Copilot Conversations in Action

0:00

Hey everyone, welcome.

0:02

Thank you all so much for attending

0:03

our second session of the day on this track.

0:06

If you've been here and hanging out in the room,

0:09

I was super excited to kick off the last session.

0:11

I'm even more excited to introduce our next speaker,

0:14

Jason Hirsch, who is also a fellow gainster.

0:17

I get to work with Jason all the time.

0:19

He's awesome, very excited to tell you a little bit more

0:21

about our co-pilot feature within GainSight

0:24

and how you can create really compelling conversations,

0:27

leveraging chat GPT and leveraging really effective

0:30

prompting.

0:31

Couple quick housekeeping notes before I hand it over

0:33

to Jason.

0:34

Sorry, we call him Hirsch internally,

0:36

so I just want to say Hirsch.

0:37

But before I hand it over to Jason,

0:39

couple quick housekeeping things.

0:40

We will have a poll at the end of today's session in Slido

0:43

to get to that poll and answer any questions you

0:45

might have as Jason is speaking.

0:48

If you go to the agenda and then you click on the session

0:51

and then click on Q&A and polls,

0:52

that's where you can insert any of those questions you

0:54

may have for us to cover at the end.

0:57

We will also be doing a breakout session survey afterward.

1:00

So if you want to fill in that survey for any session,

1:02

you attend that poll.

1:03

Every time you submit a survey, you get an entry into a raffle

1:05

to win an awesome 50 euro, I think it is prize.

1:09

So I highly recommend completing those surveys.

1:11

Also a great way for us to get feedback on every session

1:13

to continue improving for next year.

1:16

And with that, I would love to hand it over to Jason

1:18

to kick off our next session, co-pilot conversations

1:21

in action.

1:22

[APPLAUSE]

1:26

Great, thank you, Tori.

1:27

Can y'all hear me OK?

1:29

Awesome.

1:29

So whose first pulse is this?

1:32

Nice.

1:33

I've actually been to like 10 of them.

1:35

And my fun fact is, six years ago,

1:38

I brought my son to Pulse London,

1:40

and they decided to put him on stage.

1:42

And they dressed him up as Albert Einstein.

1:44

He did a bit with Nick.

1:45

So basically my fun fact is I've spent like six years

1:48

trying to be as cool as my then 17-year-old son.

1:51

And now this is my starting point I get to talk at Pulse

1:54

today.

1:54

So thank you for joining me.

1:55

And as Tori mentioned, we're going to talk about co-pilot,

1:58

conversations in action.

1:59

Has anyone heard of a gain site co-pilot?

2:02

OK, one person who works for gain site, a couple.

2:05

It's actually something where it's in beta right now.

2:08

So if you're interested in your gain site customer,

2:10

you can reach out to your CSM.

2:11

They can activate it for you.

2:13

But we're going to be talking about co-pilot in action.

2:17

And let's look at our agenda here.

2:19

So my speaker notes are really small,

2:21

so we're going to have to go a lot by memory here.

2:23

But the things we're going to talk about

2:24

is how to use chat-based AI systems effectively.

2:27

Like we've all seen chat GBT in these chat-based systems.

2:30

So we're going to talk about some of the tools

2:31

that you can use to use them more effectively.

2:33

Then we're going to talk about using co-pilot

2:35

and what co-pilot is.

2:36

It's a gain site feature that were beta in, I just mentioned.

2:40

And also some of the use cases and best practices

2:43

when you're using chat-based AI systems.

2:45

Who's using something today like a chat GBT?

2:48

Something like that.

2:49

OK, so a lot of people.

2:50

I'm going to steal a question from Tori in a previous session.

2:53

Who believes that they actually get the results

2:55

that they want the first time?

2:58

OK, exactly.

2:59

So that's something we're all learning together on.

3:01

We've all tried it, but we very infrequently

3:03

get the output that we want.

3:05

So I'm going to talk about some of the tools and techniques

3:06

that you can use in your day-to-date chat GBT tools.

3:11

And also talk about how some of these tools are frankly

3:13

built into the gain site co-pilot to try to make things

3:16

a little bit easier for you.

3:17

So chat-based tools in general--

3:19

I mean, the reason we're talking about it

3:21

is because they're going bananas, right?

3:22

You're going to go to holiday dinners,

3:24

and all your family's going to be talking about how

3:27

you work in AI, right?

3:28

It's just super popular.

3:29

You can't get away from them.

3:30

Everyone's building them.

3:32

Everyone's building AI-based chat systems.

3:35

So that's what we're talking about today.

3:36

And gain site is really no stranger to it.

3:38

We've actually been doing AI and machine learning

3:40

since, I think, 2018 with the acquisition of staircase

3:44

recently, we've really been able to turbocharge

3:46

our capabilities in there.

3:47

So what we're going to talk about is the fact

3:49

that there's a lot of chat-based AI systems out there, right?

3:53

We have gain site co-pilot.

3:55

You might have also heard of Microsoft co-pilot.

3:57

It's kind of the term everyone's using for their chat-based

4:00

system.

4:01

Agent force, just released by Salesforce, right?

4:03

It's another system that you can actually

4:05

use to query and learn things in your Salesforce environment.

4:09

And then, of course, there's agent--

4:10

I'm sorry-- open AI, which is kind of the big dog

4:14

that we're all talking about and what our product is actually

4:17

leveraging.

4:17

So these are the systems that are most commonly used out there.

4:21

And if I jump into what we're doing at Gain site

4:23

to make sure that we're taking advantage of AI not only

4:26

internally, but also making sure that it's

4:28

available for our customers.

4:30

So number one, we're customer zero.

4:32

We've been using all these AI systems for really since 2018

4:36

and really a lot more in the past 12 to 18 months.

4:40

So we're really customer zero.

4:42

Not only are we building these tools for our customers

4:44

that you're consuming right now, we also

4:47

have half of our teams internally

4:49

are actually taking advantage of AI systems.

4:51

We have a really unique program that we call AI for all,

4:54

where something I'm actually part of,

4:56

where we're doing Friday sessions, where anyone in the company

4:59

can join, learn about what others are doing,

5:02

and we can provide best practices.

5:04

So we're actually trying to do this inside out,

5:06

as well as present it to your customers.

5:07

So we have people like Seth dedicated.

5:11

That's all he does is teach people

5:13

like Gain site how to effectively use AI tools.

5:16

And we're building a lot of them internally.

5:19

We've also built custom GPTs internally.

5:22

Like I look after a lot of competitive.

