Connor Jeffers, CEO of Aptitude 8, interviews marketing, sales, and customer success leaders about how they are using artificial intelligence to innovate, optimize, and scale their go-to-market operations.
Imagine I'm a brand new user or I'm trying
to do something new, even in HubSpot.
"Hey, how do I do this thing?"
And instead of that requiring
me to open up a knowledge based
article, or poke around the app a
bunch, watch some academy content.
Instead, what if it was just done for me?
And I could just focus on the results,
and the outputs, and not learning the CRM.
Or figuring out how my property
structures are formatted.
Instead, just focus on the outcomes and
the things you're looking to do to grow.
Hello and welcome to Go to Market with
AI, a podcast for sales, marketing,
and customer success leaders using
AI to scale their growth operations.
I'm your host, Connor Jeffers, and I am
joined today by Dylan Sellberg, director
of product AI innovation at HubSpot.
What's up, Dylan?
Hey, how's it going, Connor?
Great to be here.
Likewise.
Good to see you.
But before we get into the very
exciting product AI innovation.
I've known you for a bit, and
prior to this morning, I was
racking my brain to try to remember
how we first got introduced.
Because I feel like I tell people,
with some degree of frequency,
of like, "Oh yeah, how I got
into the HubSpot universe."
And I was like, "Then I met Dylan,
and he was doing cool stuff," and I
legitimately do not remember how on earth
we originally got connected, do you?
It was one of two things.
It was either when we were working
on field level permissions.
Locking down properties, before we even
had the view level, the field level
permissions edit or it was in the
earliest, earliest days of custom objects.
And so that would put us
pre COVID just about, yeah.
Back in like the HubSpot CRM dinosaur
ages because things move so fast.
It would put us around those days.
BC before COVID, before custom objects.
Before custom objects.
HubSpot BC.
Yeah, yeah it was, it
was those days for sure.
For anyone who's here, you have a, in
my opinion, like a super interesting
HubSpot background and career, because
you have worked on most of the things
that I love that attract me to HubSpot
and got me to work with HubSpot.
If you want to start at the beginning,
you can, but give me the very, like,
we'll get to product and AI innovation,
but what's the origin story here?
Yeah it's been a while all
in, really thinking about it.
I started HubSpot in 2016 as a
consultant implementation specialist.
I spent about a year helping
customers get started with the
HubSpot CRM, HubSpot product.
At the time, that was when
actually the CRM didn't even exist.
I remember my first day sitting
next to my teammate, Joe.
And he was like, "Hey, did they teach
you about the CRM in new hire training?"
And I was like, "No, I don't think so."
And they were like, "Oh yeah,
cause like we launched it today."
And so, I spent about a year
doing that and moved over to
the product side of the house.
Worked on Facebook Messenger,
which is a really fun.
Associate product manager product
took that from zero to one.
Helping, this was like in the
earliest days of messenger bots.
I didn't know that I would like naturally
end up spending so much time thinking
about like conversational CRM at
that point, but built messenger bots.
This is before conversations.
And so we kind of like wound
down that project a bit.
Moved over to the HubSpot CRM.
And I spent a lot of time helping
customers store and structure
data inside of HubSpot working on
import, properties, CRM setup, a
lot of the admin side of house.
From there, shifted over to work on the
calling product group for a little while.
And finally came back around
to HubSpot to work on ChatSpot.
Well, you, so you bounced out a
HubSpot and then, and then something
really exciting pulled you back in.
And so, we don't have to talk too much
about what was the temptation out.
But you were like, "Hey, I built
a bunch of cool stuff at HubSpot.
I'm going to go build
cool stuff elsewhere."
And then you boomeranged back.
It was a very short boomerang.
But I mean, the same reason
out was the same reason back
in, I was just, I was hungry.
I was looking for something exciting,
something new something challenging.
And so, when the opportunity arose to
come back and work on ChatSpot, which
I had been following since the day
Dharmesh launched it on March 6th, I was
super excited for, the potential that
has for, not only HubSpot customers,
but CRM and the industry as a whole.
When that opportunity came up,
I was like, "Yeah, let's do it."
