All right. Hello. Hello.

Hi everyone. My name is Josh Kim and I'm

so thankful for all of you for joining

us here and I'm excited to share how we

on the SpaceX AI team use Grockbot for

marketing.

Here's our runner show for today. Uh

I'll be sharing what Crockpot is, how we

use it internally, uh to market and grow

the product and most importantly how you

all can start using Grockbot within your

teams. Let's jump in. So let's start

with the context. Um the way we all

started using AI was pretty simple. Uh

many of you may still be using AI today

just like this. Uh and I certainly was

in this camp uh before I started on the

SpaceX AI team. The typical pattern is

that you have one or two AI chat bots

that you really use and leverage as your

thought partner. Meaning, you chat with

them, you ask questions, you spar on

strategy, maybe you edit copy. Um, but

they're really there as your thought

partner and not a doing partner.

As AI has started advancing, there's

been this rise of generalized and

specialized agents that can start to do

tasks with you. Um, we refer to these

internally as co-pilots because while

they can get those tasks done with you,

they still typically require you to

babysit them, push them along,

particularly for those complex tasks

rather than truly being able to delegate

your work.

And that's the reason why we're so

excited about Grockbot.

With Grockbot, marketing teams can now

actually start to automate away true

pieces of their work. You can think of

Grockbot as always on asynchronous AI

teammates that actually do and finish

the work on your behalf. And most

importantly, they'll do it the way that

you do it.

And today I'm excited to demo how on the

SpaceX marketing team, we not only use

bots to automate jobs and tasks, but we

actually use them to staff up functions

and create teams of bots to actually do

the work for you.

But first, let's do a quick intro of the

product. For those of you who haven't

tried it out yet, this is Grockbot. And

now some of you may be wondering how is

Grockbot different from all the other AI

tools and chat bots that I might use

today.

What's different about Grockbot is that

rather than working with one agent or

one chat with multi-threaded

conversations,

you create individual bots for each job.

And what that means is you think of one

bot to be a singular teammate that you

then delegate away a single task or a

set of related tasks.

What this also means is that you can

actually message your bots and talk to

them just like you would Slack or

message someone on Teams. Uh, and to

take it a step further, bots can start

to talk to each other and actually

collaborate with each other within group

chats as well.

And as you continue to work with your

bots, they'll learn how you work.

They'll learn your preferences. They'll

learn your memory. And they'll get

smarter and smarter and get closer to

the way that you actually do the job

yourself. So, similar to someone on your

team or someone that you collaborate

with, the more feedback you give and the

more you invest in your bots, uh, the

more they'll be able to do for you in

the long term.

And what makes bot especially unique is

that while it can hook up to all the

MCPS and APIs that you're currently

using today, it also has its own

computer. And that's a bit of a novel

paradigm shift from how you might be

using other AI tools today. And what it

simply means is that because it has

access to its own browser and its own

computer, so long as you give it access

to it to tools that you use, it can

actually use those tools just like you

would.

So what does this all mean for

marketers? If we talk about how it can

use tools just like you, that means that

your bot, you can teach it specific ways

that you like to optimize your bidding

strategy in ad platforms or the nuances

and how you put together a campaign

brief or maybe just how you like to

manage your computer. But ultimately,

what's really unique and valuable about

bot is that it becomes customized and

flexible to the way that you work.

And with bot, you'll also find that

you're no longer having to babysit and

push along and really, you know, remind

the bot to keep working. And rather,

Grockbot is extremely ambitious and it's

also very proactive. And so now you'll

be working with a team of bots that

actually want and will finish the work

for you. Um, and that's quite different

from other tools you might be using.

Because bot is hosted in the cloud, that

also means that they're running 24/7. No

more keeping your laptop open or needing

a Mac mini setup. And for marketers,

that means if you have reporting or

metrics that you're you're looking at on

a regular basis, you can trust that

Grockbot will be able to pull those down

for you.

And the best of all, as you might have

noticed, it's really simple to use. This

UI is as familiar and intuitive as

iMessage, which really lowers a barrier

to entry for any marketer, no matter

what their specializ specialization is

to weave agents into their daily

workflows.

Okay, so let's actually dive into what

we're going to do today during the demo.

Uh what we're going to do together is

actually launch a marketing campaign

from scratch from end to end. Um,

meaning we're going to start from the

first task that we typically take on,

which is market research analysis. We're

going to go all the way through to

ideulating and betting on positioning,

uh, updating a landing page,

starting to build and monitor ad

campaigns, draw that analysis, and then

actually end with automation. Now, this

is a lot of work. Uh, there's a lot of

things that happen behind the scenes.

