The Evolution of Google Cloud GTM with GenAI

0:00

All right.

0:08

Let me get the slides up.

0:11

All right.

0:12

So, hello everyone and thank you for joining the session.

0:17

And for those who I have been met, I'm dying.

0:20

I'm the managing director for Google Cloud Marketplace and ISV Go-to-Market

0:23

Initiatives.

0:24

So I'm very happy to be here with you today.

0:27

So just a little bit about me.

0:30

So my team's primary charter is to grow the partner business through Cloud

0:34

Marketplace.

0:35

So I manage teams responsible for business development, partner engineering and

0:40

onboarding,

0:41

partner and platform strategy, Go-to-Market Initiatives, including scaling our

0:45

indirect

0:46

channel incentives and co-so initiatives and field engagement.

0:50

And I've been at Google for about nine and a half years.

0:54

And before taking this role about two and a half years ago, I spent four years

0:57

in the

0:57

product group and area of the business called the Application Modernization

1:01

Platform,

1:01

scaling our Kubernetes, serverless and developer platform businesses and boot

1:06

strapping our

1:07

multi-cloud initiatives with Anthos.

1:09

So we're here to talk about JNI.

1:12

It promises to be the transformative technology of our time.

1:15

So I think we're experiencing a whirlwind pace of innovation and it's impacting

1:21

all

1:21

industries and business functions.

1:23

It will certainly impact Cloud Go-to-Market.

1:26

And we've already been seeing over the last few years how we're seeing a

1:30

fundamental transformation

1:32

in terms of how business software is being bought and sold via Cloud

1:35

Marketplace and

1:36

how software companies are aligning with the Cloud providers to drive very

1:39

efficient routes

1:40

to market.

1:41

So certainly Go-to-Market with Google Cloud will evolve with JNI.

1:45

So very excited to spend the next 20 to 25 minutes or so to talk about this

1:49

exciting evolution.

1:51

So with that, let's dive in.

1:56

Okay.

1:57

So I think pundits and analysts are predicting that JNI has a big impact on

2:06

productivity.

2:08

And I think you can see we're potentially going to add trillions of dollars to

2:11

the global

2:12

economy.

2:13

And I truly believe that this has potential because what truly makes this

2:17

transformative

2:18

is because the previous waves of automation technology have mostly affected

2:22

physical work

2:23

activities, but JNI is likely to have really the biggest impact on knowledge

2:28

work.

2:28

So especially activities like decision making and collaboration because it can

2:32

predict patterns

2:34

and natural language and use it dynamically.

2:36

So in the last year, I think, you know, the world in the industry was really

2:40

beginning

2:41

to imagine how JNI could transform businesses.

2:43

And what we're seeing today is that transformation is underway.

2:46

So consistently, what we're seeing in hearing from customers and partners is

2:52

that 2024 is

2:53

shaping to be the year where companies are moving from experimentation and

2:56

proof of concepts

2:57

to production at scale.

2:59

So we're already seeing a lot of innovation, hundreds of technology and

3:02

services partners

3:03

have built solutions with our customers, either leveraging Google Cloud AI

3:07

technology or

3:09

cloud infrastructure.

3:10

So things like foundation models and chatbots, intelligent assistants and

3:14

others.

3:14

And across these different industries, we're seeing well-defined use cases

3:19

forming.

3:19

So for example, in financial services, we're seeing things like fraud detection

3:24

, risk assessment

3:25

and customer service.

3:27

And as an example, Scotiabank is leveraging data for predictive offers.

3:30

So they're improving these customer interactions through AI and then unifying

3:34

the data silos

3:35

across their organization.

3:37

In retail and consumer packaged goods, there's a lot of focus on things like

3:41

search and recommendation,

3:43

how to improve customer support, how to optimize pricing.

3:46

And so as an example, you'll see Home Depot has built an application called

3:51

Sidekick,

3:52

which really helps store associates manage inventory and keeps the shelves

3:56

stocked and

3:56

they're using vision models to sort of drive these associates to prioritize

4:00

what actions

4:00

to take.

4:02

In the digital enterprise space, not surprisingly, they're leveraging things

4:05

like AI assisted

4:06

software development, simplified DevOps, improving some of the back office

4:11

productivity.

4:12

And so as an example, GitLab is using Google IJNA technology to automate code

4:16

reviews and

4:17

improve developer productivity.