So hopefully anyone who is
listening to this or seeing this
is like, "Oh yeah, ChatSpot.
I know what that is."
Assuming maybe they don't.
What is ChatSpot?
ChatSpot is the conversational AI
assistant for using the HubSpot CRM.
And so, it was developed by
Dharmesh outside of HubSpot.
Basically on the tail of the
generative AI boom that ChatGPT began.
So under a year old type thing,
Dharmesh saw what folks are building
over at OpenAI and said, "You know,
this, it would be nice to be able
to talk to your CRM in such a way."
And we tried this at HubSpot with
GrowthBot back in 2015, 2016, and
the technology just wasn't there.
And so, with LLMs and the power that
we have like now been bequeathed
through OpenAI, the opportunity for
ChatSpot was real and it was here.
And so it helps you talk to your CRM,
helps you sell better, market better, and
use the entire corpus of data that you
have using nothing but natural language.
It's It's quite literally
like talking to a friend.
It's that easy.
So when you think about what you're
building with ChatSpot, and you and I had
had this really interesting conversation
several, several months ago at this point,
which feels like an eternity in AI land.
But when, when you're thinking about what
you guys are building a ChatSpot and who
ChatSpot is for, are you thinking about
it as primarily, this is something that
we're building for HubSpot customers to
be able to interact with HubSpot better?
Is it something more
or different than that?
I think we have, we have the
opportunity for, two things.
We have the opportunity to
help HubSpot customers use the
HubSpot product in a much more
enjoyable, much more powerful way.
But we also have this opportunity,
just like we had with INBOUND,
and any other sales and marketing
emerging tech, to teach non-HubSpot
customers, to help non-HubSpot
customers get value for their business.
To help them grow, regardless of
whether they're using the HubSpot CRM.
So our early passes with ChatSpot
were pretty broad, right?
The things we built in the areas that
we focus, the use cases that we were
targeting were both applicable to HubSpot
customers and non-HubSpot customers.
Can you give examples of like,
what's the initial stuff that you
guys were building for ChatSpot?
Where you're like, "What if
we did something like this,"
and how you thought about it?
Yeah.
You can think about prospecting
use case, for example, right?
Like we took natural language processing
and figured out, "Okay, so we have
something like access to Clearbit
APIs, and we can take natural language.
We can marry a customer's request
of, "Help me find companies,
help me discover prospects in a
given area, or technology use."
We can use OpenAI to translate
that request and turn that
into a Clearbit request.
And format that back as if you're having
a conversation with Clearbit data."
Right?
And we, so we did that
in a few arenas, right?
We did prospecting with
Clearbit, image generation.
For a small, very short amount of time,
we're really excited about multimodality.
You could do images and chat inside
of ChatSpots before ChatGPT had that.
Before Bard supported it.
we were super stoked about that.
The world moves fast.
Right.
You're right.
You're like, "Oh, look
how different we are."
But things like that, right?
Like really putting that, that business
application layer on top of generative AI.
Not to reduce this because
the scale of ChatSpot is big.
Like, how many, how many
ChatSpot are you at right now?
In total, we are over 150,000 signups.
We're not talking about like, and
when I say it's a, the original
impetus was, "What if there was
sort of a, business and sales and
marketing wrapper around ChatGPT?"
Before ChatGPT was like, "What if there
was an everything wrapper for ChatGPT?"
And so, this was the very initial love.
We could use this in really
cool, exciting, interesting ways.
Yes.
Yeah, precisely.
So outset is, "Let's build this thing.
Let's try to solve this problem
for some of these people."
Where have you grown to?
And if there's a multi milestone
there, feel free to take it.
Well, we, saw that working and there's
still a lot of opportunity in that space.
When you kind of look around at what's
happening in the industry, you kind of
like see two really strong undercurrents.
You have something that users
are just gravitating to,
naturally, without any business.
Which is sharing, and this is all over
LinkedIn, it's all over Twitter, it's
all over medium and anywhere you go.
The business applications for ChatGPT.
Here are the 10 prompts
you need to do to do X.
You know, how's everybody
handling X in ChatGPT, right?
Like you have that happening and users are
like kind of pulling in that direction.
