How Business Software Transforms In The Age Of AI — With Zeb Evans

Channel: Alex Kantrowitz

Published at: 2025-10-10

YouTube video id: B-uXIjDcRg4

Source: https://www.youtube.com/watch?v=B-uXIjDcRg4

Let's talk about how AI will change the
core of business software with Zeb
Evans, the founder and CEO of ClickUp in
a YouTube exclusive brought to you by
ClickUp. And I'm thrilled to be here
with Zeb today. Zeb, great to see you.
Welcome to the show.
>> Excited to be here, Alex.
>> So, I want to start with you um talking
about user interfaces. I was having a
conversation with an investor yesterday
about how AI has the possibility to
change the way that we interact with
software. So the way that this investor
described it was basically um if you
think about our desktop, we have a mail
app, we have a messaging app, we have a
calendar app, we have a browser. Uh
these may all have their own separate
databases and none of them talk to each
other. And I think in business software
that's like really uh the case on
steroids. Of course, of course there's
APIs and connectors that way. Um but
business software can often fail to be
smart because that data and that
experience is just not synthesized at
all. And I I wonder if generative AI,
you know, we're seeing the beginning
now. Everyone's talking about automation
and agents. Um, I wonder if the next
step is really going to be a change in
the way we interact with the software
completely. Uh, because instead of
having maybe the AI go and uh use the
software for you, it's possible that,
you know, the the um the actual way that
you interact with these uh software
programs is going to be very different.
You know, someone building something
like this yourself. I'd be curious to
hear your take.
>> It will definitely be very different. Um
although I do not subscribe to the
belief that everybody's going to want a
single chat interface as their only
interface uh within B2B software. Um,
and I actually think what what will
happen is every AI chat experience will
need core primitives. And what we refer
to primitives are things like a doc,
slides, spreadsheet, kind of database,
um, and of course attachments and files.
And those things will need to be built
by the AI chat providers themselves. And
we're actually already starting to to
see this. Um so our philosophy and
strategy really has has always been that
software would converge that previously
very separate um niche vertical use case
specific um software categories would
converge into to one uh and that's
certainly what we've we've been focused
on since the beginning. I think that
what will change in the the near future
and something that we're we are working
on is what we call personalized
software. Um and this is a very you know
it's a collaboration software. So you
have of course you have consistency
across the data models, you have
consistency across the collaboration um
experiences and the primitives, but your
surface areas for your software, your
homepage in ClickUp, your inbox that you
come to, your task list um will be
extremely different um based on which
user you are, based on what team you
work on, based on your current
priorities. Um and you could think of
this what we think about it as is is
almost vibe coding your front-end
experience um for a particular page or
feature uh within collaboration products
and I think that is is very much the
future especially as we start to see
more software converge right how are you
going to be able to put uh 10 15 of your
previously separate apps inside of one
platform without it bloating the user
experience and without it being
extremely complex. it will need to be
very personalized to just the things
that you particularly work on.
>> Okay. I want to ask ask you a question
about something you said at the
beginning of that answer, which is that
you don't believe that we're going to
end up having software that uh basically
condenses everything into one platform.
Uh why is that?
>> So I do believe that we'll have software
that condenses into everything in one
platform. That that is absolutely what
we're focused on. What I don't believe
is this notion that you're going to use
just AI chat experience for all of your
software needs and that you won't
actually need UI um or application
layers to to back that up. And you know
we we I think as as humans you always I
mean take chat for example right
chatting team chat you are always going
to want a consistent experience where I
can go see the history of all my
messages regardless of how I send a
message or receive it. you're going to
want a a DM that you um can go visualize
that that history of your communications
with a given person. And that same thing
is true for all of those those core
primitives. Um so I certainly don't
believe, you know, that that everybody's
going to want just to have a single AI
chat all day long and they're not going
to want to see any application layers or
experiences. Um but I do believe that,
you know, these separate software
categories will all converge. I think AI
will um ultimately also converge with
with software.
>> So there was this sort of semi famous
Benedict Evans tweet or thread uh
whatever it might be at this point uh at
the very beginning of the chat chase and
his perspective was um don't think about
how you can layer AI on top of like a
Microsoft Excel. Think about how
basically it changes Excel completely.