5:24

I built custom GPTs, where I'm consuming a lot of the information

5:27

and I'm actually able to disseminate them to our customers.

5:30

I'm sorry to our employees through different chat GPT tools.

5:35

And internally what we have is we actually also

5:37

use chat GPT enterprise.

5:39

So our teammates are taking advantage of this

5:41

to be more efficient.

5:43

And on top of that, we're also building these AI systems

5:46

for our customers.

5:47

So that's what we're doing.

5:48

And if you work for a place that's maybe a little bit slow,

5:51

what I'd really recommend you do is have discussions about

5:53

how do we basically teach people internally

5:56

how to use these different tools?

5:57

And it will eventually become pervasive,

5:59

eventually get into your product organizations.

6:02

And then you can start rolling these things out

6:04

to your end customers.

6:05

So if you're interested, I definitely recommend,

6:08

I'm happy to talk to you about our AI for all concept.

6:11

I'd really recommend that that be something

6:13

that you do in your own organizations.

6:17

So when it comes down to what I'm here to talk to you about

6:19

today, it's really the Gainsite co-pilot.

6:21

And what Gainsite co-pilot is, for those of you

6:24

that are Gainsite customers, what you'll find

6:26

is on the navigation bar on the left-hand side,

6:29

there's going to be a new button when it goes live.

6:31

It's like we stole Tories here.

6:33

And it's going to have a little prompt.

6:35

Everyone sees that little guy and knows it's AI-based stuff.

6:38

And what's super cool about it is it allows our customer--

6:42

I'm sorry-- are CSMs at our customers--

6:44

to go in there and ask questions.

6:46

So not only will it give you prompts,

6:49

but you can also ask any question you want.

6:51

Now what's really interesting is,

6:52

anyone familiar with Gainsite timeline?

6:55

OK, so for those of you who aren't familiar,

6:57

it sounds like a lot of you are.

6:58

It's basically a way to take notes.

7:00

And it's also a place that we can deliver things

7:03

like gong calls and Zoom calls directly

7:06

into the Gainsite system.

7:07

And what our co-pilot is number one

7:09

going to allow you to do is ask questions.

7:12

So actually, just today, someone said,

7:13

hey, do we have a meeting with XYZ customer?

7:16

And I was able to go into co-pilot because I'm part of the beta.

7:19

And basically go in here and say, do we have a meeting here?

7:22

And if timeline has that information, it would deliver it to me.

7:26

I could then ask a follow-up question.

7:28

What is the meeting about?

7:30

What are the risks associated with this account?

7:33

What are their use cases?

7:34

And it's actually going to query all the insights we have

7:36

in timeline, give it to me or really anyone

7:39

in our organization.

7:41

The second cool thing about co-pilot

7:43

is it also provides best practices.

7:46

So if you're a CSM and you're thinking, oh, gosh,

7:48

the boss just told me to prepare an EBR for our biggest

7:51

customer, you can actually go in here

7:52

and you're going to find Gainsite best practices.

7:54

So you can ask questions.

7:56

I'm a CSM.

7:57

I have a meeting with our largest customer.

7:59

It's an executive business review.

8:01

It's with my executive team.

8:02

What should I do?

8:03

And it's going to give you best practices.

8:05

You can also do things like, I'm struggling with scale.

8:08

What do you recommend?

8:09

And it's going to consume all of our best practices

8:11

and deliver them out to you.

8:12

And you can have a conversation.

8:14

And the third area is support.

8:18

So if you're a Gainsite user and you're thinking,

8:21

I don't know how to use this feature.

8:22

What are the best practices for timeline, et cetera, et cetera?

8:26

It actually crawls through our support system.

8:28

So number one, our first iteration of this

8:31

is going to have all the timeline information,

8:34

all the best practices of customer success,

8:36

and then information about Gainsite itself,

8:38

kind of like the how to break fix type of scenario.

8:41

So that's what our first iteration of this is.

8:43

So one of the things I also want to talk to you about

8:50

is how to use chat effectively.

8:52

So I asked earlier who uses chat GPT, everybody,

8:56

who uses it effectively, nobody.

8:58

It's really difficult to understand.

9:00

So I want to talk to you about some of the strategies

9:02

that I would recommend and Gainsite would recommend,

9:05

not only when you're using our own system,

9:07

but also when you're doing things,

9:10

whether you're looking for recipes

9:11

or looking for a vacation in Amsterdam.

9:13

Some of the strategies you can use to get better answers.

9:17

And before we dig into it, really,

9:19

if I just think of the basics,

9:21

use these tools as an assistant.

9:23

Use them as a sounding board.

9:25

Use them as if you're having a conversation with somebody.

9:30

The reason I say that is if you go to a chat based system

9:34

and really independent of work and you say,

9:35

"What should I do for a living? What should I have for dinner?"

9:38

It's going to be like, "I don't know.

9:39

"I don't know where you live.

9:40

"I don't know what kind of education you have.

9:41

"What should I have for dinner?

9:42

"Are you a vegetarian? Do you like fish?"

9:45

So without context, none of these tools

9:47

are going to work very effectively.

9:48

So if you remember anything, just remember that.

9:51

Use it as an assistant.

9:52

Use it as a sounding board.

9:53

And give it as much context as you can,

9:55

because just as if I walked up to Russell there,

9:58

we don't know each other and I say,

9:59

"What should I have for dinner?

10:00

"He's not going to know.

10:01

"What is this guy like?

10:02

"Is he a vegetarian?"

10:03

So you really have to get these things context

10:05

in order to get the best outcomes

10:07

for personal reasons and also for your business reasons.

10:10

So this is a crazy looking chart,

10:13

but the big takeaway I want you to have here is,

10:17

these are all the different things

10:18

that a chat based system does when you introduce a prompt.

10:23

So if I give away the answer to the end of my presentation,

10:28

gain sight through our co-pilot,

10:30

we actually do all this stuff behind the scenes.

10:32

So you as a user do not have to learn this type of scenario.

10:36

So these are some of the things that we're doing for you

10:38

and some of the reasons that it's beneficial

10:40

to use gain sight co-pilot.

10:41

But really the big takeaway here is,

10:43

if you introduce a text based prompt,

10:45

the concept of a zero shot,

10:47

what should I have for dinner?

10:49

Very difficult to answer.

10:51

If I go down the bridge here,

10:54

you're going to see areas where,

10:55

if I provide more context, you're going to get better answers.

10:58

So again, having a conversation, right?

11:00

Imagine like, I met someone recently who just got a tattoo

11:04

and I said, tell me about it.