So we have that opportunity with ChatSpot
to just put that layer on, right?
Like you shouldn't need to copy and
paste these mega prompts and understand
like advanced prompt engineering for
really business specific use cases.
ChatSpot can use things like
templates and these libraries
to just take the work away.
Instead, you just play Mad Libs and we
have like the religion and the prompt
engineering for business applications.
Helping customers grow
and make that really easy.
So you have like that trend.
Can I ask like a philosophical
question about what you just said?
The distillation and saying,
"Oh, well, we're just doing
Mad Libs and it's on top."
Right?
And is that, in your mind, is
that a simple framework of like,
"Oh, we'll just help these users.
And we're like skipping
a couple of steps?"
Or does that, is that like a UX
layer, on top of the database itself?
If that comparison makes sense.
I think it's, it's a bit of both.
Okay.
At the end of the day, it's,
also like a training moment for
users to help understand, right?
Like how to, how to use an LLM.
But do you like the comp that I almost
make in my head from what you just
said, is like most business software
is basically spreadsheets with extra
steps, like, more or less, right?
It's like, we're going to
build some UX is we're going to
automatically create these tables.
We're going to make it easy for you
to navigate between these tables.
And I think you could almost argue
that like, by using a CRM, we're almost
teaching you how a relational database
works and like why they're cool.
Do you think that that's a, like, I
guess the parallel for me is on one layer
where we're like, "Oh, it's just Mad
Libs and like, we're maybe helping the
user not have to copy paste something."
But it's almost this meta level above that
of, that could be said of, all software
is, it's just a layer between the machine
and the user, and the product is just
helping the user use the base machine.
That's exactly what it is.
Yeah.
That's how I think about that.
I think over time it becomes less of
a layer, less of a force layer, right?
Like you, you see a lot of
these very similar applications
popping up all over now.
And they're very heavy into, what I'd
call a quick actions, like pre filled
responses, like Facebook messengers had
those since bots even launched, right?
Like handling natural language is hard.
Even how far LLMs have come, right?
It's still pretty difficult.
And you see applications all over kind
of leaning into templating, leaning
into quick actions, things like that
as a way to interact with the system.
I think over time you start to like
remove those layers and get a little
bit better about being able to not only
understand requests, but users get a
little bit better about making them.
But right now it's, it's
just an alternative.
So I interrupted you, but you
were saying how the version of
today is sort of the two things.
One thing is we can create pre-filled
prompts, do some of this Mad
Libs, make a user and teach a user
how to interact with it better.
Yeah.
So, like, we, we saw that
happening and it's still happening.
You go on LinkedIn right now, you
can almost guarantee your feed has...
One of the first five posts is a GPT
generated GPT post about how to use GPT.
And then you saw, it was two, three
Mondays ago, OpenAI talked about GPTs.
I heard somebody call them GPTS
yesterday and I'm not actually
sure what, what they are.
IT's going to be the
new GIF, GIF argument.
Right.
I think GPTs is probably right.
Right?
Like that's, that makes sense.
Like the pluralization, there's no S.
Yeah.
But in any case, you had that, which
is like, OpenAI signaling, " Well,
gee, we kind of should put more persona
specific or application specific
use cases around this technology."
It benefits both the user and
you can actually create a better
model in that sense, right?
Like you can, you can obfuscate or
entirely remove information that
is irrelevant to that use case.
And what that does is like harden your
ability to deliver really meaningful
results for a specific use case.
And that's exactly what GPTs are doing.
They're just taking
very specific use cases.
Templating them, wrapping them,
instructions, prompt language.
Like all of that knowledge is
based in baked into a use case.
And so, that opportunity is like the exact
one that ChatSpot is looking to capture.
Folks in the industry, coaching one
another on how to prompt giant, large
language models, and then the industry
itself moving towards a more templated
way of building these interfaces.
Well, ChatSpot kind of sits right in
the middle there, with the beautiful
added bonus of every HubSpot customer
can now interact with their own data
and their own system in that way.
When you think about who your end
user is, is it, "I am a salesperson.
I am curious about how to use AI
to do what I'm trying to do better.
And we have already thought
about how you might do that.
