo everybody loves the name,
12:45
but it was it I went looking and the number of companies that have done this
12:49
pattern of take the the the acquired company's name as a parent. I there's
12:53
like five or six major companies that have ever done it. It's like the the the
12:56
only prominent example I could find was SBC and AT&T, which was like a long long
13:00
time ago. This uh doesn't happen very often. And I' I think having lived the
13:05
last three or four months of how to do this, I can understand why it's really
13:08
hard to do. But to answer your question first, what what do we love about the
13:12
name Superhum? I we're really looking for two things. One is we wanted
13:15
something that could capture the breath of our ambition. And so it's really
13:18
important for something that could flex as we enter new categories and and make
13:22
sure that you know uh people see it as a as a broad enough name to do that. Um
13:27
but the the second one is really a focus on empowerment and I and I think that is
13:32
a very different way to think about the phase we're in. So we talk a lot about
13:35
we're in this AI era. I call it the third phase of productivity. You know
13:39
clearly AI is affecting a lot of different people and there's a set of
13:42
people whose view is AI is here to replace humans. And you my view is quite
13:46
the opposite and our company's view is quite the opposite. We think we're here
13:49
to empower humans. We're here to make humans even more human. And I think that
13:54
while the name superhuman that, you know, super is going to get used as a as
13:58
an adjective a lot, but it's actually the the human part of it that we were
14:01
most drawn to. Um, and if you think about it, you know, Grammarly is a
14:04
classic tool like that. We don't, you know, Grammarly is a a very popular
14:07
assistant. It helps you everywhere, but at the end of the day, it's still you.
14:12
You still send the email. you still publish the uh the blog post, you still
14:16
submit the essay, it's still you. Our job is to bring out the best of you with
14:21
the best poss possible assistance. And so that's how we framed uh the goal for
14:25
the name and superhuman just fit fit like a glove. I mean uh we re rewrote
14:29
the mission statement, the new mission statement of the company is unlock the
14:32
superhuman potential in everyone. We believe that potential is already there.
14:37
It's not the job of AI to replace it. It's the job of AI is to bring it out to
14:40
to everyone. So that's why we picked the name. I like the focus on human. I'm,
14:44
you know, this is one of the things I love about Superhuman, the the male
14:48
client, is that there's these like just very subtle like well done features that
14:53
are just nudging you to do the right thing, helping you go faster, injecting
14:57
information and context just like right at the right point where it's still me
15:01
to your point, like still me, but like I'm getting a lot of help to to to fire
15:05
off a bunch of emails really well and I feel super human uh at the end of the
15:09
day. That's right. Um what's um so you've got
15:13
this information superighway like what what what what are you going to put on
15:17
it next? What what company are you b what's the fourth company that's getting
15:20
bought here? Um first off I'd say you know in this in
15:24
this world of building this platform this AI superighway it's um it's
15:29
actually our our default motion isn't buy our default motion is partner. So I
15:33
think the I think the most important thing we can do is open up this platform
15:36
to everybody. And so I think you know most of the companies I now talk to
15:39
that's the discussion is what would you do if you could put an agent on the
15:43
platform and really finish the journey. And you know for many people that is
15:48
about I I I've often referred to it as the last mile of AI. Um and I think it's
15:54
uh you know it's a it's an interesting way to think about what's happening in
15:57
AI that a lot of AI products are all about go to AI. you know there's this
16:01
like actually one of the customers we announced um this week as well uh one of
16:06
the first customers of go is going to be uh actually university Arizona State
16:09
University and um it's a really interesting organization they have
16:12
they're uh 180,000 students it's enormous and it's also 18,000 employees
16:17
let's say you know it's pretty big on in both dimensions and they told us about
16:21
they built something like 5,000 AI applications they have everything I mean
16:24
they have they have you know every class has a chatbot and every uh and every uh
16:29
um uh you know they have one for picking your degree, they have ones for their
16:33
employees, they have all these different chat bots. And the problem is nobody
16:35
goes and nobody remembers that that's where where I'm supposed to go. And so
16:40
their view of what we're doing is we provide this last mile of AI. We allow
16:44
that professor chatbot to actually sit next to you while you're working and be
16:48
this kind of digital twin of your professor. And you know that that just
16:52
changes completely how uh how you think about the role of a professor for
16:56
example. So there's, you know, I think our default motion here is going to be
16:59
partner. And I think there's a really broad set of of tools for which their
17:06
biggest challenge is they're actually just not in front of you at the right
17:08
moments. And if we can be that platform that gets uh gets gets you in front of
17:12
people, then that's that's that's our job. That's that's what we can provide
17:15
for the world. You know, I'm I'm sure we will end up doing more acquisitions. I
17:19
think it's a a motion we're we're getting good at. There are cases where
17:24
you know we feel like it's not the partner relationship isn't enough. Uh
17:28
part of the reason why we bought superhuman mail was we felt like it was
17:31
an opportunity to actually reinvent the surface a bit. A very important surface.