4:20

In the median entertainment area, I think there are opportunities around

4:23

content creation,

4:24

personalization and ad optimization.

4:28

And then the healthcare and medicine area, we're seeing JNAI being used for

4:31

like disease

4:32

diagnosis, drug discovery and other areas.

4:34

And as an example, Mayo Clinic, they're enabling JNAI-powered search to help

4:39

clinicians find,

4:40

understand and interpret information.

4:42

So these are just a few examples, but you can see that JNAI has the potential

4:46

just to unleash

4:47

a lot of innovation, find new ways of working, amplify some of the existing

4:52

systems and technologies

4:54

and also transform enterprises across all of these different industry verticals

4:59

So click ahead.

5:01

So Vertex AI is our enterprise AI platform.

5:04

So let me start at the bottom and work my way up.

5:07

So it sits on a world class infrastructure and it's a unified platform that

5:12

lets customers

5:13

discover, customize, augment, deploy and manage JNAI models.

5:19

So if you look, we have 130 models as part of our model garden, including the

5:24

recently

5:25

announced JNAI 1.5 Pro.

5:27

We have leading partner models from anthropic, AI21 labs, cohere and many

5:32

others and then

5:33

also popular open source models, including JAMA, LAMA2 and NISTRO.

5:38

And what Vertex AI does is allows you to tune the foundation models that you've

5:43

chosen

5:44

with your own data.

5:45

And we have a variety of different techniques, including fine tuning,

5:50

reinforcement learning

5:51

with human feedback, distilling, supervise and adapter based tuning techniques.

5:56

And of course, customers get far more from their models when they can augment

6:01

and ground

6:02

them with their own enterprise data.

6:03

So Vertex AI helps you manage tooling for extensions, function calling,

6:08

grounding and

6:09

then once you have chosen the right model, it's been tuned, it's been grounded.

6:13

Vertex AI can help you deploy, manage and monitor those models.

6:17

And then finally at the top, you'll see that we have Vertex AI agent builder,

6:21

which really

6:22

brings together everything, the foundation models, Google search, other

6:26

developer tooling

6:28

and it really makes you easy to build and deploy agents and agents really are

6:32

helping

6:33

users achieve very specific goals.

6:36

And they can understand these multimodal information, whether it's processing

6:40

video,

6:41

audio, text together, connecting it and rationalizing across different inputs.

6:45

And they can learn over time and facilitate these transactions and business

6:48

processes.

6:49

And what we're seeing is that many organizations are building AI agents that

6:53

serve customers,

6:54

they support employees, they can help them create content, they can accelerate

6:59

software

6:59

development, they can unlock the potential with data and they can also improve

7:03

our security

7:03

posture and much, much more.

7:06

So and of course at the very top here is potentially the most important part of

7:10

this discussion,

7:11

which is our ecosystem of partners.

7:14

So click ahead.

7:16

Now to adopt JNI broadly, I think customers really need not only enterprise

7:22

platform

7:23

that provides a broadest set of end-to-end capabilities, but it's optimized for

7:27

cost

7:27

and performance.

7:28

And it's an open platform.

7:30

It really offers choice.

7:31

So it's easy to integrate with existing systems.

7:34

And of course it's supported by the broadest ecosystem of partners.

7:37

And so what you see here is that we offer a choice, a first party and ext

7:41

ensible partner

7:43

enable solutions at every layer of this AI stack and really provides a choice

7:47

across infrastructure,

7:50

models, data solutions, AI tooling and help customers really build those

7:55

applications

7:56

and create business value.

7:57

So broadly speaking, our partners really have this tremendous opportunity to

8:02

help customers

8:03

transform this expansive open AI platform and we're really committed to

8:08

supporting our

8:09

partners at every layer of the stack.

8:11

And of course we have a partner-led approach to services delivery.

8:15

So again, let me start at the bottom and work my way up.

8:18

So I think at first if you look at the foundation models on Google Cloud, from

8:22

the start we've

8:22

really prioritized giving customers and partners access to a very broad set of

8:27

curated AI models.

8:28

And as I mentioned earlier, 130 foundation models, including our first party

8:32

models,

8:33

open models and then popular third party models.

8:36

In March we actually announced that anthropics latest, a large language models,

8:41

Cloud 3,

8:42

SONIT and HIKU are now available both on model, garden and Google Cloud

8:45

marketplace.