For all of us in the room as as

marketers, you know that this slide

might look simple, but actually it

actually spans across many different

teams, functions, disciplines within a

typical marketing or meaning the person

who conducts market research is, you

know, having to work with a product

marketer to identify and best

positioning who then has to pass off a

task to the web team and so on and so

forth. And for anyone who's done

marketing before, that context

management, the coordination, the

orchestration of a campaign like this,

that's truly where the craft of the work

happens because it's the crux of our

job.

So today, I'm going to introduce you to

my marketing team of bots. And I would

be remiss to say that I built all these.

I was crowdsourcing these across the

entire SpaceXi team, uh, which is a

feature I'll talk about. But I'm excited

to show you how you can orchestrate an

entire campaign build across the entire

flow. So you can say hi to our market

researcher, our product marketer, our

website ops performance marketer,

marketing analyst, and lastly our

project manager.

So with that, let's dive into our demo.

Now for the sake of our demo we have uh

launched a brand new product and it is

called XAIR.

So this is an airline product that we

built as something that we can use to

build a net new campaign for. And you'll

see that this landing page is set up.

It's our homepage. People can do some

airline route searches, look for fairs,

and then we have some positioning and

value proposition, value propositions

currently in place. So, I'm going to

navigate to my Grockbot instance. And

just to walk through what you see on the

screen on the left side, sidebar, that's

where the bots live. I've pinned them to

the top, similar to what you might do in

iMessage. And I'll be flipping through

each bot and I'll explain what I'm doing

as I'm going through it. But let's get

started with the market research bot. So

what I'll say first is, hey market

researcher, what I'd like you to do is

study the X error website. Get a deeper

understanding of what the product is,

what market we're operating in. And now

I want you to go do a competitive

analysis.

identify and deeply understand our

competitors. Look at their marketing

websites. understand their positioning

and then importantly identify the gaps

and opportunities that we have to

actually strategically

um competitively position against them

within our marketing strategy.

So just dictated a prompt. I'm going to

set this off. Now the market researcher

bot is actually going to go and scrape

the website. It's going to understand

the product and then it's actually going

to go find the competitor competitors

their landing pages and start performing

this analysis.

So, while that's running, let me just do

a quick walkthrough of the additional

bots that we have. Um, we have a product

marketer bot, website ops performance

marketer,

marketing analyst, and a project

manager.

And

importantly, what I'm about to show is

that as this works, you can actually

kick off jobs asynchronously. Meaning,

you might have one bot that you work

with uh as your main point of contact,

but you can actually kick off work in

parallel. across a team of bots and that

means that orchestration is happening in

the background uh because the product is

always on. So market researcher is

saying it's pulled xair the competitive

asset is being identified. It's going to

show us a screenshot first of the

website and then go through gaps and

positioning.

Great. So it has its first output of an

analysis. It's talking about the

product. It's talking about the market.

is has identified key competitors. These

are all airlines that we're all familiar

with. Um I think most importantly it's

identify the gaps and where to lean in.

So similar to how you might work with a

market researcher, you ask, "Hey, I'm

launching this new feature. I'm going to

have this new exciting campaign launch.

Help me understand what's currently in

market." So we un we are optimizing for

breaking through the noise and actually

landing and resonating with our target

audience. So I think the most important

note here is you know these gaps around

useful time day design uh long hauls and

then really leading every message with

useful time which is something I think

is resonating at least with me and it's

also shown me that it's pulled up the

website and this is actually a view of

its browser that it has access to and

that it controls and it's showing me

progress updates of it working in the

background and being able to actually go

to the website itself, pull down a

screenshot and show to me uh how it's

progressing.

Now, marketer researcher has done one

task, and that's great. What I could do

is actually kick off another one, but

what I'm going to do instead is actually

go to the next bot, and that's the

product marketer. And I'm going to have

the product marketer actually do a

handoff of that work that the researcher

just did and take that into their work.

Okay, product marketer, what I'd like

you to do is now go to the market

researcher bot, do a handoff of the

analysis that just performed and I want

you to draw up a draft of a positioning

brief where you identify how we should

go to market, what our one oneliners

are, our positioning and packaging, um

some value statements, and then also

show me some examples of how this should

come to life across two or three

different marketing services.

So, as this is running, what you'll see

is in the chat, the product marketer bot

is starting to kick off this work. I

told it to actually go and speak to the

market researcher and actually grab that

context and information. Um, as it's

working, you'll see there's a visual

indicators. It's, you know, doing

things. It's running in the background.