17:35
Email is the number one use case for Grammarly. It's actually the number one
17:38
thing people do with Grammarly is write emails. Um and which kind of makes
17:41
sense. You know I think we're people picture this as docs but actually you
17:44
spend way more time writing communication than you spend writing
17:46
documents. Um and uh so we looked at it as a sort of fundamental as part of the
17:52
the platform for what we can do and there we had to own it to reinvent it.
17:56
And so I think that will be more of our bar. We're going to default to
17:58
partnership everywhere we can. But if we find cases where you know we think we
18:03
can make that a core part of the platform. I think that's when we'll be
18:07
more interested in uh in an ownership model.
18:11
I love it. Um so let's shift gears here. Um, we talk a lot about AI
18:16
transformation on this podcast and you're doing a very unique kind of
18:21
transformation. You're taking three companies, three distinct cultures,
18:26
three distinct product road maps, three distinct customer bases, three distinct
18:30
brands, and you're all coalesing them into one. At the same time, AI is
18:35
changing how all these companies work both internally and their product road
18:39
maps. And so I guess the question is what are you
18:44
learning about how to do this? Well, well I so first off it's four. So we
18:49
have four products in it. I guess that's true. Four companies.
18:51
We have Go now as well. So So it's four. But Go is new, right? Like that that at
18:56
least gets to be although interestingly it's you know so
18:58
that the the uh I I wrote a uh a bit of a a manifesto
19:03
on how we were going to run as a company and I titled it the compound startup
19:06
manifesto. Um, and uh, it's a term I borrowed. There's a bunch of people that
19:10
have been using it lately. I think Parker at Rippling uses it a lot. I
19:14
really love it. I think it's a really good way to think about what happens to
19:17
companies as they as they grow. And I've seen every pattern. I mean, I've seen
19:21
companies that grow and they transition to being multi-product and it all sort
19:25
of falls over. And I've seen ones where it completely flourishes and it sort of
19:29
unlocks the company. Um the the you know from my experience the the best success
19:34
story in my history about this was my time at Google. I used to run the
19:38
YouTube group there and you know when I got there was right at the time when uh
19:43
Google sort of figured out this compound startup motion and it was a huge unluck
19:48
I think but most people probably uh don't remember or forgotten that you
19:51
know 2004 2005 Google was a single product company and you know to the
19:56
point of there was 10 principles for Google and uh uh listed on the website
20:00
and one of the principles is we only do one thing and they actually had to pull
20:04
it off the website when all these products started to launch. ing and in
20:08
the period um uh actually there's a there's a really good podcast about
20:11
this. The acquired team just did a podcast on how the period from 2005 to
20:14
2012 Google launched eight billion user products um and got them to got them all
20:20
to scale. And you know I happened to be driving one of them the the YouTube one.
20:25
But it's actually interestingly how in the same time you know Gmail was being
20:29
created and Google Docs was being created and doubleclick was being uh
20:31
grown and the Google maps business was growing and Chrome and Android and all
20:35
of these different businesses were happening all at the same time. And uh
20:38
so anyways I think there's a lot of lessons for companies that figure out
20:42
how to do that compound startup motion well. And there's always this tension of
20:47
you want to create these uh these units, these teams of people that feel like
20:51
they have really good autonomy that they can iterate like you would as a startup.
20:54
They know their customer base really well. They know the the the customer
20:58
alternatives, the other competitors. They know that really well. They
21:01
understand the technology stack really well and they can dream and iterate
21:04
really really quickly and respond to the market quickly. Um and at the same time
21:09
you want an infrastructure that gives gives an advantage to all these startups
21:12
because you know the whole point of being together is that they get to do
21:16
some things together. And so how do you how do you find that that uh that
21:20
magical interlock is a big piece of what we're doing. So we call that our
21:24
compound startup model. Um I think we're early in it but I I do think it's a it's
21:29
an interesting test how we're doing it. And you know these are four business as
21:32
you mentioned at totally different scales. goes brand new, you know, just
21:36
they're two days they're two days old. I just went through the first set of
21:38
metrics reviews with them and you know it's like what you feel like after two
21:42
days of company. You got grammarly at 16 years old and it's a lot and you know
21:46
has a massive install base and revenue line and has a very tuned motion and so
21:51
on. But you have to run each of these things with the right level of uh
21:56
attention process, you know, support in some cases, leave it alone in other
22:01
cases. Um so that's what we're doing. What were the key tenants or principles
22:05
in that compound manifesto that you wrote?
22:09
Uh, oh boy. Um, so, uh, this will be an interesting discussion. I don't think
22:15
I've talked about this publicly yet. The the, uh, uh, I I think the, it's fine.