8:46

So we're very committed to making it easy for developers to use Google Cloud

8:50

infrastructure

8:51

for training and inference.

8:53

So we have a wide range of open models that can leverage as well.

8:57

So in January we announced a partnership with HuggingFace which allows

9:00

developers to quickly

9:01

deploy hundreds of thousands of models through the HuggingFace platform on

9:06

Google Cloud.

9:07

Now if you move up and look at technology and platform partners, we give

9:11

customers and

9:12

partners the freedom of choice of infrastructure they want to use the best

9:15

suits they need.

9:16

So we have a selection of tens of processing units, we have NVIDIA GPUs, we

9:22

announced at

9:23

next the TPU V5P which is our fastest powerful TPU to date.

9:28

And we've expanded our partnerships with NVIDIA.

9:30

In fact, we're going to be the first Cloud provider to make the Grace Blackwell

9:33

platform

9:33

available to our customers.

9:35

And of course customers can also access the A3 mega instance which is powered

9:40

by the NVIDIA

9:41

H100 tensor core GPUs and has doubled the energetic bandwidth speed of the

9:45

normal A3 instances.

9:47

Now of course if you go up, every AI project starts with getting a handle on

9:52

their data

9:53

estates.

9:54

So our customers and partners are certainly using BigQuery, the leading data

9:57

warehouse

9:58

service to build the underlying data infrastructure for AI implementation.

10:02

But we also have a very open approach as well.

10:04

So we can connect the third party data platforms with partners like Confluent

10:08

and Databricks

10:09

and Elastic, MongoDB and others.

10:12

And customers or partners are also connecting their enterprise applications

10:15

such as like

10:15

Workday and Salesforce with Google databases like AlloyDB.

10:20

So I think in short you can think of Vertex AI just really lets customers

10:24

connect their

10:25

AI models to the data platform and databases as well as third party platforms

10:29

ensuring that

10:30

they're grounded and have the most relevant business data which of course

10:33

delivers the

10:34

best results.

10:36

Now moving up, if you look at developer tooling and applications, we want to be

10:42

the place

10:43

to drive customers, partners and developers to build these AI models,

10:47

capabilities and

10:48

applications.

10:49

And so we've developed helping developers with JNI and we've introduced our

10:53

state of

10:53

the art models and incorporated into popular developer tools like Colabs, VS

10:58

Code, JetBrains,

10:59

Replits, Stack Overflow and others.

11:01

And of course as you go up, it's not just about technology and tooling

11:06

companies, we're

11:07

also partnering with the market leading technology and ISV partners to embrace

11:11

Gemini models and

11:13

bring it into the capabilities of the projects and services.

11:16

So companies like Salesforce and Workday and Canva and UKG, OpenText, HubSpot,

11:23

these partners

11:23

are using our GNI capabilities to launch important customer facing workflows

11:28

like summarizing

11:29

documents, building job descriptions from scratch, helping legal teams parse

11:34

through

11:34

contractual languages and of course the cybersecurity and data analytics side

11:39

and DevOps as well.

11:41

Partners are incorporating even more AI powered capabilities into the products.

11:45

So companies like Palo Alto Networks and CrowdStrike and Exibim and Optiv and

11:51

Altirix and Dynatrace,

11:54

are launching new features built with Google Cloud AI to help their customers

11:58

drive more

11:59

value from business data, improve the productivity of anyone who works with

12:04

data and of course

12:05

automate workflows, support data governance, create better capabilities around

12:10

observability

12:11

and manage data around these critical applications.

12:15

And then at the very top, as I mentioned before, we're a partner led services

12:21

and delivery

12:21

company which means that the vast majority of our customers are going to be

12:24

working with

12:27

expert partners to implement our AI services.

12:29

And so I think this unique approach really allows us to scale this AI

12:33

opportunity to

12:34

thousands of experts and services and delivery providers around the world.

12:38

So we'll continue to provide customers with that expert capacity.

12:42

They need to execute on their AI driven transformation.

12:46

One of the things we did introduce at Next was we launched a GNAI services

12:49

specialization

12:54

for partners who demonstrate the highest level of technical proficiency with

12:57

Google Cloud

12:58

GNAI.