And I think where he was going with that
is basically that business software or
something like a Microsoft Excel um
would be instead of like you you know
putting your numbers into a spreadsheet
and then uh doing calculations on them,
you just upload all those numbers into a
chat interface uh and then ask for
insights. Um that's not how you see it
going. So, you're basically saying we're
going to need some of those uh initial
programs, but uh AI will make them
smarter or tell me exactly where you
land there.
Oh,
>> look, ultimately you need to to have it
all. Um and I think this is, you know,
it's very timely because you start to
see Grock add spreadsheets. Um you start
to see chat GPT also added Excel
spreadsheets and CSV embedded inside of
that that chat experience. Uh and you
know more importantly right now AI we're
in a world where where AI is not good at
calculating things yet. Um so generally
you're having to spin up a virtual
machine and run Python or something on
there uh to actually perform those those
spreadsheet calculations. So either way
you have to build both sides of it. So
you can you can start from this the
spreadsheet um and you can add the
question the chat on top which is what
Jim and I did. Um you also can start
from the AI chat um like chatgpt or gro
and then you can add the spreadsheet on
top of it. Uh but ultimately you you do
need both of them.
>> Okay. And so talk a little bit you
referenced ClickUp obviously that's the
company you run. Talk a little bit about
what you're building there. You
mentioned that there's going to be you
know personalized software for people
that are using it. Uh how does how does
that work?
We try to give you the best experience
based on your intent. Um, and intent is
is very tricky to figure out sometimes.
If you have enough data, of course, it's
very easy. So, if after a ClickUp user
is using us for a week, two weeks, we
really start to be able to personalize
that experience uh pretty much
automatically for them. They can give
inputs. They can they can refine how um
things are prioritized and visualized. I
am experimenting right now of viewing my
inbox uh like a newspaper. Um and you
know like an old school kind of news
newspaper. We have a a branch that I was
just trying that out on. Um and so it I
I do think that it gives you a new level
of like creativity and control as an
enduser that just doesn't really exist
today um inside of at least inside of
B2B software. I think you can go vibe
code something that's consumerish but
you struggle with adoption um across
teams. Uh so we have always believed in
horizontal massive horizontal platform
primitives that are very flexible uh
that allow you to accomplish any given
use case any vertical use case uh and
really any core work management or
productivity need inside of one platform
uh because you know similar to I think
what is getting a lot of attention now
with AI is like underneath um the hood
why aren't all of these primitives the
same you just have data models you have
a database with a certain schema. Um,
you have services for infrastructure,
you have websockets, etc. Um, all of
those things are are are very much
needed and and leveraged appropriately
for AI. Um, but they're also needed for
any given they're common core components
in any given SAS application that that
you have. So, we've always taken the
approach of build that very flexible. We
call it convergence. Um, the only other
company I'm aware of that that had a
similar strategy is Rippling. They
called it compound software. Um but it's
the same the same approach and it just
enables significantly greater
flexibility in use cases and and also
velocity in shipping product. So then
talk a little bit from an end users
perspective how they would go through it
dayto-day.
>> So let's take the inbox for example.
Inbox in ClickUp is kind of a universal
inbox for all of your notifications,
your action items. Um, you can send
events there, emails, and you can
imagine typing, I essentially vibe
coding your inbox and saying, "Hey, I
want these things to appear first. Um, I
want everything about AI agents to
appear at the top. I want all of the
urgent things that mention me
prioritized in order one through five
um, in a second section. And instead of
it looking like a traditional inbox, um,
I'll go back to that newspaper example.
make it look like an old school New York
Times news newspaper. Um, and so it just
gives you this this kind of like
limitless creativity in in how you
consume um your notifications, consume
your your your inbox. And that example
can be relevant to really any given
surface area, any any front-end
application layer. You should be able to
really deeply customize that experience.
I think you can go too far um if you
don't add constraint and it becomes uh
you know AI is look I'm total optimist
in AI but it's not there yet in a in a
lot of ways and it certainly is is also
very it's very costly um it's pretty
pretty slow and it's ripe for errors if
if you don't give it a bunch of guard
rails um so we are taking the approach
where we will give it opinionated UI
components and and framework to build on
top of um which minimizes issues that
that would be created um otherwise if
you you let it do essentially anything
>> and so people can customize with natural
language.