11:05

He's like, well, I went to the tattoo place

11:07

and I sat down with the person

11:09

and we talked for 30 minutes

11:10

so they got an idea of who I was.

11:12

And based on that, they were able to design his tattoo.

11:15

Again, these don't have to be all work scenarios.

11:18

I mean, it sounds a little wonky,

11:19

but that's an example of,

11:20

you can't walk into a tattoo parlor

11:22

or a restaurant and ask for something without context.

11:25

So that's a, I thought that was a really interesting example.

11:27

So really the big takeaway here is,

11:29

because gain site co-pilot is all based on post-sale activities,

11:34

you don't have to go through all the context that I mentioned.

11:38

If you go to gain site co-pilot and you ask,

11:40

what's the best thing to put on top of vanilla ice cream,

11:43

it's probably not going to do a good job

11:45

because we spent all of our time and energy

11:47

teaching it post-sale activity.

11:50

It's going to look at insights.

11:51

It's going to look at all the information coming

11:53

and look at it specifically through the lens of risks,

11:56

opportunities, renewals, executive business reviews,

12:00

usage behavior.

12:02

So we've done a lot of that hard work behind the scenes.

12:05

And frankly, that's what a lot of these other companies

12:07

that have AI-based chat systems have done.

12:11

So while you can go to a chat GPT,

12:13

and yes, you can pour all this stuff in there

12:15

and you can get a reasonable answer.

12:17

If you use a purpose-built chat system,

12:19

like a gain site co-pilot or some of the other vendors

12:21

like Slack as an example,

12:23

the reason they work so well is because those AI agents

12:26

are specifically tuned for what you're trying to accomplish.

12:29

So they have all the context in here.

12:31

So that's one of the benefits of that we see

12:34

in using a gain site co-pilot.

12:36

So as you're building a prompt,

12:41

so a prompt is when you get the box

12:43

and we used to go to Google and type in

12:46

like some cryptic thing, right?

12:49

Dinner in Amsterdam, right?

12:51

Now when you're moving to a chat-based AI system,

12:55

it's better to ask layered questions.

12:59

So typically, somebody would ask a question

13:03

like restaurants in Amsterdam, right?

13:05

And what do you get?

13:05

You literally get like 10 million responses, right?

13:08

Because you don't provide any context.

13:10

If you're using an AI chat system,

13:13

the recommendation is, and especially if you're thinking

13:15

about the lens of customer success,

13:17

you wanna ask questions and start with something simple.

13:20

I am an executive.

13:23

I am an individual contributor.

13:26

So by providing context like that upfront,

13:29

you're basically telling the AI chat bot

13:32

the types of information you're trying to get.

13:34

Because if you just put garbage in it,

13:36

you're gonna get garbage out,

13:37

and who's been frustrated by these things, right?

13:39

They don't provide you the right answer.

13:41

It's because we haven't thought about

13:43

how to talk to these things like, not a human,

13:46

but talk to it like a robot,

13:47

or you're having a conversation where this robot

13:50

doesn't know anything about you.

13:51

So what I'd really recommend is starting off by saying,

13:54

I am this type of person,

13:55

and I'm looking to achieve this type of result.

13:58

So if you go into Gainsite co-pilot,

14:00

I am an executive, and I'm looking to do X, Y, and Z,

14:03

you're gonna get different answers.

14:04

And if you say, I'm an individual customer success manager,

14:08

I have a book of business of 50 accounts.

14:10

I'm really struggling with efficiency.

14:12

I would like three bullets.

14:15

I would like a paragraph.

14:17

I would like you to help me build a presentation.

14:21

And I've actually done this really recently, right?

14:23

Where you can actually build all these things.

14:25

So you're basically saying, I am this type of person.

14:27

I am looking to get this type of output from the discussion.

14:32

The next thing you wanna look at are things like,

14:34

where should you look?

14:35

So when you're doing this professionally,

14:37

you can add maybe links to knowledge bases.

14:41

I would like you to look in these different areas

14:43

to find the different answers.

14:45

I did this recently.

14:46

I do a lot of competitive research,

14:47

where I'll put in the competitors' support sites.

14:50

So I can have the chat GPT know the context

14:53

that I'm looking for.

14:54

So giving it sources is great.

14:56

And again, one of the benefits of the Gainsite co-pilot,

14:59

just like Microsoft co-pilot,

15:00

which is based on their system,

15:02

is it already knows these things

15:04

and it knows exactly where to look.

15:05

So if you're using Gainsite co-pilot,

15:07

it's gonna know where to look

15:09

'cause it has the corpus of information

15:10

from your Gainsite instance.

15:12

If you're doing it personally,

15:14

I would give you a recommendation

15:15

of giving it direction of where you want it to look.

15:18

And then finally, any additional context

15:20

that you want for a good output.

15:23

So these are some of the things

15:24

that we definitely recommend experimenting with

15:27

as you become more proficient

15:29

in these chat-based AI systems.

15:31

So basically the more information,

15:32

the better information you give it up front,

15:34

the better results you're gonna get,

15:36

don't forget to have a conversation with it.

15:39

Don't expect the zero shot to work perfectly.

15:42

Have a conversation,

15:43

just like if we're having a conversation

15:45

and trying to brainstorm.

15:47

So brainstorm with these things

15:48

and you're gonna get the best results.

15:51

Now, based on that AI for all program

15:56

that we're doing internally at Gainsite,

15:58

we actually did a survey just a couple months ago

16:00

where we're trying to determine

16:02

why do our teammates use AI

16:05

and what are the benefits that they're getting.

16:07

So we got about 100 people,

16:09

the top 50 or so are in this chart.

16:12

And you can see here that people are primarily using it

16:14

'cause they're trying to be more efficient.

16:16

Right, that's the outcome you wanna get.

16:19

But don't forget to use it for things like problem solving.

16:22

It can really help you if you're stuck.

16:25

Another thing we've seen a lot of benefit from

16:28

is skill extension, which is something we talk about.

16:30

So the primary use cases for AI personally are efficiency.

16:35

I wanna do more things quicker,

16:37

but skill extension is exceptional as well.

16:41

And really what that means is

16:43

if you're expected to do something

16:44

that's not in your core competency,

16:46

use these systems to help compliment

16:48

the things that you don't know

16:49

and you'll eventually learn them.

16:51

So if you're asked to kind of stretch

16:52

beyond what you normally do,

16:54

use these chat based systems

16:57

to help you do beyond what you're typically doing.

17:00

Another one that's really interesting

17:02

is this creativity and invention.