22:19
The, um, so maybe first off, just to give the extremes for a moment. When
22:23
people call talk about a compound startup model, there's in my mind,
22:28
there's two extremes. So there's one extreme is the the fully functional
22:32
overlaid model. So that that's the the case where you know you have the entire
22:37
company's organized functionally all the way up to usually up to the CEO and
22:41
everything you're talking about is is sort of teams uh uh built together
22:45
across that. Um and you know there are some companies that that still work that
22:49
way. Apple still works that way. Um is uh is somewhat amazing. actually most of
22:54
that time at Google when all those com when all those products were created you
22:57
know that's how Google worked as well um the you know Eric Schmidt ran the
23:01
company with a head of engineering and a head of product and a head of sales and
23:04
and so on that's how you know that's how I operated at most of my time at YouTube
23:08
um the uh the other extreme is you organize into business units and uh you
23:14
know that that uh in that frame you end up generally with something like a
23:18
general manager for each business unit and all the functions sort of report
23:21
into that person and you kind of tolerate some level of duplication and
23:25
and some level of difference between how you know that's an engineering team over
23:29
here and an engineering team over there and a sales team over here and a sales
23:31
team over there and you know when you're a huge company like uh maybe I'll pick
23:34
like uh G like they have you know they have a unit that makes aircraft engines
23:37
and a unit that makes dishwashers and they don't have a lot to do with each
23:40
other and so they they they run that way and so that in my mind those are the
23:44
extremes and then there's a lot of options in in the middle um and you want
23:48
to try to find that that middle sweet spot and you know by the way this
23:53
this discussion. My first time even thinking about this
23:57
discussion actually happened when I was in college and I was um for whatever
24:01
reason at MIT they let you take classes at the business school which I I think
24:05
is uh um for some people it's great but for me I think it I was just way too
24:09
early for this 17 years old I take my first management science class and they
24:12
spend like literally like a third of the class on matrix management and this
24:17
tension between functional organization and business units and and I'm just 17
24:20
years old this is so stupid like this is you know just pick one how does it
24:24
matter and Now here I am 30 years later and that's like we spend all the time
24:28
innovating in the space and it turns out it's really hard and there's no right
24:30
answer between these things and you're going to pick the the pros of one and
24:34
tensions of the other. Um but there are a couple tests I've I've landed on that
24:38
are that are really really helpful. Um and I I can share lots of them but the
24:41
one that I think is the most helpful is something I call the hoodie test. And
24:46
it's one we we we really um somebody made this analogy at YouTube and it
24:51
really stuck with me through every job. Uh, and it was a really simple test is
24:55
you come in every day and or you come in the next day and you say, "Uh, hey,
24:59
tomorrow everybody please wear a hoodie with your team's logo on it." We used to
25:04
give out a lot of hoodies of swag. So, wear the hoodie with your team's logo on
25:06
it and see which logo everybody wears. And one of the things that happen is
25:12
that everybody gets very focused on am I functionally organized or business unit
25:15
organized? But actually doesn't matter. What matters is what's on your hoodie.
25:19
And uh, and that leads to a lot of different dynamics. So, first off, you
25:22
know, your matrix in this organization. Usually, it's easy with your EPD teams,
25:25
your engineering team, your product team, your design team. They're always
25:28
kind of organized this way anyway. It's like there's the go engineering team and
25:31
the go product team and the go design team. But then you get out to these next
25:34
functions and it's way harder, right? So, now you got you've got a marketing
25:37
team and it usually doesn't make sense to organize a marketing team that way or
25:40
a sales team. You almost never organize a sales team by product. Even you get to
25:45
your lawyers like do you do you give, you know, does each of these teams kind
25:48
of have their own general counsel? um you know your finance team and so on.