>> Exactly.
>> And are you you're sitting on top of
email um is it its own email server or
is it something that sits on top of
email and various different programs
>> we synchronize in in real time um pretty
much any application outside of ClickUp.
So you can connect Gmail or Outlook um
and we will in real time bring in all of
your emails.
>> Okay. And so then people can basically
prompt their own not only UI but
dashboards as well. We'll be able to
give them new data uh and ways to look
at the data that's coming through their
platforms.
>> Absolutely. That's that's the uh the
power use case you you hit on is is the
dashboard uh because it can be used for
reporting, right? Those are the obvious
things for reporting and analytics, but
we're seeing some really cool use cases
of of people almost like creating their
own like apps as dashboards. Um because
it can do more. Our dashboards can do
way more than reporting. Anyway, you can
have a chat. I can have a dock in there.
I can have um tasks. I can have a
whiteboard inside there. You can have a
spreadsheet. So, you imagine all of
these like primitives, right? Going back
to that that primitives analogy, it
being created with AI and and if you
give AI these these core primitives like
a spreadsheet uh plus a doc plus these
widgets that are able to show charts and
really embedded anything inside of it,
then you can create pretty cool
applications and pretty cool use cases.
Uh we saw a a customer create a a
bespoke like CRM uh solution that also
had their clients have a a very
personalized login experience that feels
like their company. Wow, that's
fascinating. So where do you think uh
traditional business software is heading
right now? I mean obviously there's like
a lot of talk about um you know does
business software collapse into sort of
this singular user interface or um can
you replace a lot of it with AI
solutions. I'm curious if you think
that's if you think that's the case and
and how that uh how these these hopes
might stack against the reality today
which is that a lot of companies are
struggling to uh even implement some of
the basic use cases.
I look I think we're we're very early
still and things don't change as quick
as everyone thinks that they do, right?
Everyone thinks it's o overnight
everything's going to be blown up. Uh
but I I don't subscribe to that. I do
think a lot will will change to be clear
and and I do think that software
businesses that have a very simple
product offering um are in trouble for
sure. You can argue what will happen
with software, but I think everyone
agrees that you will be able and you
already are able to build software much
more efficiently than we could in the
past. Therefore, every company will
build more software, right? Not less.
You're going to build more software if
it's more efficient, if it's easier to
create new products, easier for you to
go into different markets. Of course,
you will create more software. Um, so
that goes along with our thesis of
convergence of software. that software
will converge ultimately given a long
enough period of time. Uh all of your
software will be provided by one
software provider. I believe one AI
provider also um and I don't think that
there will just be one of them as far as
monopoly goes. I think that there will
there will be several not that many um
but several. And then along with that
you will have this whole new category of
what we call bespoke software where the
small business that previously was not
able to afford their own customized
mobile application now can afford that.
We don't believe that uh those people
are going to go create their own though.
You know I think that that's that there
is definitely that notion of software
gets eaten because everybody's going to
go build their own software. If that was
true, everybody would have built their
own website by now. I mean, you've had
website builders for for decade plus,
good ones, really great ones. Um, but
that is not the case with the vast
majority of especially small businesses.
They're not creating their own stuff.
They're usually using an agency. They're
usually using somebody that's using
Canva or that's using a website builder
them themselves to go do that. Um, and I
think that that's what will happen in
this next age. also
>> uh for a unit software I guess that's
what you would call it right this one
software to system to rule them all
couple questions for you what does it
look like I mean are you able to does
the ERP and the uh sales database and
the marketing database just live in the
same product um and and then is this
something that is going to be built by
the incumbents like is this something
you already mentioned openai and grock
uh embedding uh spreadsheets. So, is
this something that's going to be built
by these big foundational research
houses that need to turn profits or do
you think it will be something that
others build?