17:04

Now there's also one up here about the concept

17:07

of a white piece of paper,

17:08

but I think this is also really important

17:10

because you can use it to,

17:12

if you're kind of stuck with a white sheet

17:14

and you're not sure what to do,

17:15

use these chat based systems to help be creative,

17:18

kind of get the juice flowing.

17:20

So that's what our,

17:21

those are the things that our team has been working on.

17:24

And we're also feeding this back to our product team

17:26

to make sure that they're building,

17:27

as they're building out our own AI systems and AI tools

17:30

that they're getting these outputs.

17:34

So these are some of the things

17:34

that I definitely encourage you all to use.

17:37

So as I mentioned earlier, AI is not human, right?

17:43

It doesn't know what it doesn't know.

17:45

And while the chat GPT is off the shelf,

17:48

kind of know everything as a result,

17:50

they kind of know nothing

17:51

because they're not going to give you anything with context.

17:54

So what they want to do is when you're talking

17:56

to a chat based AI system,

17:59

it doesn't act like a human

18:01

because it wants to dump everything on you at one time

18:03

and that's very difficult to consume.

18:05

So really the recommend,

18:06

the recommendation is dissecting these large requests,

18:10

having a conversation,

18:11

don't expect it to act like a Google search.

18:15

So start small and build on it.

18:17

Another thing I'd recommend is if you're vague,

18:24

it's going to give you a vague response,

18:25

the whole garbage in job concept.

18:28

So giving a context, like as I mentioned,

18:31

Gainside Co-pilot has context of post-sale

18:33

and it knows everything about it.

18:35

If you're using personal,

18:36

if you're using a personally, give it context,

18:39

tell it why you're trying to do something,

18:42

a little bit about your background

18:44

and the things that you're trying to express

18:45

because if you tell it,

18:46

I'm trying to build a presentation,

18:48

it's actually going to help you

18:50

because it's going to know how to summarize things

18:51

and be succinct versus if you're looking to write a paper.

18:55

And the last one is off the shelf,

18:59

these systems are trained for the masses.

19:01

They're trained for a typical person.

19:03

Now, Gainside's Co-pilot is trained again for post-sale,

19:06

you're probably sick of me hearing that,

19:08

but if you're using this personally,

19:10

you want to make sure that you train it

19:11

and you tell it who you are,

19:13

what you're trying to accomplish,

19:14

and then the outputs that you want.

19:17

Now, I have a story here about how I didn't listen

19:22

to what I said and I did something that wasn't effective.

19:25

So recently, I had to send 85 different emails

19:29

to 85 different people based on some context.

19:33

So I was like, oh, I'm really smart, right?

19:35

I know how to spell AI, I can do this.

19:37

I'm going to go in there and I'm going to say,

19:39

hey, AI thing, I need you to write an email

19:42

to these 85 people and here's just a little bit

19:44

of information about each of them.

19:46

I put it all in there and it was trash.

19:49

And I struggled with it for an hour.

19:52

I'm like, this is what I want you to say.

19:53

Here is the content and here is the type of email I want.

19:57

And over and over and over, it was just providing me gibberish.

20:00

And I got frustrated because in retrospect,

20:04

I was giving me a garbage and it was giving me garbage.

20:06

I mean, that's what was happening.

20:08

I think I told Tori about this too.

20:09

So what I said is, okay, I give up,

20:11

I've been fighting with this thing for an hour,

20:13

I'm just going to start writing the emails myself.

20:16

So I'm writing the emails and I'm getting them ready

20:20

to send out later and then it dawned on me

20:23

after I did like 15 of them, I'm like, oh my gosh,

20:26

what if I were to now teach the AI system

20:30

examples of what I'm trying to accomplish?

20:34

So after I fought with it for hours

20:35

and tried to get it to teach itself,

20:38

I said, hey AI system, here is the input

20:42

that I'm trying to write emails.

20:43

And here are 15 examples of good ones

20:46

that I want you to model.

20:48

I uploaded it and it was like, bam, it was perfect.

20:51

And the reason it was perfect is because I gave it,

20:53

I started giving it more context.

20:56

I gave it great examples of what good looks like.

20:59

And I also gave it examples of what I didn't want

21:01

and it was tremendously effective to the point

21:03

that I actually built like a custom GPT

21:06

where I took those examples.

21:09

I taught it the writing style.

21:11

And in order to teach it the writing style,

21:12

what I did is I actually grabbed some of my prior writing

21:15

and I uploaded it and I said, this is how I speak.

21:19

I'm kind of laid back, I'm kind of casual in my speak.

21:22

So I taught it more about myself,

21:24

I taught it more about what I was looking to accomplish

21:26

and then I gave it real examples.

21:28

Guess what, just like a real human being.

21:30

Like I was trying to talk to this thing in an effective way

21:33

but once I taught it what I was trying to accomplish

21:35

and I talked to it like a robot and I built it up,

21:38

the results were amazing.

21:40

So another thing I'd recommend is if you have a big task,

21:43

give it examples, tell it what good looks like

21:45

and tell it what bad looks like.

21:47

And also ask it to do things in iteratively.

21:50

So if you're doing like an email project like this,

21:52

what I would say is, write one of these for me

21:56

and I'm gonna give you feedback.

21:57

It would write it for me and I would give it feedback.

21:58

So again, do it in baby steps.

22:01

Do it just like you're talking to a person.

22:03

Don't expect it to be perfect right out of the box.

22:06

Okay, so co-pilot itself.

22:11

So as I mentioned, the Gainsite co-pilot,

22:14

is anyone in the beta?

22:15

Okay, one person, we only have 50 people

22:18

so you're the winner.

22:20

So what's super cool about it is we're actually doing,

22:22

we're doing a beta right now, kind of a limited result

22:26

and what we're trying to do is get feedback.

22:28

So we're using it internally at Gainsite

22:30

we're giving feedback back to our AI teams

22:32

and we're also having a select group of customers

22:35

in this beta as well.

22:37

And as I mentioned, what we're trying to do

22:39

with the product is, it's actually available,

22:44

let me start over.

22:46

It's actually able to find information

22:48

about one specific account.

22:50

So as an example, when I was out there at the EMC desk,

22:53

trying to get information, so I said,

22:54

give me some information about this account

22:56

and it was able to prep me.

22:59

Something else that co-pilot is able to do

23:01

is look at multiple accounts.

23:04

So if I am permissioned through role-based access permissions,

23:07

I can not only ask, how is ABC company doing,

23:11

I can ask broader questions.