And we are going to give you
pre-made functionality and tools
to help you achieve that goal,
whatever that goal might be?"
When you think about who are we building
this thing for, what's your avatar for it?
Yeah.
So right now it's folks who are using
the CRM and helping make tasks and
the things that they're doing day to
day more quick and more effective.
Over time you start to see actually like
net new opportunities show up, right?
Like a proactive assistant,
an agent, all running through
something like a ChatSpot interface.
But right now it's, there's,
there's so much opportunity to make,
using, learning the CRM not only
easier, but like plain quicker.
Should you need to know the entire
corpus of your accounts properties?
And each value?
Or should you just be able to like ask
a question and kind of figure it out?
If I want to GTM with AI Dylan.
If I want to do that, how
does ChatSpot help me?
What is the workflow or what is the
user journey of, "I'm somebody and
I'm trying to do X and how do you
help me do that if I'm using ChatSpot?
What are you trying to do?
I mean, I could give you a really
hard one, but I don't want to
completely curveball you here.
Last time I watched Hot Ones on YouTube.
Have you ever seen Hot Ones?
Yeah, yeah, yeah,
Yeah, I think we should do
that for the next interview.
You should bring the hardest question
with the hottest chicken wing.
We like match them in advance and
then we come in and we're like, "All
right, you got to get this one right.
Let's do it."
Yeah, now we're really sweating use cases.
When you think about who the persona
you're solving for, is it mostly sales?
Is it mostly marketing?
Is it anyone who's doing anything?
It's doers.
It's just, it's folks who are in the CRM
and need the CRM to operate in their role.
But ChatSpot lives, and this is
something that like, I think is
very interesting as an outsider.
And I think I look at and evaluate
something like a ChatSpot.
And I think at the outset, prior to
you and I having a conversation about
this, actually, I was like, "Oh man,
ChatSpot is this natural language
interface for my HubSpot portal."
And what you just described in the
way that you're speaking about it
is, different than that, which is,
"We're going to give this GPT layer,
that is, attuned to what folks
that are trying to engage in sales
marketing and service activities.
And we're going to give them a GPT
that already knows that they're
trying to do that, and it's going
to help them attain that goal."
Versus them needing to go and prompt their
way to get something that's of value.
ChatSpot does not live inside
of HubSpot, it is separate.
Today.
To me, right?
When I think about it from my
perspective, I'm like, "That's weird."
And, is that weird?
And how is that intentional?
Is it, that's a current state and
maybe there's a world where it changes?
Like, how do you think about it?
Purely a current state.
I mean, we're bringing
ChatSpot inside the product.
Going to happen at some point.
Maybe.
Yeah.
I mean, how's Monday sound?
What'd you say?
Monday sound?
Amazing.
We'll get you access.
No the way we're thinking about
ChatSpot and moving inside the HubSpot
product is almost exactly the way we
thought about building custom objects.
From a, like a tactical
sequencing perspective.
Can you actually, what would be super
helpful is to, I think a lot of people,
if you were not like in the trenches
of HubSpot CRM before custom objects
existed and before it was there.
There was not a world where, and I
think this is what you're alluding
to, and going through like how you
did this, and then maybe creating
that parallel is interesting.
But I think for anybody seeing this,
if you were not in the trenches,
there was not no custom objects,
and then one day there were.
Is not how this went down.
And I think it's helpful, because I think
it's actually a very good avatar for how
HubSpot tends to build and launch stuff.
Yeah.
That's helpful context.
Like it's not a switch that we will flip.
It is a progression.
It's a continuation.
So what we're focused on right now
is building something that can scale
across the entire product at, I don't
know, call it the platform layer.
It's an overloaded term,
but that's what it is.
Something that both from a look
and feel and interface perspective,
can be used anywhere in HubSpot.
But then also from a functionality
perspective, I mean just the way
that products are built, right?
Like you have a team that builds a
platform, and you team have teams
that apply a use case applications.
ChatSpot today has been a single
unit trying to do both, right?
And the way that we can achieve scale
across a very broad HubSpot product
now, is to build a platform and
enable any team building anything.