I tend to think that the giants will
build offerings
uh but it will take them a long time. I
think also that the AI labs um will
struggle cracking into B2B especially
with collaboration. um that is like a
you know a core block in AI right now of
AI chat it's it's your own chat right
you're chatting with your own stuff and
that's that's it there is no like
element of collaboration um and the
context is also missing in in in the AI
labs today but they know that right and
and so that is is certainly part of of
their strategy in the long term um but
yeah I do think to answer your question
yeah I think that you will have one
software provider that has ERP system of
record it has all of your customer
system of record. It has your marketing.
Um certainly has HR and project
management and work management along
with team communication.
>> Okay. But here's here's where it gets
interesting. Um and that is that some of
these companies uh everyone owns like a
little FFTM or a little piece of the
puzzle. And what happens if the
providers end up shutting it off and
then making it impossible to build this
single software? Here's an article that
you wrote uh in Fortune I think today.
I'm the founder of a $4 billion software
unicorn and I see data wars coming.
Enterprise AI beware. You talk about how
Salesforce for instance has cut off
access in some ways to data from Slack.
Yeah. And but I I over time I think that
that is is is the reason why you'll see
other providers pop up and have all of
the offerings in one place. Um it it's a
short very short-sighted move um you
know to to cut off your data access
completely uh especially for something
that is is not truly competitive right
it was more about unifying their data to
provide AI value um so what will happen
is when context becomes valuable it
already is valuable in my eyes and
anybody that I think has worked with AI
deeply context is really the most
important thing but as consumers and as
businesses start to realize how valuable
it is and that AI value um actually
becomes a a reality with that context,
they will strive and demand all of that
context together in in one place. Um and
so if you know Salesforce doesn't offer
that um and they don't or they don't
want to buy from Salesforce or another
competitor comes along and does offer
that that's where I think you start to
see um you know these single software
providers take off.
>> So is this basically the Salesforce's
move you're saying? is its attempt to
become this sort of central uh unit
software. I mean there I was just
talking about sales data. So sales data,
Slack data. So it's it's what you think
the move that they're making is is
basically saying we want to control and
be that singular software player and
we're going to prevent others from doing
it. So we're going to limit access from
Slack Slack.
>> I think that is is part of the strategy
long term. I think that immediately
though it they see it as a risk. I I
think that there's there's there's no
way of getting around that that they see
it as as a big risk. Um that they have
customers that are exporting all of
their data and they're using all of that
data outside of their core application.
They have no engagement anymore, right?
Then therefore they really are just a
database. They just are that that system
of record. Um and certainly it would
make it a lot easier for customers to
switch to another provider, a single
software provider. Whereas Salesforce,
yeah, they will try to provide all of
the software, but all of their software
um is really buil built in very much in
isolation. it it does not play well with
each other and and it's not great um for
what I see as the future of software uh
where you really have to build
everything with the same application
primitives the same framework um and and
the same modules the same components uh
that's where you get a lot of leverage
from and you're not going to get
leverage from you know having five
separate applications even though
they're sold by one provider uh you
don't really get the same convergence
leverage and efficiency and productivity
ity and AI context value uh from that
>> the whole value of AI that people have
talked about is the fact that you don't
need everything to be sort of primitive
primitivized and structured and you can
take unstructured data from various
systems that don't speak to each other
and sort of use AI to extract that
synthesize it make sense of it and then
let you act on it you don't buy that
>> no I do buy that from a so from a data
perspective, it's it's different versus
the front end the front-end application
layer p perspective, right? And there's
they're they're really two very unique
things. Um because you have the back-end
platform as the databases, as the data,
the the context there, but then you have
the front-end applications where all of
the engagement happens, where the humans
actually work, right? Where they see and
visualize that stuff and where they
engage, where they collaborate, where
they actually do the do the work. Those
two things are are needed uh to provide
the greatest AI value. You need both of
them. You need the context. You need
that data. Uh but you also need the
engagement. You need that that human
that is actually using that and
providing really really valuable
feedback. This is how we are able to to
really make incredible strides with the
quality of our AI and the value is the
human feedback and and these these loops
that start to happen autonomously um
where people downvote something and we
learn from that automatically. We have
an agent that goes and and fixes it
really and adds an eval for it. Um so it
continuously gets better and better and
better. U but you certainly need that
like unified data in in a very
digestible way. So it's it it is to be
clear I don't actually think that you I
I believe to certainly that all the data
in in one place is in similar data
models is 100x more effective than
trying to unify five 10 different data
models and databases through an API and
you can see this how it works I mean if
you go connect MCP um or even the
connections from chat GPT you're very
very very limited on the data that you
can access. It sounds cool uh but you
know I was I was I was connected my
Google calendar uh the other day to
claude and and to open AI and I was like
hey what are my events for next week?