23:12

Can you tell me what other customers are like ABC

23:16

and that are trying to accomplish this type of task?

23:18

I've also done things in there where I've said,

23:20

hey, I'm talking to a prospect who has this use case,

23:25

tell me other customers who have a similar use case.

23:29

So I'm using that to maybe mine references.

23:32

I'm also using it so I can name drop accounts.

23:35

So I definitely encourage you to do that if you're a CSM

23:37

or even a seller.

23:38

The last thing we're talking about is advice

23:43

based on your role.

23:44

So the system, co-pilot is already trained

23:47

on how to speak to executives, how to speak to CSMs.

23:51

So when you're using co-pilot or really any of these systems,

23:54

I wanna make sure that you look at

23:55

and you give it information about who you are

23:57

and you're gonna get a better output.

23:59

So again, use it to find information about your accounts

24:02

and also use it to find information

24:04

about your best practices.

24:06

So how is it actually working behind the scenes?

24:10

So who has security people breathing down your neck?

24:13

I'm not letting you use this thing, everybody.

24:16

So we have tons of information about our security,

24:18

it's super safe, but I'll give you the walking away message

24:22

is behind the scenes we're actually using chat GPT-40.

24:26

We're using a combination of chat GPT delivered directly

24:30

from OpenAI and also Microsoft,

24:31

they run on their Azure information.

24:34

Currently we're looking at timeline information,

24:37

we hope to expand that in the future

24:39

and in the future we're gonna be looking at things

24:40

like different objects, account objects,

24:43

opportunity objects, custom objects,

24:45

but we gotta start somewhere.

24:46

So currently it's looking just at timeline.

24:50

The way that it's working today is,

24:52

it's also for your security folks,

24:54

it's also looking at the information

24:55

that you have access to.

24:56

'Cause a lot of security teams are gonna say,

24:59

I don't want you to have access to all the information.

25:01

So what's really interesting about it is right out of the bat,

25:04

we're also making sure that we're granting access

25:07

based on your ability.

25:09

So if you're an executive, you can see everything,

25:11

but we're making sure that we're also restricting access

25:14

so we don't give people insights

25:15

that they shouldn't have access to.

25:19

So this is just one example of how the system is working.

25:23

So the input query is coming in,

25:25

we're consuming all these different data sources

25:28

on the left-hand side, you can see some of the ones

25:30

that are coming up.

25:31

We have all this context built into it, right?

25:33

The context of being post-sale,

25:35

it knows how to ask for clarifying questions.

25:38

And then finally it's gonna give you a response.

25:39

So this is what we're doing right now.

25:42

- Insights co-pilot lets you skip searching

25:47

and get instant answers.

25:49

Let me set this up first.

25:51

This is just a quick demo example

25:52

from our co-host Tori here,

25:55

which is gonna kinda go through an example

25:57

of how we're using co-pilot today.

25:59

So I'll start this one over.

26:01

- Insights co-pilot lets you skip searching

26:06

and get instant answers to your

26:07

pressing customer questions.

26:09

Spend less time clicking with summaries

26:11

and link timeline entries

26:12

to dive deeper into customer content.

26:15

Ask follow-up questions and let co-pilot work for you

26:17

to eliminate grunt work.

26:20

(upbeat music)

26:22

For execs, quickly assess what customers need your attention

26:25

and get more information about risk situations

26:27

so you can be best prepared to take action.

26:29

Let other teams such as product

26:32

benefit from your collected customer data

26:34

by better understanding feature needs

26:36

and roadmap prioritization

26:38

or finding customer references.

26:39

With co-pilot, we're super-charging gainsite

26:43

with a whole new way to engage

26:45

and ultimately change how our customers deliver

26:48

customer success.

26:50

- So I wanna show that for two reasons.

26:51

Number one is it makes Tori really uncomfortable

26:54

to hear her voice and to watch her do presentations.

26:57

So that's my number one goal.

26:58

The number two goal is did you notice

27:00

that when she was doing the searches,

27:02

it asks follow-up questions?

27:03

Like do you want more information on X, Y, and Z?

27:07

The other thing I thought was super interesting

27:08

is it gives you, it cites the source, right?

27:12

Because a lot of us are like,

27:13

"Where are you getting this information?

27:14

Are you making it up?"

27:15

So it's super interesting.

27:17

I don't know if you saw on the bottom,

27:18

but it would give you the answer

27:19

and then there were links.

27:21

I got this from timeline.

27:23

I got this from this area.

27:24

I got this from that area.

27:25

And then when you click on that,

27:26

it actually gives you the root of the information.

27:29

Because it's natural to look at these things

27:32

and say like, "What did you get this information?"

27:33

And when the boss asks you, you're like,

27:34

"I don't know, the chat GPT gave it to me."

27:37

So we thought it was really important

27:38

to make sure we cite the sources

27:40

so you're confident in the information

27:41

and you can actually go down.

27:42

You can actually get into the details.

27:46

(chuckles)

27:48

Okay, so a couple more,

27:49

then we'll have some questions and answers.

27:52

We also did a study about six months ago,

27:55

actually with our customers,

27:57

to try to figure out,

27:58

what should we actually be building?

27:59

So not only did we do research internally

28:02

to see what our team and our CSM's were looking for,

28:05

but we also asked our customers,

28:07

like what are the things you're looking to accomplish

28:09

and what are the things that you

28:11

are currently doing today?

28:12

And really the lion's share was about data analysis

28:14

and reporting.

28:16

What are we doing with this?

28:17

We're actually building,

28:18

and a few of us have access to a generative AI reporting system,

28:23

where you can actually go in there and say,

28:27

"I would like a report to do this."

28:30

It is integrating and basically writing code behind the scenes

28:34

and building reports.

28:35

So this is something we're working on today.

28:37

Deeper customer insights.

28:38

What we're doing with this is,

28:40

anyone who use cheat sheet,

28:42

that's a new feature we have.

28:43

What cheat sheet does is it basically

28:44

looks through timeline, again, those notes,

28:46

and it gives you kind of a recap

28:48

of everything about an account.

28:50

So based on this information,

28:51

we're looking at more objects, more insights,

28:53

because we want to get deeper into the insights.

28:55

Faster information on calls.

28:56

I think we're also doing that with cheat sheet today.

29:00

Like Nick, our CEO in the old days,

29:01

you'd have to fill out a piece of paper.

29:04

Now he just knows to go to cheat sheet,

29:07

and he's going to find exactly what he needed.

29:09

I don't have a good answer on contract management yet,

29:11

but that's kind of low on the list.