Um, and then it'll actually show you,

hey, I exchanged a few messages with the

market researcher bot. And that's

showing that they're starting to

collaborate. And that's something that

you can tell it to do in the beginning,

but over time, it gets smarter and

smarter and smarter. And it will

actually do it proactively, meaning you

have to be less involved in doing the

orchestration of context and knowledge

between your bots, and rather they

actually start to work with each other

organically.

So, it's telling me it's grabbed the

market researcher analysis. It's

drafting the positioning brief. Um, it's

created a Google doc where it's actually

starting to draft that position brief

which we'll jump into. And inside it's

telling me what exactly it did.

So,

let's bring this up.

And if we go through this, this is

looking like a solid first draft.

Um I think what what I will do is start

to read through this. It has a research

handoff that I got from the market

researcher is identify the target

audience long haul travelers who want

useful time back. I'm going to leave a

comment and say this is great. Lean into

this in our messaging.

It's talking about positioning and

packaging and how we should be thinking

about our marketing strategy and the way

go we go out with messaging.

And it's also left some notes on go to

market, some additional oneliners about

different types and angles to lean into

and then some value statements and

examples of how it shows up. So there's

a paid landing page here that is

drafted. I think what we'll do for our

campaign is actually take this page,

shove it up into the website, launch the

landing page, and actually start

directing and building an ads campaign

to do a messaging test to see what's

landing the most and what's resonating.

So, I'm going to close this and I'm what

I'm actually going to do is tell it,

"Hey, product marketer, I've just left

some comments inside of the Google doc.

What I want you to do now is go through

the Google doc, take the comments and

the feedback, incorporate it into an

updated draft. And while you're at it, I

also want you to build out a full

outline of the landing page. And then I

also want you to ideulate and draft some

Google search campaigns to do variant

testing between the copy angles that you

drafted.

Now again, this is all happening within

one bot right now, but what it's doing

is flipping it back and forth between

market researcher bot to make sure that

it has the right context. is going into

the Google Docs which I've connected

through uh the MCP to be able to

actually look at the comments, look at

the feedback and incorporate into a new

draft. So, it's telling me, okay, it's

I'm reading your comments now. I'm going

to update the brief, expand the landing

page outline, and then start drafting uh

search ad variance inside the sheet,

which is great because that's typically

how performance marketers work.

Now, while this is happening, it's

reddrafting the file. I'm actually going

to flip over to my performance marketer.

I'm going to say, "Hey, performance

marketer. Uh, I would like you to start

creating a shell campaign inside of

Google Ads. And what I want you to do is

build it so that it's optimizing for

clicks because we're going to do some

copy and messaging testing that's being

drafted by the product marketer right

now.

So, before this demo, I actually hooked

up um this bot into my our Google Ads

account. What's happening is it now has

access to it. And when I kick this off,

it'll actually go into it and you'll be

able to see because it'll show

screenshots of how it's starting that

build and it'll be able to actually do

the clicks for me um and start the shell

campaign that can start uh trafficing

the ad copy into.

So, let's flip back to product marketer.

There's an updated brief. We can open

it. Comments are resolved.

Looks like it's in a good state. It has

a fuller landing page outline uh that we

can implement. It also has more built

out campaign variants which I believe

are also inside this sheet where it has

variant name, hypothesis, ad group name,

etc. as well as the URLs and then it's

also built some additional copy variants

here. So all this is happening in the

background. You know I've kicked off two

or three jobs in parallel. Um, but the

best part is that they're all sharing

context with each other while not

tripping over themselves to get their

work done.

Okay, this is feeling pretty good. What

I'm going to do now is I'm going to go

to website ops and I'm going to tell

them, okay, so landing page outline is

drafted. What I want you to do now is

take the landing page from the latest

brief that the product marketer just

drafted and spin up a PR to actually

push that as a new landing page into the

website and send me screenshots as

you're working so I can keep monitoring

your progress

on the back end. Again, this bot is

hooked up into our repo. It has access

right access to our landing page and our

marketing site. So, it's now able to

actually leverage what's under the hood

is cursor to be able to go write the

code, look at the copy, get the handoff

from the product marketer and actually

ship a new landing page to the marketing

site.

So, it's going to pull down the latest

product marketer brief. It's going to

open the PR. It's also going to show me

progress screenshots as it works.