25:52
And what you want is for this team to feel like a unit to feel like a
25:57
business. You don't actually need all these people to report there. What you
26:01
need is them to feel like they have a staff meeting full of people all wearing
26:04
the same hoodie. Um and the the more complete that staff feels, the more this
26:08
unit will feel like they can run uh run their business. And I would say that's
26:14
the sort of middle spot we focused on is it's really not about, you know, so we
26:17
actually all report functionally. So the the company is is uh from that
26:20
perspective closer to Apple or how Google was in those days. Um but we put
26:24
a lot of emphasis on building this cross layer and giving really clear
26:29
accountability and uh you know participation to who's the
26:32
representative from each of these functions that make this thing feel like
26:35
a unit. How do you get those folks to um when it
26:39
matters to care about superhuman versus team Grammarly or team kod or team mail
26:45
or team go? Yeah. Yeah, I mean it's that's a really
26:48
good question and I think uh that's another thing that happens as um as
26:52
teams sometimes migrate away towards that business unit model you lose that
26:56
connection and you get you get tension on it. Um uh I think some of it first
27:00
off is structural. Like if you just start with what which company do you
27:04
have equity in? Um and I think that's a I think that's usually the first test is
27:07
like you know I worked on YouTube I got Google stock and it was really clear
27:11
when push came to shove and you know the google.com team called me and said hey
27:15
can you help us out with this thing? You know we all understood we own we own
27:18
Google stock. So I do think there's a there's a structural alignment and
27:21
there's a clear line you cross. When you cross that line you have a different
27:24
structure. now you're a conglomerate and that has different there there are pros
27:27
and cons of that as well but that's not that's not what what we're doing um uh a
27:32
lot of it is also in your values and behaviors you know we have one of our uh
27:36
we actually just we wrote the mission for the company we wrote uh our six
27:39
values as well one of them is this thing we call prioritize the pond um and the
27:43
basic analogy is that sometimes you focus on the size of the fish or the
27:46
size of the pond um and it's meant all the way from the individual up to the
27:49
company um but it's meant to capture this idea of you know when when in doubt
27:53
you're supposed to prioritize the pond you you talk about it, you repeat it,
27:56
you you amplify those examples so people people understand them. Um so I think
28:00
that's really important. Um I think that also leads to for example in that value
28:05
we say you want people to prioritize the pond you need to be transparent with
28:07
them so they understand the pond like they they're going to understand their
28:10
part of the they're going to understand their part uh anyways you know they they
28:15
what they don't understand is what's happening uh uh overall um and then you
28:19
lead by example. you you know you make every every team does it and you you uh
28:24
uh you help them get there. But I think if you get the structure right and the
28:27
values right you know I I haven't you know honestly here I haven't seen any
28:30
issue with this um and uh you know if I go back to Google or my time before that
28:36
at Microsoft I mean it took time eventually you had issue with this but
28:39
for like you can do a lot with culture. Yeah. Uh what have you learned about
28:45
smashing these three disperate cultures together?
28:50
Oh boy. Um, you know, I think I think it's uh it can
28:57
be really hard. I mean, I think the the and these are uh as you said, these are
29:00
three companies with very deep cultures. I mean, you ask any employee of any of
29:03
those three companies about their the the values of the company or the stories
29:08
or so on. These are, you know, two 11-year-old companies and a 16-year-old
29:12
company coming together. And so, everybody coming in had a view of this
29:15
is this is how we work. Um, and so first off, you need to do a lot of sharing.
29:19
You need to let you need to help people understand who who each other are and,
29:24
you know, get them together and and make sure that they're seeing each other. One
29:27
of the things we started doing right away was we moved from doing a monthly
29:30
all hands to doing a weekly all hands. Just like we're going to see each other
29:33
more often and that's a, you know, I think it's a a a small thing but an
29:37
important thing that, you know, it's one company, we're all going to see each
29:40
other. This is how it's going to work. Um, I think you really want to amplify
29:44
some of the best parts of each culture. Every company comes in with some
29:47
strengths and some weaknesses and you want to you want to you know sort of
29:50
gently ignore some of the weaknesses but you really want to go and pull up
29:54
examples of like hey that's a thing that team is really good at and I want
29:57
everybody to notice this is a this is you know for example the superhuman team
30:01
you know one of their values is to create remarkable delight which if
30:04
you're if you're a user of the product you immediately know what that means I
30:08
mean it's a it's a it's a value I think is um it meets this test I always use
30:12
for values is uh it's reverse identifiable like if I told you What's a
30:16
company that has a value like create remarkable delight? You'd probably name
30:19
three or four companies and and my guess is superhuman would be one of them on
30:22
the list. Yeah. Superhuman, linear, like Yeah.
30:25
You're just there's just not Apple, right? There's not there's not
30:27
that many, right? There's a handful where you can see it in their DNA is the
30:31
hey, we're going to make uh and and the words are very carefully chosen.
30:35
Remarkable, delight. It's you you delight is a goal. Like you want people
30:39
to have that sense of joy, that sense of glee. And remarkable means it needs to
30:43
be good enough that they actually remark on it. They actually they actually talk
30:47
about it. And I think that that's a really high bar for for delight. And you
30:51
know, so we took that and we said, "Okay, this is a thing that this team is
30:54
really really good at." And we just amplified it. We we put it into our
30:57
joint values. We we made sure there was lots of examples of it. I made sure
31:01
Rahul uh talked about it with the team, gave people some sense of what it means,
31:05
what examples where it showed up. Um and it was, you know, really helpful to do
31:09
that. So I think with each each company you just have to you have to do that
31:13
amplification. Uh and the last thing I'd say is you want to intermix a bit and
31:16
and I think the you know sometimes you you build up a company out of multiple
31:21
uh companies and you try to separate them as much as possible and I do think
31:24
there's something to the don't don't shake up too much too fast. Um but a
31:29
little bit of that cross-pollination you know goes a long way to creating the new
31:34
identity. Um and I think I think we've done a pretty good job of that so far.