29:13

So I just want to make sure everyone understood

29:15

that we're asking these questions,

29:16

because we're actually doing things with them.

29:17

So this is just one way we're trying to learn

29:19

from our community.

29:20

Okay, so as we wrap up here,

29:23

and then we'll have about 15 minutes for questions,

29:25

suggested prompts.

29:27

So I'm going to look at it through three different lenses.

29:29

Number one, I need strategy advice.

29:32

Like how do I do this as a CSM, et cetera, et cetera.

29:35

I'm also going to talk about how you can get better

29:37

information about your customers,

29:38

and also your book of business.

29:40

So here's an example of some of the questions

29:44

that you can ask.

29:45

I particularly like this one,

29:46

like what topics have trending negative

29:49

or positive sentiment.

29:50

The reason I like that one is because you can kind of throw

29:53

a net across all your customers,

29:55

and as a leader, you can determine what are the macro things

29:58

that are either positive or negative with my business.

30:01

So I thought that was a really interesting examples.

30:04

You can also explore using feature information.

30:09

So as we start expanding and getting usage statistics in here,

30:13

your product teams may also use this to determine

30:15

what are the features that our people are using.

30:18

So these are some great examples

30:19

of how to get strategy advice.

30:21

As we kind of go down a little bit further,

30:23

account level information, something I did today.

30:26

Suggest similar customers to this one, right?

30:29

So if you're going to have a meeting with the prospect,

30:31

and you say this prospect is looking to have efficiency,

30:33

they're looking to send emails,

30:34

they're looking to do X or Y,

30:37

you can actually go in and ask these questions,

30:38

and you're going to find similar use cases,

30:42

and potentially similar customers.

30:44

So I think this is a really good one as well.

30:46

Something else that our CSMs love

30:48

are the top business objectives.

30:50

So if you're prepping for a call,

30:54

this is a great way to figure out what is this customer,

30:56

what have they been talking about

30:58

through our relationship and timeline.

31:00

So that way when you have these discussions,

31:02

you actually know exactly what they're trying to accomplish,

31:04

and you can see some of the things

31:05

that they have accomplished.

31:08

And then the last one, things like role levels.

31:10

So if you're an individual contributor,

31:12

like I mentioned, you're a manager, you're an executive,

31:15

you can use these type of prompts to figure out

31:18

like how should I measure myself,

31:20

how should I measure my customers,

31:22

how should I measure my team.

31:24

So another great example of using a prompting

31:26

to figure out what are the best practices,

31:28

and what are some other people doing that I would recommend.

31:31

Okay, so as we wrap up here in our slide,

31:35

we have 13 minutes for questions.

31:38

So top takeaways, when you have a chat GPT

31:41

and you wanna ask prompts, this thing is a robot,

31:44

you need to teach it, imagine it's like a three year old child,

31:46

right, you need to build it up and help it understand

31:49

what you're trying to accomplish.

31:51

Number one, I would definitely look for

31:53

purpose built AI systems,

31:55

that's why you see all of the different software vendors

31:58

building them, because they are tuned,

32:00

just like I mentioned, co-pilot is,

32:01

and they're gonna give you better results

32:02

through the lens of what we do.

32:05

And then the last one is about prompting,

32:07

get asked the right questions, experiment,

32:10

learn from me, learn from others to figure out

32:12

how you find that needle in a haystack.

32:14

And I think we're up to Q and A.

32:17

(audience applauds)

32:19

- Thank you so much.

32:21

Awesome, massive round of applause,

32:25

like her study's been at Gansite what, 10 years?

32:27

- 100 or 10, something like that, yeah.

32:29

- 100, feels like 100, baby.

32:30

- Feels like 100, but yeah.

32:32

- But super excited to have a couple questions here

32:35

for her to answer on anything Gansite,

32:37

but particularly co-pilot.

32:39

Couple questions, I'll just go ahead and read these, Harsh,

32:41

if you wanna give us a great answer.

32:44

Is co-pilot already available, or still on the roadmap,

32:46

does it require a separate license or included?

32:48

- I'm so happy, I know the answer to this one.

32:50

So co-pilot is currently in beta,

32:53

we have 50 customers using it, it will open up,

32:55

I just don't know when, just to be straight with you,

32:57

but I know it will eventually.

32:59

So yes on the roadmap, but it is for real, I'm not up.

33:03

What do I say, Tori?

33:04

Ship it or zip it?

33:05

I'm a ship it or zip it fan,

33:07

like I don't wanna talk about things that aren't real,

33:08

it is real, people are using it.

33:10

It's gonna get better before we release it.

33:13

The second question on there is about,

33:15

does it require a separate license?

33:17

No, we made the decision over a year ago

33:20

that the core Gansite customer success platform

33:23

is gonna have AI built into it.

33:25

The reason is we wanna make sure that you're getting value,

33:28

we wanna make sure that we're making it easier

33:30

for you to use the Gansite system.

33:32

And just to be straight with you,

33:33

you heard about staircase this morning.

33:35

So staircase is a separate product,

33:37

it kinda does a different thing,

33:39

even though it's kinda still within the lens

33:40

of customer success and that is a separately licensed product.

33:44

- I'll add just one other point there, license wise,

33:47

it is available for viewer licenses

33:49

within Gansite CS as well.

33:51

So you don't have to have a full licensed

33:52

to leverage co-pilot,

33:53

anyone that has access to Gansite

33:55

can leverage that functionality.

33:56

- Yeah, we have about, I think, 10 different AI live,

33:59

real AI things out there today.

34:01

So yes, everyone talks about it on stage,

34:03

we actually have real stuff, which makes me happy.

34:05

We have the ability to write emails,

34:07

we have the cheat sheet,

34:09

we have sentiment analysis in surveys,

34:12

we have likelihood to renew scores,

34:13

so we have about 10 things like out the door.

34:15

- And Hirsch does say shippeter's a bit several times.

34:18

- Yeah.

34:19

- So next question here,

34:20

what are the plans to plug in external sources

34:22

to co-pilot, like confluence,

34:24

and desk articles, et cetera?

34:25

- Okay, good question.

34:26

I know, I don't know the answer on timing,

34:29

I know it is something that they're thinking about,

34:32

the things that they're looking at first

34:34

are exposing more areas of core Gansite to co-pilot.

34:38

I'm sorry, you have to co-pilot.

34:40

So first, as timeline as I mentioned,

34:42

they're looking to expand it to the other objects,

34:44

which is just more data.