Meanwhile, if we go back to performance

marketer, um, it's successfully gone and

found the copy that was drafted. Uh, it

has shown me some screenshots of it

actually starting to build the campaign

inside of the platform. And we'll let

this keep running as we finish up the

landing page.

Okay.

So, you can see how it's working

through. uh it is going to take a

landing page, it's going to update it

with the latest full uh paid landing

page outline that's passed off from the

product marketer and we'll actually be

able to ship this and you know ship it

to production um and see the updates

come to life.

Okay. So, while this is running, um I

would would like to show a feature that

uh I think is really special. It's quite

unique from the way that you might be

using other a AI agents now, which is

our sharing feature. Um, so something

really unique about Grockbot is that

while you can create these really

robust, powerful bots within your

workflow that are customized to the way

you work, you can continue to feed it

context and memory and build routines.

You can also turn those into templates.

And this is where I'll show you. You

share as a template. You're prompting it

to say, "Hey, I want you to create a

template of yourself that I can then

share with someone else." And the

implications of that are pretty powerful

because that means you can not only take

a snapshot of your bot in time and turn

that into a duplicatable

uh bot and agent that you can share with

your team but also people externally.

And what that means is for the receiver,

they are then have access to literally

the way you work because you've injected

all this context, your expertise, your

specialization directly into this bot

and they can easily take it, copy and

paste it directly into their Grockbot

instance and then start using it uh

exactly the way you would.

All right. So, uh, looks like it is

working on getting this page up.

It's still writing the page in the PR

and it'll show me when the preview is

up. So, while that's working, what we've

done so far, just as a quick recap, we

started with the market researcher.

We asked it to do an audit of our

product page, understand the product,

identify identify competitors, and then

give us strategic guidance on the gaps

and opportunities. That was passed off

to the product marketer to then draft up

a positioning brief, which it did in

Google Docs. I was then able to go into

that document, leave comments and

feedback, and then go back to the bot

and say, "Hey, fix this." The same way

that you might give feedback to someone

on your team when they're drafting a

document. It's done that rev. and come

back and said, "Great, I did that. I

also created an outline for a paid paid

landing page, and I'm also drafted this

copy that you can then traffic into your

ads."

That outline is now being worked on by

the website ops, which I'll give it um

it's currently working through this, and

that is going to be turned into a

landing page update that I'll push up

and we'll be able to host on uh the main

site.

performance marketer has started to

build the campaigns, built the shell,

and that copy is being fed into the way

that the campaign will be structured so

that we're getting that message test

live. And as this is running, um, rather

than needing to actually push the

campaign live here, because that would

require obviously like some additional

guidance on spending the budget and

hooking up a credit card, um, I've

actually already launched a campaign uh,

this week. And so what I'm going to do

is actually tell my marketing analyst,

hey, you're already hooked up to my

Google Ads account. I want you to

actually pull down that data and give me

insight and analysis into how the test

performed.

Hey, marketing analyst, what I want you

to do now is go into Google Ads. I want

you to pull down the data from our last

uh experiment on messaging, analyze it,

tell me the TLDDR of the insights, and

then give me recommendations on how we

should incorporate it into our marketing

strategy as well as update the other

assets that you've seen.

Okay. So, what this is doing now, it's

actually going into uh Google Ads

platform. It has access through that

through the API. It's going to pull down

the information, pull down the data, and

it'll be able to structure it into an

analysis that's all going to be living

within this chat here.

So, it gives me the TLDDR. It's pulled

this copy test down from copy from the

Google ads. Um, it has a clear winner uh

that's analyzed across different

variants of brand, customer promise,

hours in between, cheap that cost a day,

etc. Um, it has key metrics, spend, CTR,

CVR, etc. All the things that we care

about. I think most importantly, what

it's done is it's actually made

recommendations. And this is obviously

where there's some debate of how much,

you know, liberty do you give your agent

to, you know, actually make those

decisions for you. But I think in this

case, it's making recommendations that I

can then intake, you know, discuss with

the team, but also guide it to actually

make decisions on my behalf. Um but it's

made that entire process smooth and

without any dependencies that you might

otherwise have uh in a traditional or

okay so now we've finally gone through

every single piece research positioning

landing page which I will nudge one more

time push the page to prod

uh building the campaign inside a bot uh

that has access into your account from

the Grockbot instance

We've run the analysis. Let's pull that

down. And you might have noticed that

during this entire demo, I have to be

the one actually orchestrating. And to

go back to my earlier point, that's

really where the taxes, right? It's a

context switching. It's a having to know

the players, understand who's good at

what, and sort of like trafficking and

managing that entire process across the

entire workflow. Um, and that is

actually something else that you can

dedicate or delegate to Grockbot.