31:38
Uh but we'll see. So we're getting a sense of like uh what
31:42
I would call like Shashir's bag of tricks. But um another thing you're
31:45
known for is collecting rituals. Like you study companies, you pay attention
31:49
to how they operate. Like this is something since I've known you, you've
31:52
talked about this. Um are there common rituals that you think
31:57
are especially effective at driving organizational change in the AI era? And
32:03
the flip side, are there common rituals that you think are actually no longer
32:07
relevant and maybe actively harmful in the age of AI?
32:11
Oh, that's interesting. Um, okay. So, first off, the comment on rituals, and I
32:14
I do think the term rituals is a really interesting one. I think it's uh um when
32:19
when you hear it this way, I it's it's a very sticky concept. And I think, you
32:24
know, my my history of Ritual started with a um was on the board of this small
32:30
startup with um fellow board member, a guy named Bing Gordon. And Bing is um he
32:35
was a chief creative chief creative officer at Electronic Arts. Um he's
32:39
since gone on to be a great investor. He's uh you know, Amazon, Zingga, many
32:43
of the companies he's he's been very deeply involved with. And he kept
32:46
harassing this board member with this question. He said, "What are your golden
32:50
rituals?" Um, and uh, the the CEO of this company at some point stopped and
32:55
said, "Bing, I don't know what you mean. What's a golden ritual?" And Bing gave a
32:58
really clear definition. It's really stuck with me. And Bing said, "Look,
33:01
great companies have a small list of golden rituals and they meet three
33:04
tests. Um, number one, um, uh, every employee knows them by their first
33:10
Friday. Number two, they're named. And number three, uh, they're templatized."
33:16
And each of these turns out to be a pretty essential part of these these
33:20
great rituals. And so once I heard that, I got really interested in the idea. I
33:24
started talking to everybody I could about what are your rituals? And and
33:27
actually during co we started doing this dinner series um where people would get
33:31
on virtual dinner and we shared rituals and I took those and I turned them into
33:34
I've been turning them solely into a book. I wish I was moving faster on
33:37
this. I've now interviewed over a thousand different teams companies on
33:41
and I'm taking sort of the best hundred rituals and turning them then to this uh
33:45
into this book that's all about rituals and if anybody wants to read it rituals
33:48
of great teams.com I'm publishing a chapter at a time um and uh you know
33:53
people can get a sense of lots of different rituals including a bunch from
33:56
uh from Zapier uh thank you wait for your contributions the uh um so I think
34:01
I think the concept of rituals is is is really interesting could be really
34:04
powerful maybe one last thing I'd say about the frame on rituals is um Darash
34:08
gave me Dashshai is the the this uh the co-founder of HubSpot. He gave me a way
34:14
to think about that I thought was really interesting. He was sharing one of his
34:16
rituals with me. The ritual he shared is called flash tags which is a a really
34:20
fun ritual worth looking up. Um but he also gave me this analogy. He said he he
34:24
was really excited to talk to me and he said you the reason I'm so excited to
34:27
talk about rituals is he said I think as companies we actually build two
34:30
products. We build one for our customers and we build another one for our
34:33
employees. And when people talk about the product you build for your
34:36
employees, they often give it a name like culture. And culture is a good
34:40
name, but actually if you ask people to describe culture, they will describe it
34:44
through rituals. And that and his point was that rituals are the mirror of
34:48
culture. Um they usually go hand in hand. And we put so much thought into
34:52
designing the product for our customers. We should put that much thought into
34:55
designing the product for our team. And the way to do that is to really obsess
35:00
over our rituals. So that's a little bit of uh you know maybe background on on
35:05
the concept you know in terms of uh what does it mean in the uh age of AI. I mean
35:10
I think it means something different in every different age. I mean I think you
35:13
know you guys are Zapier is known for being uh a distributed company before it
35:18
was cool. Um and I think that some of the rituals you've shared with me
35:21
started with that insight of hey we're going to be distributed. We're going to
35:25
do things a little bit differently. And you know I borrowed and mimicked a bunch
35:27
of those rituals. We now run our our pulse at the beginning of our staff
35:30
meetings. I very similarly I think to to how you showed me you did it at at at
35:35
Zapier. Um so I think there's each you know big change can change some rituals.
35:40
I also point out there's many rituals that don't change and I I think there's
35:43
a lot of rituals for which um you know the essence of how you make good
35:48
decisions or so on. There's a lot of things that didn't didn't change that
35:51
much. Um, so in terms of things that have changed in both directions, the new
35:56
things and things that we probably should do less of, I think there's sort
35:59
of two obvious examples. Um, the new thing that's happened with AI is I think
36:04
the the simplest version of it is, uh, we can go from idea to prototype to
36:11
actualization faster than we ever had before. And that just fundamentally
36:15
changes how many of our rituals work, how decision-m works, how idea
36:20
generation works. It's just your expectation starts in a very different
36:23
spot. Um, and you just expect that even when we're talking about the uh the idea
36:28
in the very first interaction, you know, we expect there to be the live
36:32
prototype. We expect that that the question of hey, can you try it again in
36:36
a different color? Can you try it again with the button over there? Doesn't wait
36:39
a week or two weeks. It happens in that meeting. We're going to see it again and
36:43
maybe that but does that look better? Um and I think that that idea of we are
36:49
working with work artifacts in a malleable uh you know concrete way much
36:55
much earlier in the process I think is really exciting and you know I I was
36:59
giving some of the analogy of when I I my first software job was working at
37:02
Microsoft and I um at the time it took us three years to ship product of which
37:07
the last year and a half was test and release manufacturing was printing CDs.