34:45

I don't have an answer on the other ones,

34:47

do you, Tori?

34:49

- I don't know specifically,

34:50

but one thing that we're leveraging co-pilot for

34:52

is Gansite source content, right?

34:54

Like supporting through Gansite,

34:55

so the natural extension of that

34:56

is to leverage other sources similarly.

34:59

But like her said,

35:00

I think the initial focus is just our own internal data.

35:02

- Mm-hmm.

35:03

- Great question though.

35:04

Next question, how much of co-pilot is based on best practices

35:09

and how much can we train to our needs?

35:11

- That's a good one.

35:14

I know half the answer.

35:15

I know that currently it is all based on best practices

35:19

that are based on things that we have taught.

35:21

There's no training needed,

35:23

and frankly, I'm not sure if it can be trained

35:25

for your specific use cases.

35:27

What it is is it's designed based on your use case

35:30

and your instance of Gansite.

35:32

So it is a tenant, right?

35:35

A little piece that is tuned for your environment,

35:37

and it's not sharing information with other environments.

35:40

So it's really gonna only be trained for you,

35:42

but I don't have a good answer on whether or not

35:43

you can train it for additional insights.

35:46

Although when we leave,

35:47

we're gonna tell Shonton,

35:48

who's our product manager, that that's on the list.

35:50

- Yeah.

35:51

- I will also add it, it is trained like her said

35:53

on your instance.

35:54

So if you have in your data somewhere flagged,

35:57

like certain risks are high priority or things like that,

35:59

co-pilot will know that information right now.

36:02

But as far as other stuff,

36:03

I think that's probably more good now.

36:05

Next question, did you add any sense of humor to co-pilot?

36:09

I love this question.

36:10

- Yeah, I definitely have had conversations with Shonton

36:12

about adding fun Easter eggs in there.

36:15

If you tell it the tone in which you would like to talk to it,

36:19

the tone will adjust.

36:21

He probably hasn't listened to any of my dumb ideas

36:24

about putting jokes in there,

36:26

but maybe I will have to tell them now.

36:27

So just like I mentioned,

36:28

I'm an executive, I'm an individual contributor.

36:30

If you tell it, I'm a casual person,

36:33

I like to make jokes, do it in a riddle.

36:35

Nick, ask it to write in haikus.

36:38

So whatever output you want, you can tell it.

36:41

It just, unfortunately, they didn't listen to me

36:43

and make it all funny.

36:44

- There are some fun prompts in there,

36:45

like while you're waiting in the loading screens, but--

36:47

- Oh, that's true, yeah.

36:49

Great question, though.

36:51

The next one, is it possible to extend

36:52

co-pilot training data sources

36:54

with other company sources, mail ERP CRM?

36:58

- The answer to that one is not yet.

37:00

It's currently only looking at gain site data.

37:03

Period.

37:06

The next part of that is the staircase product

37:08

is actually trained on a lot more data sources

37:12

because it's really intended to do,

37:13

it's intended to be more of like a management product

37:16

where managers can kind of look at all the communications

37:20

that are having across the organization

37:22

and then immediately figure out if there is, you know,

37:24

negative or positive sentiment.

37:26

Staircase currently looks at a lot of different data sources

37:29

and my hope is that eventually it'll make it into CS.

37:33

Hopefully I said, save Harbor earlier,

37:35

so you don't say, hey, that Jason guy told me it's in there,

37:38

but I know that's one of the goals.

37:40

- And it does search timeline right now,

37:41

so if you have that data in your timeline,

37:43

like if you're logging that in timeline,

37:45

then it will capture that data today.

37:47

- Mm-hmm, yeah, one of the common things I see,

37:49

you're like right now we have out of the boxy ability

37:51

to bring in like your Zoom calls, your GON calls,

37:54

your emails, all of your notes.

37:56

They can also be synchronized with like CRM systems,

37:59

so you can actually sync it over from Salesforce.

38:01

So if that information is in there,

38:02

like Tori said, absolutely.

38:04

Can it pop out to like an SAP ERP system, not yet,

38:09

unless it is brought into gain site?

38:10

- Awesome, and I think you kind of answered this next question

38:13

a little bit as well, Hersh.

38:15

It sounds like Copilot will be able to access

38:16

attached GON recordings, is able to read attached file types

38:20

like GON PDF PowerPoint emails.

38:23

- So yes, we have an integration with GON and Zoom.

38:26

What that does is it brings in their verbatim transcript

38:30

into gain site timeline.

38:32

Think about like a little note, right?

38:33

So it's gonna have the verbatim of all the content,

38:37

it's gonna have a video from those different,

38:40

it's gonna also have the video,

38:42

and what timeline also does is it does a recap.

38:44

So it says based on this phone call,

38:46

here are the five things Tori needs to do,

38:48

here are the five things Jason needs to do,

38:50

by the way this had a negative sentiment

38:51

because they said they were going out for an RFP.

38:55

So it'll be able to look through all those different things

38:58

today, it'll also look through PDFs in timeline,

39:02

so if there are PDFs in timeline,

39:04

it will look through the content of that as well.

39:06

- Awesome, next question.

39:09

Are these recommendations only applicable to Copilot

39:12

for other AI chat tools as well?

39:14

- The recommendations I gave,

39:16

you don't have to actually be as explicit in Copilot.

39:20

The reason is, is because we have basically tuned it

39:22

to be all about post-sale.

39:25

The recommendations I gave you will help it,

39:28

and I would say it is necessary when you're using more,

39:32

like if you just go to chat GBT and use our typing stuff in,

39:36

I would definitely recommend some of the recommendations

39:39

I made about content and et cetera.

39:41

- Copilot will know a lot of this stuff out of the box,

39:44

but the more context you give it,

39:45

the better outcomes you're gonna get.

39:47

- Awesome, next question.

39:49

Does Copilot learn about me?

39:51

If I tell it that I'm a vegetarian,

39:53

to use your example and a manager,

39:54

would it save this information and use it next time?

39:57

- Yes, Copilot will remember things that you told it,

40:00

but it only does it for a month.

40:02

So it's like a rolling month,

40:03

so if you tell it you're a vegetarian monthly,

40:05

it'll probably remember, but it will learn who you are,

40:08

and it will remember the fact that you had

40:10

a sense of humor.

40:12

What it also does is on the left-hand side,

40:15

you might've noticed there's like different,

40:17

like with chat GBT, there's like different chats,

40:20

so you can actually go back and look at a prior chat,

40:23

and it will have the context from that conversation.