I'm limited to maximum of 10 events, you
know, hey, can you create a calendar
event? I can't create a calendar event
yet, you know, like it's it's just like
it's very surface level. What you
actually need is to synchronize all of
that data and have optimized AI
retrieval for for that data so that you
can get the right context out because
there's way too much context there uh
for AI to be able to actually like
decipher what matters. You need to be
able to to first limit that and then
give it to the LLM. And that's that's
very difficult to do,
>> right? I mean, that's been the problem
with uh Apple Intelligence, I think, and
some of the Alexa Plus delays that we've
seen is that when you ask these bots to
go in and find some information for you,
uh if I'm reading you right, if I'm
hearing you right, the answer is that
they just are dealing with a con, you
know, just an unbelievable amount of
data. They're struggling to pull it out.
Um, so how do you how do you then go and
and put that infrastructure into limit
and then find the right stuff
>> from a data perspective? You know, by
the way, we acquired a company a few
years ago for enterprise search and the
infrastructure and platform that they
built was allowing us to synchronize any
application in real time and store that
data in a universal database. So, a data
model that is consistent across the
board regardless of what application it
is. We can plug into HubSpot and
Salesforce and Slack uh and your task
management platform if it was outside of
ClickUp and we can synchronize that into
one database in real time with
permissions and privacy aware. And this
what is what we optimize for AI
retrieval. So, our our agents understand
this this singular data model. It can do
semantic search in one place. So it
doesn't have to go kick off 10 APIs and
then try and pull that back together uh
and then do re-ranking. It's very slow.
Um and it just doesn't work at scale
using using APIs to to do that. Uh so
certainly this this is where the world
will will be headed is is actually
unifying that data into like a single
database. And I do think that's part of
the reason why you know Salesforce
started putting up those those walls um
is because they they do want to be that
provider. They want those those that
database that full universal data model
inside of their ecosystem, not somebody
else's.
>> And so, how do you adapt if they put the
walls up like that for the Slack data
for instance?
>> Well, it's been a it's been a tailwind
for us because and and I think it will
continue to to be because we have a our
team chat product which is our fastest
growing product by far. Um, and it's
accelerated the the growth ever since
since did that. there is a you know that
I I would say it's more early adopters
that care about it right now that see
that oh holy like I want to get
away from this you know I need to be in
more of like an open system um and and
they they started you know switching to
our our our ecosystem um and I think
over over time as people see more and
more value from context and team chat is
extremely valuable context right the
most up-to-date information generally is
going to be there as people see uh that
value is is there. They will demand um
more ecosystems that are friendly and
allow you to pull in your data and and
kind of use it use it freely without you
know being handcuffed to a single uh
Salesforce provider if they don't even
have the other offerings. I think that's
that's the the difference here too is
you we have a lot of a lot of customers
that yeah they they use Salesforce and
they use ClickUp for everything else,
right? they have Salesforce as their CRM
backbone and ClickUp for for everything
else. Um, and so, you know, Salesforce
doesn't have a ClickUp offering, right?
So, so they can't even to your point ear
earlier like, yeah, maybe they'll get to
that world where they do have an
offering for everything, but they're not
there right now. So, what they're doing
is is they're forcing customers to
separate their con their context. There
is no way to unify the context. Even if
they want to pay a million dollars, you
can't really do it in Salesforce today.
>> Interesting. So when in the early going
my thought was maybe you would have um a
set of companies that are databases and
then a set of companies that might you
know draw from those databases and be
the user experience on top of it. But I
think what you're saying is
>> it isn't going to be two separate ones.
Like the only way to make this work is
everything is in the same one together.
>> Yep. Okay.
>> Exactly.