37:10
Um, and so the the the early part of it was, you know, you would come and you
37:14
would, you know, write these really elaborate specs and you would like you'd
37:18
spend all this time in definition, cover all your cases and you try to get
37:21
everything down and then three years later you'd have product. And then I got
37:24
to Google and we shipped every day and it was like dramatically different. You
37:29
know, the whole process felt different. And now all of a sudden you can go and
37:33
ship like in the meeting and you can ship at you can ship many variations of
37:37
the idea. So in my mind that's the thing that has become uh the rituals that are
37:42
going to dramatically change for the positive. The rituals that I think are
37:46
going to go away hopefully is all the busy work rituals. Um and I think
37:51
there's there's so many rituals in companies that are focused on collecting
37:55
up the status reports and the you know there's a whole movie made about it with
37:59
office space. they, you know, have you submitted your TPS report? And, you
38:02
know, we all watched it and we we laughed about it, but actually, you
38:06
know, I was talking to a product manager at this other company and they said, you
38:09
know, they counted up their time spent in a week and they were like, I probably
38:12
spend 60 70% of my time taking the same set of updates and recasting them
38:18
slightly different ways for every audience. And in the meantime, by the
38:21
time at at the end of it, they're all stale already and actually not they're
38:25
actually not that not that important. Um, and so I think that kind of busy
38:29
work is really going to go away and we we're going to be able to get a real
38:33
time, you know, on one side we're going to get these concrete representations of
38:36
what we're working on. The other side we're going to free up our people to not
38:40
do this sort of mundane uh collection process, busy work process because
38:45
that's like the easiest thing that you can imagine our AI systems doing for us.
38:49
So those are the sort of two extremes I think. I was wondering what examples you
38:54
would give and those were the two that were in the back of my head as key
38:57
changes we've seen as well too is I I can't tell you how often I'll get off of
39:00
like a call with a customer and just like feed the transcript into Claude and
39:04
be like hey can you like turn this into a PRD and then take that over to Claude
39:08
Code or Lovable or something like that and be like all right make this look
39:10
this particular way and it's just yeah you can literally go in 15 minutes like
39:15
be like all right I got something I can show you which is crazy and then the
39:18
update side of the house my goodness I can't tell you how often in In the past,
39:22
I've had to like try and get folks to be like, "Look, I don't need you to overdo
39:25
it. I don't need you to polish. I don't need like But to your point, it's like
39:28
it's so hard to do a good update. Like, it's so challenging." And with AI, like
39:32
you can get these ambient updates sort of like out of the like you can just
39:36
extract it from the exhaust of the company. It's it's it's fantastic.
39:40
Yeah. Um, last topic for you. So, you have um
39:47
a chance to like work with three pretty distinct customer bases now. Grammarly,
39:51
of course, has 40 million active users. Everyone, you know, lots of, from my
39:54
understanding, lots of students, lots of um non-native English speakers. Um, just
39:59
a huge variety of customers there. With KOD, you know, you're in a lot of tech
40:03
companies and, uh, and a lot of enterprises and, you know, superhumans
40:06
probably got a lot of or male, I should say. Now, it's going to take me a minute
40:09
to get that right. It's okay. I think I do.
40:13
Um, it's got a lot of, you know, um, you know, busy people, executives, leaders
40:18
that are sort of doing these products. Maybe I don't have the quite like the
40:21
right uh personas, but that's my mental model for these different customers.