12:59

So this specialization is going to unlock access to our products, has

13:04

additional funding for

13:06

GNAI assessment work, can increase access to AI resources and partner marketing

13:11

funds

13:11

and more.

13:12

The initial group of partners who have achieved this level of specialization

13:16

include companies

13:17

like Accenture, Capgemini, Cognizant, Deloitte, Quantify, Sears, TCS, Whitpro

13:24

and others as

13:26

well.

13:28

So just to give you a sense of the momentum we're seeing with the ecosystem,

13:32

here are

13:33

some interesting metrics.

13:34

So more than 60% of funded GNAI startups and nearly 90% of GNAI unicorns are

13:40

Google Cloud

13:41

customers.

13:43

And if you look at our developers, today we've helped more than a million

13:47

developers get

13:47

started with GNAI and our GNAI trainings have been taking millions of times.

13:53

And it's amazing when you look back, this past year as I mentioned before, how

13:57

quickly

13:57

our customers have moved from experimentation to actually implementing AI tools

14:02

and launching

14:03

these early stage products.

14:06

And then the last metric, if you look at our services partners, they've taken

14:09

more than

14:10

a half million GNAI courses to support customers.

14:14

And if you look at our global system integrators and global consulting partners

14:18

alone, they've

14:19

committed to train more than 200,000 experts on our GNAI solution.

14:23

So companies such as Accenture, Capgemini, Cognizant, Deloitte, HCL, Tech, KPMG

14:31

, Kindrol,

14:32

McKinsey, PwC and others.

14:36

So let me talk a little bit about marketplace.

14:40

So Google Cloud Marketplace and the business momentum we're seeing, both in

14:44

terms of the

14:45

gross transaction value, our co-so activities and the partner velocity.

14:49

So a few notable metrics on the left.

14:52

Our third party gross transaction value, more than doubled last year.

14:57

If you look at over 2022 to 2023, the number of partners actively transacting

15:03

on Google

15:04

Cloud Marketplace more than doubled in the past two years.

15:07

And then last year, we added hundreds of new partners and new solution listings

15:11

And I think the significantly increases our selection and scale.

15:15

In fact, if you look at the third party listings on Google Cloud Marketplace,

15:18

that nearly doubled

15:20

in the past two years.

15:21

So I believe collectively, if you look at these metrics on the left-hand side,

15:25

this

15:26

likely makes Google Cloud the fastest growing cloud marketplace among the

15:30

hyperscalers.

15:31

Now on the right-hand side, of course, we've talked about how partners gain a

15:35

very efficient

15:36

route to market.

15:37

Ultimately, they want to sell where the buyers are buying, which is

15:40

increasingly marketplace.

15:42

So we conducted a survey with our top ISV partners last year.

15:47

And they're seeing significant benefits in terms of selling through marketplace

15:50

versus

15:50

traditional sales channels.

15:52

So just share a couple of the findings here.

15:54

So first, 42% acceleration in deal cycle time.

15:57

And I think that's primarily driven by standardized agreements, simplified

16:01

negotiations,

16:01

and enabled customers to procure without engaging in some of the lengthy

16:04

procurement

16:05

vendor review cycle times.

16:07

They're seeing a 35% increase in win rate.

16:09

And I think this is primarily driven by our strong co-selling engagement we

16:13

have with our

16:14

partners, but also they have access to the customer committed cloud spend.

16:18

And many customers see that these packaged third party solutions are the

16:21

fastest way

16:22

to consume GCP.

16:24

And then lastly, a 32% increase in deal cycle time.

16:28

So private offers, of course, is driving a big part of that growth, moving that

16:32

direct

16:32

sales led motion online.

16:34

And we're consistently seeing deals, total contract value of millions and tens

16:38

of millions

16:39

of dollars.

16:40

And in fact, if you look, we've actually had multiple nine figure deals, again,

16:44

total

16:44

contract value transacted in marketplace.

16:46

So you can see the deals are just continue to grow.

16:49

So I think we all know that marketplace is becoming mainstream.

16:53

So at minimum, it needs to be a strong component of any SAS go to market

16:57

playbook, but it's

16:58

quickly becoming that channel for selling software.

17:01

And I see one of our top tier partners, they're driving 50, 70% of their GCP

17:06

business to marketplace.

17:08

So now let's dive into a couple of areas of cloud go to market where I think G

17:14

NAI will

17:15

help, we'll make this evolve quite a bit.