40:26

So Tori and I were experimenting with the other day,

40:28

and we asked questions like,

40:29

"What's the best practice for this?"

40:31

New chat, tell me about ABC Company,

40:34

and I could go back to either of those independently,

40:36

and it's gonna have memory of those conversations.

40:39

- Yeah, which is great if you have multiple customers

40:40

and you wanna have a chat per customer,

40:42

that way it's easy for you to scroll back

40:43

and see what other questions you've asked in the past.

40:46

Great question so far.

40:49

Next one, "Would uploading bulk males to timeline

40:52

"make sense to benefit from Co-Pilot

40:54

"if things weren't logged before?"

40:55

- Yeah, totally.

40:57

What I've seen some customers do is they synchronize,

41:00

as an example, a lot of customers have emails in their CRM.

41:05

So as an example, I know Salesforce has a basic button

41:08

you can press, and it basically syncs them all over.

41:10

So bringing all that information

41:12

will absolutely help Co-Pilot.

41:15

There's also the ability to,

41:17

for those of you who don't want a disaster in your timeline,

41:20

you can actually code the,

41:21

I'm sorry, tag them as a specific topic.

41:25

That way you could potentially filter them out

41:27

if you're reading through there.

41:28

- Awesome.

41:30

Next question.

41:32

Oh, okay.

41:33

Still, but real question.

41:34

Do we need to add please when we talk to AI?

41:39

No, but I always do.

41:40

Like I always say please, I always say thank you.

41:43

There's also this theory that,

41:45

I'm not sure it's been debunked yet,

41:46

but if you are polite and you incent it,

41:51

it will act differently.

41:53

So there is research that no one knows why,

41:56

but if you say stuff like, you know,

41:58

if you do this, you'll be the winner.

42:00

If you do this, you'll get a special prize.

42:02

It sounds ridiculous, but some of the concepts

42:05

actually give you better results

42:06

if it knows that it's gonna get a good outcome.

42:10

So Google that when you're all done,

42:11

or check in chat, GPT, it's bizarre.

42:14

And I've experimented with it, but it seems to work.

42:18

- Awesome.

42:19

That was a good question, a real question.

42:20

Next one, can you tell us about the safeguards

42:23

regarding data access that Gainsite is implementing?

42:26

- Oh, sure.

42:26

So within Gainsite, you have the ability

42:30

to have different types of access control.

42:34

The access control can be at a account type level.

42:39

So if you have Enterprise and SMB,

42:41

I can say Tori, if it's tagged SMB,

42:44

she can see all of it, and I can say period, right?

42:48

So it's a definite, and she will not have access to SMB.

42:52

So within there, you do have role-based access control

42:55

based on the person, and also based on the account.

42:58

I can also go in there and say IBM as a customer

43:01

has sensitive information, therefore only

43:04

Nick meta has access to it.

43:05

So you can actually do role-based

43:08

and also account-based access controls.

43:10

And then when you use the chat-based AI,

43:12

if I ask a question about IBM,

43:14

because I'm not logged in as Nick meta,

43:16

it will not allow me to have those insights.

43:18

So as a manager, if you implement those types of strategies,

43:22

you can look at things more holistically,

43:24

tell me more customers like this.

43:26

If you're an individual and you have 50 accounts,

43:28

it will only look at those 50 accounts.

43:32

- Awesome, great questions.

43:33

I think we have time for maybe two more.

43:35

Very similar to the last question about security.

43:37

Is Co-pilot powered by a third-party provider?

43:39

Will they get our clients data or train on that data?

43:41

- Oh sure, okay, so Co-pilot, the brains behind that,

43:44

we use chat GBT directly through OpenAI

43:49

and also through Microsoft Azure.

43:51

So that is the brains behind the operation.

43:55

On top of that, we do train it

43:58

based on what customer success means.

44:00

And yet the part of me, I don't remember the correct phrasing,

44:03

but we basically train those different models.

44:06

The information is sent to those vendors

44:09

in order to accomplish the transactions.

44:13

The data is not stored in memory.

44:16

We also have DPAs and the other acronym is escaping me,

44:21

but we have all of the security protocols

44:24

that I have not seen any CIOs reject.

44:27

It's kind of the basics of data protection.

44:30

Oh, and on training, you can choose to not have it trained

44:34

based on your data.

44:36

- Yep, and it's also not cross-tenant as well.

44:38

So it'll only like within your gain site instance,

44:40

no data is ever shared with another gain site instance.

44:43

I'm actually gonna skip this question

44:45

'cause I think it's kind of similar,

44:46

but I wanna get to the one below it.

44:48

How can gain site support and overcoming

44:49

data protection challenges related to leveraging

44:52

and integrating AI that we typically face in Europe?

44:54

Any best practices available?

44:56

- I might need to grab somebody else

44:59

to help you with that one.

44:59

So I'll always be straight with you.

45:01

I don't have a good answer for that one.

45:02

I do know that we currently host

45:04

our European customers in Germany.

45:07

We have a German tenant, we have a German AWS instance,

45:10

and we also use a German,

45:12

I believe we use a German open AI tenant,

45:14

but I have to double check on that one.

45:16

But if anyone's interested,

45:17

I'm happy to provide all the documentation

45:19

and we can kind of pour through it together.

45:20

- Yeah, and if you go to,

45:22

quick plug, if you go to gainsite.com/security/ai,

45:25

we have a whole bunch of resources

45:27

that do provide a lot of good data,

45:28

like Hersh was just describing in terms of

45:30

what security protocols we have,

45:32

all the different hosting and other parameters

45:33

for EU customers, European customers,

45:36

and all of that as well.

45:37

- Yeah, super important,

45:38

and if you do not want the data going back

45:40

into the model, your security team

45:43

can absolutely ask for a not to.

45:45

- Awesome, well, I think we are out of time.

45:47

So many great questions, so much great information

45:50

and knowledge that Jason shared with us today,

45:52

very fortunate to have him on stage.

45:54

I think it is time for lunch from an agenda perspective.

45:58

So if you guys wanna head out,

45:59

we actually get first dibs on lunch,

46:00

sessions are staggered, so you all get to be first,

46:03

which means you avoid all of the lines, which is a huge win.

46:06

So I highly recommend going to grab some food

46:08

before lines you start to grow there.

46:11

Hopefully you'll join us again on this session,

46:12

we have an awesome panel coming up after lunch,

46:14

but otherwise, I hope you all have an amazing pulse

46:16

and thank you all so much.

46:17

- Yeah, thank you for coming.

46:18

- You're a proud of applause.

46:19

(applause)