40:24
Um what I'm curious about with this
40:26
question is when you look across these audience, how are you seeing AI adoption
40:31
spread? Like who is who is doing the best with these tools? What is actually
40:35
working? And what are the best doing that's enabling to to change the way
40:39
they work to achieve bigger results compared to maybe the rest of us? I
40:44
mean, first off, I'd say for all three and now four products, the audiences are
40:48
there's definitely a home base for each product, but they're incredibly broad
40:51
audiences. I mean, Grammarly is certainly known for its uh um
40:56
uh uh starting at foothold in education and it's about a third of of the users
41:01
are students. Um but it means two-thirds of the users are professionals. And it's
41:05
very common that we'll show up to the largest enterprise in the world before
41:09
we've done a sales call and they'll have 10,000 active grammar users. And so
41:13
there's a it's a really wide spectrum. You can't get to 40 million daily active
41:17
users without without seeing that kind of uh that kind of spread. Um you know
41:22
similarly if I look at Kota you know KOD is is very popular with tech teams and
41:26
so on but you know recently for whatever reason Kota's gotten very hot with
41:31
sports teams. Um you know so I I probably can't say which ones without
41:35
taking credit for the records but you know it's a really wide set of uh
41:40
companies. you know, the the New York Times and Snap and, you know, there's
41:44
all these different companies that are that are in different er different parts
41:47
of the the world and and same with male. Mail's very well known for the busy
41:51
professional, but actually the the the foothold in the sales audience and so on
41:55
is is just as big. But interestingly, when I looked up and part of what
41:59
motivated me to buy the superhuman mail company was I looked at our own data at
42:04
KOD and uh we had I was a very early superhuman mail user. been a fan for a
42:09
very long time and I think when the CEO use it, it sort of tends to trickle down
42:13
and half of the company was active users of superhuman mail. Um, and it wasn't
42:17
the CEO and the and the sales team and the recruiters, it was the engineers and
42:22
the designers and they just wanted that that better design product. And and so I
42:27
do think there's a um there are products that are very narrow in a function
42:31
definitionally like it just doesn't make a lot of sense to use Salesforce if
42:35
you're not a salesperson. Um, and then there's products that are horizontal by
42:38
nature, but have have a have a a wedge in their in their product market fit
42:44
journey. Um, and I think those those are really interesting to me because I I
42:48
think those are many cases when you bring those products together, you start
42:52
to stretch people's definitions and you start to see uh actually I did want a
42:57
better mail experience. And yeah, I'm not the one that sends all the sales
43:00
requests, but you know, I'm I'm an engineer and I get a [ __ ] ton of emails
43:04
about every pull request and about uh about every recruiting candidate, and
43:07
I'm pretty busy, too. Um, and so I think that that idea, I think, is is uh is
43:12
really helpful. Um, you know, in terms of what what people are are doing within
43:17
AI in each of those places. And I I do think it leads to very different
43:21
experiences, right? So, you know, for for Grammarly, it's um it's really
43:25
surprising. you mentioned a couple of the audiences, you know, how do students
43:27
use it? Students use it, they get a better grade, they get they they learn
43:30
grammar, they get a better grade, and I think it's a it's a, you know, easy
43:33
value proposition. Um, you know, ESL, uh, learners, uh, people people are
43:38
picking up English as a second language, a big audience. Um, and you know, that
43:42
audience is they're trying to to look smart in a language they're not as
43:45
familiar in. Uh, we now support multilingual, so we get the other way
43:47
around. Um, so we get people where where uh, English is their first language and
43:51
they're picking up something else and that's also really helpful. um the uh
43:56
you know I think in each case it's a it's a little bit different in KOD the
44:00
the AI use tends to be blended with other workflows um so for example you
44:04
know we did this big launch I wanted to send out a personal sounding email to a
44:08
lot of our investors and advisers uh it's a pretty big list um so I had AI
44:13
help me write all those different uh mails knowing enough about each person
44:17
what we were doing and customizing it a bit for yeah I saw you last week at this
44:21
event um and yeah I want to let you know about this new thing we're we're
44:25
launching. Uh you know for mail I think a lot of the a lot of the attention has
44:30
been on the the two hottest AI features autodrafting and auto labeling. Um
44:34
they're both really really productive features. Um autodrafting is a somewhat
44:38
obvious one. I think that's a um I think the ability to autodraft emails to to
44:43
people is something kind of you know it's designed around the salesperson,
44:46
the recruiter or so on but actually is pretty magical for just about anything.
44:50
But I'll tell you autoleing is the the the sleeper hit. I mean I I I fell in
44:55
love with that feature because it for me I actually respond to a very very small
44:59
portion of email and my guess is you're the same and so I spent most of my time
45:03
spent in email is processing uh and I think that ability to presort as I you
45:07
know I was moving my mail system from our Koda domain to our Grammarly domain
45:13
and I had over the years I developed um I think it was like a thousand filter
45:18
rules in in Gmail because you know 11 years of like every time I'm like All
45:22
right, I'll write a rule that keeps this stuff out of my inbox. And I was I've
45:25
gotten pretty good at doing it. It was a little bit of my Saturday morning ritual
45:28
to write these rules. And then I come in, I'm like, what the hell am I going
45:31
to do? I'm going to move all these rules. It's going to be impossible. And
45:34
so I went and wrote like three or four of these AI based prompts. And it just
45:38
like magically worked and I moved none of the rules. And it just it sort of it
45:42
was amazing how you could take a task like that and just sort of transform uh
45:47
the the application. And then, you know, and then of course with Go, I think it's
45:50
it's we're going to see we're early in how people are using Go. Uh but the
45:55
really surprising use cases. I mean, you know, people getting warned of uh issues
45:59
while they're writing something. Hey, that's the wrong uh that's the wrong
46:03
name of the customer. I mean, one of my favorites is how often you misspell your
46:06
own name. That's kind of a fun one. It's like, you know, you want to be
46:10
embarrassed. Uh and all the way to I mean, Go connects to everything. And so,
46:13
you know, I I was talking to one of the engineers and they were just so excited
46:16
like you know what I did today and I said what'd you do? He said I told go
46:20
can you extend this meeting by two hours and decline all the rest of the
46:23
meetings. I just went and did it. Uh and so it's you know lots of different
46:27
things that you can start doing with these products. So, you know, pretty
46:30
broad set of users and a very broad set of use cases.