So now with my next and final bot, which

is the project manager, what I'm going

to tell it to is to do is actually study

the entire orchestration flow that we

just went through and then automate it

so that I don't have to be managing the

context, information, and communication

between all the different bots all the

time.

Hey project manager, I want you to now

study and talk to each of the bots in my

team and understand what each of their

roles is in bringing a campaign to life.

I want you to also look at our

conversations and look at where I had to

interject or give guidance or give

feedback and weave that into the way

that you work with them. And I want you

to also now kick this off and automate

it completely with three new campaigns.

send me screenshots and progress updates

that can stay in the loop. And most

importantly, I want you to now be my

point of contact, meaning I don't want

to talk to any other bots. I only want

to talk to you to save myself the

context switching.

So, I've now delegated away the entire

process of a campaign build and

throughout that identified, okay,

there's like tax and additional work

needed to actually make a campaign

happen from start to end. And the final

piece is actually taking another step

back and saying I'm just going to have

Grockbot do that and play the role of

project manager, coordinator, handoffs,

sharing the context, pushing the team

along. And I think most importantly,

what's most valuable is the fact that

you no longer have to context switch and

be bombarded with five different

messages from different uh individuals

or bots on your team. And that gives you

the clarity to be able to zoom back out

and really start to direct and uh direct

your energy and time towards where the

most leverage is.

So you'll be able to see it's saying

that it's the only um this bot is now

the only point of contact here. It's

studying the five specialists and it's

going to create and kick off three net

new campaigns that it'll then show me

screenshots of. So, it's identified

these are the campaigns we're going to

do research, positioning, landing page

build, etc. That's all queued up. Um,

and it will continue to keep me in the

loop so that I can continue monitor, but

I don't have to be directly involved in

every step of the process.

All right. And finally, we have uh a

website that we built up and running.

It's shown that it's actually pushed it

to prod.

And if I ask it, give me the URL.

It'll also give me the URL

that we can open C. And so this is the

URL that we drafted, went through

positioning, etc., and actually pushed

up to prod all from the one place where

all the work is happening, the Grockbot

workspace.

Okay. And that's it. That's our demo.

[applause]

So, just to wrap that up, um, what have

we learned by going through an entire

campaign build from start to end?

The first point is, uh, what's really

important when you're starting off with

Grockbot is to ensure that you're

scoping your bots properly. And the way

that I like to think about this is that

when you identify a bot or a job that

needs to be done, it's almost like

crafting a job description. Meaning

you're identifying a new teammate that

you need help from. You understand the

swim lanes and roles and

responsibilities and you're ensuring

that it's scoped tightly to ensure that

they are specialized and they're able to

get that specific uh piece of work done.

Uh because that's the way you squeeze

the most efficiency out of your team of

bots.

Second, you want to trust your bots. Uh

as you might have seen, these teammates

are ambitious. They're proactive.

They're hungry. They want to do your

work. And that starts with giving them

the access they need. Meaning, one of

the first steps you might take is

hooking them up into your Slack

instance, your email, and actually maybe

asking it like, "Hey, go study all the

context, all the messages, all of my

organizational history, and tell me what

you can do for me and take a take a job

off my plate." Um, but I think if you do

that, you give them the access, you

throw ambitious tasks against them,

you'll be surprised uh by what they're

capable of.

And finally, I've said this over and

over, but you want to invest in your

bots. They really get better and better

over time. You know, the point of

collaboration between botto is something

that gets uh more organic, especially as

you prompt it. Um, but they also improve

steadily as you start to feed it more

feedback, more context, more memory,

similar to how you might work or manage

a teammate, where the more feedback you

give them, the more time you invest into

them, the more they'll compound in value

and be able to take more and more off

your plate.

Lastly, here's a plug. Uh, this QR code

goes to our bot marketplace. And the

marketplace is actually where we take

the very best bot templates like the

ones that I shared, um, and host them.

Meaning, you might be able to create

your own bots and prompt them and build

up, you know, really rich context and

memory, but you can also just copy

others. And that's what I did is when

building out this demo, I crowdsourced

across the entire team. I said, "Hey,

who has the best positioning bot? who

has the best ads bought, etc. Um, and

you can find all of those on our

marketplace. So, I can encourage you to

go here and be able to copy them down.

And with that, that's it for me. Thank

you so much for your time. And I believe

we have some time for questions.