From blogging, to quoting, to social
media, and anywhere in between
to build atop that functionality.
And so, what we'll see with ChatSpot
moving into the product, is core
applications sell first, some of
the more popular use cases inside of
HubSpot, the more visited frequent
apps, the more, the bigger problems.
And you'll start to see ChatSpot
appear, help you with those tasks.
And then over time, start to proliferate
across the entire product suite, and
you'll be able to see ChatSpot anywhere.
So, current state is
ChatSpot is its own domain.
ChatSpot is its own thing.
It lives outside of HubSpot.
You connect it to HubSpot.
Very close future state.
Tomorrow, maybe Monday, maybe later
than that ChatSpot is in product.
When you're thinking about
how ChatSpot is in product and
how people interact with it.
Is this a Clippy plus type of
experience that there's somebody,
there's, there's this agent that's
following me around and I can ask it
stuff or how am I interacting with it?
Yeah, well, I think you're going
to interact with it very different
than our, than our average
user as a very professional.
But what I'd like to think about it
is like, imagine every HubSpot user
had Connor sitting next to them.
Somebody who knew HubSpot
in and out, right?
You know, you'll have to deal
with some downsides, but there
is upside having that, right?
And what that will look like is, imagine
I'm a brand new user or I'm trying
to do something new, even in HubSpot.
Like you can be a tenured user and
there's always new arenas to explore.
And I can ask an expert like Connor
and say, "Hey, how do I do this thing?
Or can you help me do this thing?"
And instead of that requiring me to open
up a knowledge based article, or I'm self
discovering, like, poke around the app a
bunch, maybe watch some academy content.
Instead, what if it was just done for me?
And I could just focus on the results and
the outputs, and not learning the CRM.
Or figuring out how my property
structures are formatted.
Instead, just focus on the outcomes and
the things you're looking to do to grow.
Extremely interesting here.
From what you had just said,
because I agree with you.
I think that this is the truth.
What makes me so interested in what you're
doing is I think you're manifesting this
more than most of us that are talking
about it, which is, my perspective is
that all of these AI tools essentially
make it so that knowing what buttons
to click and knowing how to operate
the machine is not a prerequisite
to getting value out of the machine.
And therefore, by extension, the value
of knowing what buttons to click and the
value of knowing how to pull the levers,
goes significantly down and maybe to zero.
And I would argue that, we'll start
with like full time people and then i'll
probably transition a little bit more like
services generally and what that happens.
But there are a lot of people, for HubSpot
and for lots of other business software,
that their core function, their job, the
value that they add to an organization
is knowing what buttons to click.
And I would describe their primary
function is they interface with
a business user who does not know
what buttons to click and does
not know what levers to pull.
And that business user describes to
them, "Here's what I'm trying to achieve,
or here's what I want to have happen."
And then they go and click the buttons
and pull the levers to make that happen.
And I think the premise of what you're
describing, and I would argue almost
every AI tool in every application, and
I think this is the right direction, is
ultimately how do we make that user who
doesn't know what buttons to click, how
do we get them to their result faster?
There is very real intangible
disintermediation there.
And what do you think, not so
much to those people necessarily.
How does that shift how people
interact with the product?
And what happens to the
HubSpot-y world as that occurs?
Yeah it's a great question.
It's on a lot of folks minds these days.
I think a few things will happen there.
I mean, the overall sequence and
timeline of all of that is probably
much longer than most folks are thinking
at this time, I would love for that.
The nuance of an individual business'
data and the way that they think about
their operations is so incredible.
What we'll see is the ability to
use a pure non-customized HubSpot
out of the box using nothing but
your keyboard or even your voice.
Well, that timeline is
probably pretty short.
But the ability to use a HubSpot account
that is, has years of data and customer
context, and customization, and external
apps, and everything in between, like that
timeline is probably like exponentially
longer and potentially infinite.
And so, like, I, I think the timelines
are really interesting there.
And like, depending on the level
of business there, they're probably
stretched out in some sense.
And then the other part of your
question is like, "What happens to the
buttons or the way of using HubSpot?"
I think it just becomes different
buttons over time that are more
creative like applications, right?