17:18

So I think last year, we launched a new GNAI category on Google Cloud

17:23

marketplace.

17:24

So this gives our customers a very seamless experience in terms of being able

17:29

to find,

17:30

buy, use all the different GNAI offerings from companies like Meta and Vidya,

17:35

Cohere,

17:36

AI 21 Labs, Typeface, Mishkel, Able, GLEEN, Gerato and many others.

17:42

And we have a very robust solution validation process that enables us to

17:46

surface the security

17:47

and such a solution.

17:48

So we knew this was just the beginning, because when you think about the

17:52

opportunity, when

17:53

you combine ML models, data assets, AI frameworks, it's unlocking a tremendous

18:00

amount of innovation

18:02

and monetization opportunities, because many customers want to use and/or ret

18:06

rain models

18:07

with their own data, their own corpus of data, but also data they may actually

18:10

procure

18:10

or subscribe from third party providers.

18:14

And then many app developers want to use data sets and they want to build

18:18

enhanced data

18:18

services powered by these AI models.

18:20

So the opportunity I see is that Google working very closely with our partners

18:26

have this leverage

18:28

the combination of commerce provided by marketplace, big query powered data

18:35

analytics and then

18:36

vertex support for prediction inference.

18:39

And collectively, we have the ability to accelerate the monetization of all

18:44

these AI and data

18:45

assets and services for partners and really establish Google Cloud as that

18:50

preferred provider

18:51

of AI and data solutions.

18:53

And at the heart of this, we will establish marketplace as the primary platform

18:58

for onboarding,

18:59

publishing, contracting all of our ecosystem AI and data offering.

19:04

So data providers are integrating into BigQuery and data provisioning and

19:09

access is being

19:10

handled with analytics hub.

19:12

We have a pretty thriving data ecosystem, 3,500 plus listings.

19:18

We have 350 petabytes of data shared per week.

19:21

It's a combination of public data sets, free data sets from Google, commercial

19:26

data sets

19:27

from companies like Dunne and Bradstreet, Axiom, Zoom Info, CoreLogic, and then

19:32

we have a number

19:33

of SaaS applications in areas like retail, marketing, manufacturing, supply

19:38

chain, sustainability

19:39

and others.

19:41

And of course, with Model Garden, that will continue to be the platform

19:45

experience for

19:46

app developers and data science to discover and use and manage ML models, which

19:51

is in

19:51

context, but in terms of commercialization and transactional capabilities,

19:55

marketplace

19:56

will be the place.

19:57

And so, marketplace is really going to provide this consolidated catalog of all

20:02

AI and data

20:03

ecosystem listings, which will include first party, third party, open source

20:07

models.

20:08

And it really becomes that single location for partners to onboard on to Google

20:12

Cloud

20:12

and allow customers to discover, browse and buy all the Google Cloud ecosystem

20:17

AI solution.

20:18

And for both third party, commercial models, as well as open source models, we

20:22

're going

20:22

to enable the full procurement lifecycle on marketplace.

20:27

So it's partner onboarding, it's search, it's discover, it's procurer, it's

20:31

deploy, it's

20:32

govern, it's manage.

20:33

And we're really going to provide a range of deployment options.

20:36

So partner can choose SaaS where they manage the infrastructure themselves, or

20:41

they can

20:41

choose Vertex AI where Google manages both the model and the infrastructure, or

20:46

for those

20:46

who want really the maximum control and customization, they can also leverage G

20:50

KE, Kubernetes engine,

20:52

where the partners can manage the infrastructure to their preferences.

20:55

And of course, with Vertex AI, partners, the model weight protection, so that

21:01

ensures

21:02

the security of the IP.

21:04

And then customers are more confident because they can use the models knowing

21:07

that their

21:07

query and grounding data does not go back into the partner models.

21:11

So as I mentioned before, anthropic Cloud 3, haiku, sonnet, opus, they're all

21:15

listed

21:16

on marketplace.

21:17

We have commercial data sets available from Dunne Bradstreet, Weather Source,

21:22

IP info,

21:23

true elements and more, and you'll see a lot more of these listing and

21:26

solutions coming

21:27

online.

21:28

So very exciting area.

21:29

All right, let me shift to one other area is around partner networks.