>> Figma's doing it too. Why is Figma also
it's one of one that you've listed as
well? Yeah, you know, they are and and I
look, I I think if I was them in in
their case, it's it's hard to say, you
know, is it the right decision? Is it is
it the wrong decision? Um because in in
Figma's case, you know, they do have
that engagement layer, like where people
are actually working. They also have the
platform, the data platform layer, like
the source of truth of designs and the
very detailed specs along with that.
It's not just a screenshot. they have
all of all of the actual components
within the wireframes and yeah they they
started locking it down too where where
you know we were previously uh
synchronizing all of your Figma
prototypes and you could search across
them and you know you could do AI across
the the prototypes um and recently they
uh notified us that uh you know they've
they are deprecating that across the
board not just not just for us for for
everyone um and so that will you know
that will will be a bit of I actually
think that it does it will create
problems where consumers are pissed off
because they're using their favorite
vibe coding tool right now and they're
trying to plug it into Figma and when
that access gets cut off you have to go
use Figma's product right which is
imagine I imagine that is is kind of the
intention here is to go use their Figma
make product for for vibe coding rather
than using something else o over time I
think maybe over the next quarter two
quarters is that that's what we'll we'll
Um so they are certainly headed towards
that same direction though that
convergence and and trying to get single
you know provide a single offering
platform that that provides nearly all
the software for their for their given
market which is you know designers right
now.
>> Yeah. It's interesting because your
headline was you know there's going to
be data wars and I was like oh is wars
too strong of a term but no really is
it's going to be like real business wars
here.
>> Absolutely. It'll it will be it will be
it will be wars for sure. Um, and I
think you're you're going to end up like
waking up every day and you know see I
think you'll see another headline of
this company cut off this access to this
access and you're using you know your
your tool that you thought you were
using to control everything um connected
to it. But but that those connections
will will certainly stop I think as as
convergence continues to happen.
>> All right, Zeb, I I want to end here. Uh
there was this MIT study that I'm sure
you've heard of it. 95% of businesses,
I'm getting tired of repeating it
already because we've talked about it so
much. 95% of businesses get no ROI on uh
their their AI spend. Let's not talk
about a study. Um I'm actually curious
to hear how you think about uh
calculating ROI uh on enterprise AI.
Like what is the right way to calculate
ROI on enterprise AI spend?
time saved is is what we care about and
that is easier said than done to measure
it. U but directionally you you can
start to measure efficiency. You know,
it's going to sound like I'm selling it,
but it goes back to you have to be able
to see all of the context. You you need
to be able to see everything in order to
measure efficiency. Any given missing
context just throws all of all of the
data off. But I would say that high
level, you know, I'm not surprised at
all. What we've always focused on is is
the engagement piece. Everybody can, you
know, go take the latest most complex
agentic platforms and build something
cool for an agent, right? A cool use
case for an agent. But what matters is
humans actually using that agent, right,
at the right time, the right place um in
in the right way. And I believe that
ambient AI is is the future here. And
this is where we started to see the
value really cracked. You know, of
course you can ask brain our AI product
any question about, you know, your work
or outside of your work. Great. But it
requires the human going to AI to ask
that question. And humans default to
asking other humans questions, right?
We're just it's very hard to change
behavior. And so what we started doing
was every time a human asks another
human a question, we just ambiently sit
there and see if we can answer this
question with 80% confidence, we will
actually just answer it. We will
intercept it and answer it
automatically. And we've answered over a
million questions automatically without
a user asking AI. They're asking an
actual human human to do that. That's
where I see the the real value coming
from is is where AI is really just
providing the value automatically. um or
asking you and confirming you that they
to to provide the value, but it's just
doing it there. You shouldn't have to go
invoke AI to realize the value from it.
>> More proactive AI.
>> What you're proactive AI.
>> Okay.
>> Uh Zeb, if people want to learn more
about ClickUp, where do they go?
>> ClickUp.com.
>> Okay. All right. Well, you heard it
here, folks. clickup.com. Zeb, thank you
so much for coming on the show. Great
speaking with you.
>> Thanks so much for having me, Alex.
>> All right. Thank you everybody for
watching. We'll be back on the feed
soon.