Like instead of using HubSpot and the
expertise is point and click, it's
creativity and customer context, and
strategy, and less point and click.
Yeah, I think when I think about
it, that's a lot and I think
what you're saying is right.
And I'm going to talk a little
bit about service providers just
because I think that this matters.
And I think people are either
thinking about it a lot and
maybe too much or not at all.
And maybe not enough.
I think there's like a lot of oscillation
away from that, like, correct mean.
But when I think about sort of
what the AI tools do generally.
And then my perspective, what
products do generally and specific
to sort of like digital products
and software is, interacting with
machines is hard and has been hard.
And historically, you had to have
many layers of technical engineering,
software expertise, and in the original
sort of developer days, it's not
enough to say, "Oh, I can write a
script of code and it does something."
Like, how much ram time does that take it?
Does the machine have
enough processing power?
And how is that wired?
And all of that has to
go into consideration.
And I think you go to the example, which
I love of how everyone was wowed by
Pokemon existing on a single cartridge.
And it was just so big for
what should should fit there.
And the reason is there's a lot of
engineering behind reusing assets and
making sure that everything actually
can function on something so small.
And I think most development today,
because computer is so inexpensive, no one
really cares or thinks about that anymore.
And I think that this is kind of
a further extrapolation layer of,
we're making it easier and easier and
easier for people without a degree of
technical acumen to use the machine.
Which means more people can use the
machine, which means net productivity
increases across the board.
And, to your point on, on sort of
that role of disintermediation,
how do you think this impacts?
I mean, in the HubSpot I was just looking
up to try to go and see the number
that I say expands because there's so
many new HubSpot partners all the time.
And there are 6,612
currently in the directory.
And so there are 6,600 companies
globally who have dedicated some degree
of their business towards helping
people know what buttons to click,
and how to click them, and know what
levers to pull, and how to pull them.
And I, I would say that historically,
and I think a large percentage of those
are allocated to, "Let's help someone
who's brand new to HubSpot, maybe
brand new to CRM, start to use it."
And there's like onboarding
and the rest of it.
How do you think that ChatSpot
changes where they focus?
Because I imagine what they're going
to have to do changes a lot because
HubSpot is going to be able to make it
a lot easier for a net new customer to
use the platform for the first time.
And by no means does that mean all
these people have nothing to do.
It means that what they're doing
and where their focus is changes.
How do you see it changing?
Yeah.
I think, I mean, in my ideal, all
6,600 of those, those partners have
higher order of problems to solve.
And I think that looks a lot like building
account specific GPTs using ChatSpot.
And now we're, a few moons
away from this, right?
But like, imagine if your work
went from, instead of customizing
the way your CRM records work, or
like instructing somebody on how to
find the right filters, or drawing
property diagrams or association maps.
What if the higher order problems were
like, customizing the AI for your account,
customizing ChatSpot for your account.
For ideal outcomes, for brand
styles, for brand tones, for exactly
the way that your business needs
to use something like ChatSpot.
Like, I think it actually opens a new
door for all 6,000 of those partners.
Okay, so you see a world where
there's not just a sales and
marketing and HubSpot GPT.
There is every organization
has their own GPT and agent.
That GPT and agent lives on
and is inside of HubSpot.
And there is a world of consulting
and services providers who are doing
what to make that agent useful.
I mean, it boils down to
something like customization.
And you can look, you can look at
basically if you take the parallels of how
somebody would build a GPT today, you kind
of like squint at those and see where,
the future may lie for somebody doing that
within a hub or a CRM instance, right?
It's providing it, you know, key
knowledge sets, stylistic instructions,
core data sets, providing unique
data, data unique to the business,
connecting external data sources.
And things like that, like I
think that that becomes the...
Wild.
Yeah, I think that's the exact paradigm.
Yeah.
I mean, it, it makes sense.
It flows, right?
You think about the, the, if you
do a tremendous distillation.
The CRM at its core is effectively
a bundle of software layers that
lives between the business and the
database and makes that database
more interesting and useful.
This is just that again.
And, how are we going
to not give you a CRM?
What if you had a place that
had all your customer data and
had it all available to you?