21:36

And when we think about it, taking a step back, I think when you think about

21:40

our buyers,

21:40

they're very digitally savvy.

21:43

That's why a lot of these transactions are moving online to marketplace.

21:46

They're consumer-like, they have different buying behaviors, and they're

21:50

supported by

21:51

many different partners as part of the customer journeys, right?

21:54

Which is their partners that they trust and it helps reduce some of the

21:57

friction in the

21:58

buying process.

21:59

So the other really exciting area is really enabling all partner types to

22:05

participate in

22:07

marketplace value creation and delivery.

22:09

So even those that are more focused on selling services, because I think the

22:13

opportunity here

22:14

is the role of marketplace is going to really be the orchestrator aligning the

22:19

customer's

22:20

preference for more integrated solutions.

22:22

And you can simplify some of these complex transactions by bringing a more

22:27

diverse array

22:28

of products and services together under a single umbrella.

22:31

Now the concept of partner networks to drive customer business outcomes is not

22:36

necessarily

22:37

new.

22:38

But I think what makes this exciting is that cloud marketplace is really that

22:42

connective

22:42

tissue for the broader ecosystem.

22:45

Because historically, if you think about partner networks, they've kind of c

22:48

atered to very specific

22:49

partner business models like a channel partner or an ISV.

22:53

They have different partner management.

22:54

They have different differentiation.

22:56

They have different Co-Cell motions.

22:57

They have different incentives.

22:58

And I think cloud marketplace can really become that central mechanism to

23:02

really connect

23:03

both the channel partners and the ISV and the services partners to facilitate

23:08

this Co-Cell

23:09

at scale.

23:10

So think about marketplaces.

23:12

It's not only just bringing sellers and buyers, but it's also bringing the

23:15

different partner

23:16

types together to collaborate.

23:18

And so we're enabling a lot of these different partners and partner types.

23:22

So it's certainly data providers and resellers and AI foundational model

23:27

providers, distributors,

23:28

specialized SIs, multinational service providers, MSPs and much, much more.

23:34

And I think it's going to play this critical role marketplace where we shift

23:37

from selling

23:38

products to selling customer outcomes and solutions.

23:43

And we're going to enable deals with multiple ISVs, ISV plus a service provider

23:48

, create these

23:49

bundle service offerings through marketplace.

23:51

So one of the areas I want to just highlight is where this is a good example is

23:58

industry

23:59

value network.

24:00

So industry value networks really combine expertise and offerings from the

24:04

system integrators,

24:05

the ISV, the content providers, a domain or industry specific AI models to

24:10

create a very

24:10

comprehensive, differentiated, repeatable and high value solution.

24:16

And it minimizes the need for clients to build bespoke solutions and creates

24:20

and addresses

24:21

some of these common challenges.

24:22

So just to give a couple of examples, quantifying on Cork, for example, they

24:26

have an AI led underwriting

24:27

platform.

24:28

You know, quote divine offers a very seamless underwriting experience with

24:32

reduced effort

24:33

and high extraction accuracy.

24:36

We have a publ assist and safe and working with live ramp, litics, bloom reach,

24:40

growth

24:40

loop and centralized retail media network planning to allow brands to activate

24:45

audience in a

24:45

very private, you know, privacy safe manner.

24:48

So it's a fast moving space and you can see the potential here in terms of

24:52

collaboration

24:53

and monetization.

24:54

So I know I went through this content fairly quickly, but you know, hopefully

24:58

this overview

24:59

was helpful.

25:01

I don't think we have any time for Q&A, but I have my Google colleagues.

25:03

Hopefully they answered most of these questions along the way, but you can see

25:07

there's a lot

25:07

of momentum, a lot of platform investments.

25:10

And as I mentioned, 2024 is shaping to be the year where companies are moving

25:14

from experimentation

25:15

and proof of concept to production at scale.

25:18

So my team is here to support you.

25:20

And really incredibly exciting time.

25:23

I look forward to building this joint business together.

25:25

Thank you.

25:26

[BLANK_AUDIO]

25:36

[BLANK_AUDIO]

25:46

[BLANK_AUDIO]

25:56

[BLANK_AUDIO]

26:06

[BLANK_AUDIO]

26:16

[BLANK_AUDIO]

26:26

[BLANK_AUDIO]

26:36

[BLANK_AUDIO]

26:46

[BLANK_AUDIO]

26:56

You