46:34
That uh uh labeling example is one of my favorite AI use cases too. I have a Zap
46:40
year agent setup that does this. And to your point,
46:42
like it sifts through probably 100 plus emails that I get a day, but in my
46:48
inbox, I'm usually looking at less than 10 at any given time. Like it's a pretty
46:53
small number of folk because it just turns out there's just a bunch of stuff
46:55
that comes in that it's not necessarily for me. It's for
46:59
something else. Yeah.
47:00
And I I bet that zap went from being a bunch of structured rules. So you stuck
47:04
a little AI prompt in the middle of it. Yeah. It's pretty simple. Yeah. And then
47:07
I can go tweak it if it doesn't quite get something right and, you know, or
47:10
not. But it's, you know, it's it's like, hey, if someone looks like they're
47:13
looking for a job, like toss a hiring label on it and forward it to these
47:16
hiring teams. Like it's pretty natural language, which is incredible.
47:21
I mean, that's also an interesting observation. I think Zap Zapier and Kota
47:24
share that sort of view of, you know, AI is even more powerful when it can
47:30
connect to your tools and it can be an, you know, the when the actuators are
47:33
there. It's I I kind of envision it a little bit like, you know, the the the
47:38
robots are, you know, interesting. you need the actuators, you need them to
47:41
actually be able to do things. Um, and I think that's a it's a similar AI
47:45
philosophy. I think the metaphor we use a lot internally is
47:48
that um, we have, you know, AI like whether it's, you know, chatg or cloud
47:53
or Gemini, like this is the brains, but it needs arms and legs
47:57
to like go do stuff. And so that's where, you know, providing it access to
48:00
tools and uh, your the apps you use, the surfaces you spend your time in is like
48:05
so powerful. That's right. Yeah. Absolutely. Um, last
48:08
question for you. What advice do you have for leaders who
48:13
are out there trying to change their culture, trying to get the teams to
48:16
adopt AI more, trying to get them to shift from one way of being to another?
48:21
You've, you know, obviously had a massive change management exercise over
48:26
these companies. And I'm curious if you could leave them with one advice on uh
48:29
what to do next or or how to best navigate this moment, what would it be?
48:36
You know, I I feel like I get this question a lot. And the the um uh
48:42
I I think there are sort of two different kinds of AI that we're seeing
48:45
really matter in in people's workplaces. Um and there's the supervisible and the
48:51
super invisible AI. And I and I end up uh and I'll I'll I'll talk my own book a
48:56
little bit on this, but you know the the the super visible thing is probably
49:00
close to what we were talking about with rituals is the you need to take every
49:04
you need to run a hackathon and teach every single person in your team that
49:08
they can take whe what whatever it is it's called code it's lovable it's
49:11
cursor so on and we just did this and we took our recruiting team and had them
49:15
build apps and it was just they were just shocked they could rebuild Koda
49:19
they could rebuild Grammarly and it was like literally people that had never
49:22
looked at or thought about a line of code had a functioning prototype of a
49:26
dream thing that they wanted and it it's it totally changes your perspective. So
49:30
I'd say you got to unlock people's people have a mental block of I'm not a
49:34
builder and you got to unblock it for them and you got to give them give them
49:37
that chance. And so I think that's really important and a lot of people's
49:41
heads go there with with hey I want my my team is AI native and you the reason
49:45
I know is because everybody can build prototypes but then I think the other
49:48
piece is what what I call the invisible AI. If you go to the other side of those
49:51
rituals is you got to deploy the tools that give people the ability to to have
49:56
blend AI into the workflow. And you know that's that's certainly what we do here.
49:59
I mean I think we build you know Grammarly Koda Mail and Go are all you
50:03
know the they're they're all built to be that horizontal AI where the little
50:09
things you're doing are all pulled together. And you know Grammarly people
50:13
don't even think about as an AI provider. We do Gramly does a 100red
50:16
billion LLM calls a week. it would be a top, you know, four or five AI provider
50:20
in the world um if we were if we were sort of judging that way. But it's, you
50:24
know, these things are they're sort of they're just built into the fabric of
50:27
how your how your team works and you got to embrace and allow all those tools to
50:31
to get deployed and spread as well. So maybe two too extreme for what I would
50:35
do. I love it. That's it for today's episode
50:38
of Agents of Scale. Shashir, thanks for coming on and congrats on Superhum.
50:44
All right. Thanks, Wade.