But what if you had an agent that had all
your customer data and had all the context
and knew it was happening, and every
person in that organization is now 10,
100x more effective and efficient because
they're just interacting with that thing
and telling them what they want it to do.
Now you can't go build that GPT
without the access or the data
store that HubSpot provides.
Right?
So if we...
How do you think that applies to...
so I had a really interesting
conversation with somebody that's
relevant to the show actually.
But I don't know what the order will be.
So I don't want to reference it,
but you'll hear it at some point.
And, and they were talking a
lot about how they're trying to
do this for their organization.
Like what you are describing as
something that they are trying to do.
And they're doing it from more of a
product and an engineering perspective.
But the big blocker for them is on
security and data share, and if they go
and connect to this, they're supplying
that customer information to OpenAI and
it goes into train the models, and they're
concerned about all of these pieces.
That seems to me from what you are
describing to be maybe, an extremely
important, if not the most important
differentiation, from what you are
doing and how you're positioning it.
Yeah, yeah.
So one thing I was talking to somebody
the other day, actually a very
knowledgeable HubSpot partner who didn't
know that when you're using something
like ChatSpot or any AI feature through
HubSpot, whether we're using OpenAI
or any other vendor that we're using,
their models downstream are not trained.
And that's like something we need
to do a better job at it, screaming
from the rooftops because...
I don't know that anybody really
understands what all this means
just yet in their defense.
Yeah.
Right.
And so like, that is, that is a
benefit to using an API versus
the, the native product, right?
Like if you're, if you're kind of
working through HubSpot through
OpenAI or any other vendor we
choose to use, their downstream
models are not trained on that data.
The other long term security lever
we have here is like, you can imagine
HubSpot going down the route of
creating fine tuned LLMs at the
portal level and things like that.
And at that point, the data
is still within one place.
Do you think that the...
Okay, so the future of businesses
is everyone has their own GPT.
It is connected to your data sources.
It's trained on how you do things.
And there is a whole universe of
consulting and service providers and maybe
full time professionals whose job it is
to tailor and hone and customize that.
And I keep using the word GPT
and it's probably reductive.
It could be AI agent, it could be a
lot of different things, but to make
that thing a lot more effective.
How does that impact the future
of how people, specifically
people in these GTM roles, work?
Because I think it's really easy
for people to think about...
Like if you, if you've watched anybody
code, and they're using sort of the
VS code, or the Copilot, or any of
the GitHubby stuff that's going on.
And you're like, "Oh, wow,
this is suggesting stuff for
people who are writing emails."
And like, that seems very easily
understandable to connect it back through.
If an organization as one of these
deployed, and it's really tied to
the CRM, which things where people
work, how does what I'm doing as a
salesperson or a marketer or service
person, like, what does my day look like?
I think your day looks, your day
looks similar to as it does now.
And if your pie chart is 60 things you
love doing and 40 things you need to do,
it's probably more like 70 things you love
doing and 30 things you need to do, right?
Like I want to take the toil work out
of using CRM or selling or marketing
away using something like ChatSpot.
How, how can we allow you to just do the
things you love doing about your role?
The thinking, the customer
conversations and all that.
And that looks like a proactive product.
You know, finding you in the right
moments and auto filling, making
suggestions where you need them.
But it also looks like a companion.
We've been thinking about use cases
where folks kind of like set their
mental timers to do things of like,
"Oh, you know, we should check
in on these property definitions,
make sure they weren't updated."
Or like, you know, "Refresh these
workflows," things like that.
could you kind of just deploy
an assistant like ChatSpot to to
take care of all that, for you?
I think so.
And I, I hope that, the, shift of what
somebody does in a given day can be more
along the things that they love doing.
One last thing that I'd love to talk
about before we wrap is something...
I mean, we just did a bunch of AI
research that we're releasing at some
point, depending on when this airs.
Might may have already come
out, maybe coming out tomorrow,
maybe Monday, as Dylan says.
But I think one thing that we found
is, people that have AI tools that are
native to the platforms that they're
working with, tend to report much higher
efficacy and much higher outcomes.
I think one of the things that makes,
in my opinion, HubSpot very uniquely
positioned to leverage this moment, is
that HubSpot has a system and a tool
that is a CRM layer with a whole bunch
of functionality that is layered on top.
As opposed to a bunch of
independent software suites that
solve for a particular use case
that are then connected together.
And I think that framework and that
platform infrastructure allows for
the impact of sort of an AI solution
to be much, much higher because
it's expanding across everything.
As opposed to, you have to
wire it into each place.
If you have well thought out concepts
and ideas on that in terms of speaking
to why or what you think makes that
different I'd be super interested.
What specifically?
The thought for me, right, is like, if,
I am another business software company,
and I have a lot of different products,
and a lot of different tools, and those
tools maybe talk to each other, but
they're talking to each other via APIs.
But their fundamental data architecture
is different, the systems are different,
and that means if I'm going to go build
and add AI to my products, I'm not
building an AI for my suite of products.
I'm building an AI that has to be tailored
to every single tool that I, I sell.
Versus, with HubSpot, all of these
different hubs are just software
functionality on top of the same core CRM.
And therefore, my assumption is
that the ease with which that, that
AI can be deployed is much higher
and the impact of what the AI can
do is also theoretically higher.
Yeah, I agree with that.
I think the, the ease of
deployment is, is mostly the same.
But the impact and the efficacy
has much more opportunity.
We have much more opportunity there.
The way I look at it is like the flywheel
just got a lot of torque, the ability
to have a unified data store and things
that look, feel, and act the same.
Our ability to deploy AI features and
functionality atop something like that
has the potential to deliver vastly,
vastly different customer results.
Then the need to, to kind of essentially
like the story doesn't change to,
to connect a bunch of systems, try
to unify language there, right?
The locus of centralized data, the
flywheel of all business functions, GTM,
OPS, all related, all living on that same
core infrastructure, makes our ability to
deliver those features really exciting.
I have one last unrelated
question for you.
Which is, I think if I think
about what's the what's the most
exciting and interesting job to
do in in the universe right now?
And if you think about if CRM platform is
where most people are going to experience,
if most people are going to experience
AI at work, and the CRM platform is
the place where most people work.
Then building AI for CRM platforms
is one of the highest impact
AI things that you could do.
And probably one of the
most exciting things.
And ergo, I think you specifically
have one of the most interesting and
exciting jobs that exists right now.
What do you think is cooler
than what you're doing?
If you were to do anything
else, what would it be?
Actually there's a PM on my team.
Her name is Rachel.
She has a cooler job than me.
She uh,
Dylan wants to be Rachel.
That's, that's the
answer!
Uh, Rachel's a, uh, she's
brilliant senior product manager.
She's been at HubSpot for
a few a few months now.
And she leads our specific
team called AI Innovation Labs.
And they almost have a metric of failure.
Yeah.
Can they experiment and try
such cutting edge functionality?
Take such big swings, and in hit a
few, a few home runs along the way.
That is, is the one gig I'd take over.
All right.
That's, that's a good answer.
Still at HubSpot, still in product, still
in AI, but not necessarily having to
build stuff that works and instead just...
Just taking huge swings all day and
just at some juncture, whether it takes
a year, five years, 10 years, there
will be an absolute home run out of
that, out of that team that'll work.
And I'm absolutely thrilled
for what it's going to be.
Incredible.
Well, Dylan, thank you so, so
much for talking to me about this.
I'm being incredibly genuine when I said
I was absolutely looking forward to it.
I think that you are an incredibly
interesting and amazing and
intelligent person and I think
you are doing incredible things.
And so just having the opportunity
to talk to you about it is
something I, I cherish for real.
Thank you.
Thank you.
I appreciate it.
And that feeling's mutual.
It's been quite the ride.
A couple of years, four or five
years now, a couple INBOUNDs.
We've gone from talking about, "Can we
just please figure out how to change
property permissions, to something as big
as training custom GPTs for your portal?"
So it's cool evolution.
The future is bright and, and we've
had a good, we've had a good, a good
track record of delivering on it.
So we'll, we'll expect more of the same.
Yeah, exactly.
Awesome.
Sounds good, man.
I'll see you soon.
Thanks, Connor.