WEBVTT - Salesforce CEO Marc Benioff Talks AI Strategy

0:00:02.400 --> 0:00:08.119
<v Speaker 1>Bloomberg Audio Studios, Podcasts, Radio News. Market's always great to

0:00:08.160 --> 0:00:10.799
<v Speaker 1>be here with you at Dreamforest. At the start of

0:00:10.840 --> 0:00:13.440
<v Speaker 1>the show, the rock star of the show, thank you

0:00:14.280 --> 0:00:16.360
<v Speaker 1>agent for us, of course, and you rolling that out

0:00:16.440 --> 0:00:17.800
<v Speaker 1>is a big thing. I've been talking to some of

0:00:17.840 --> 0:00:21.640
<v Speaker 1>your customers. They're very excited. Tell us about what lllms

0:00:21.680 --> 0:00:24.599
<v Speaker 1>you built this on. Is it open AI as it is.

0:00:24.760 --> 0:00:27.800
<v Speaker 2>I've never seen the customers more excited. And the reason

0:00:27.840 --> 0:00:31.200
<v Speaker 2>why is these customers have been convinced something that's not true,

0:00:31.680 --> 0:00:33.519
<v Speaker 2>that they have to buy an LLM, they have to

0:00:33.560 --> 0:00:36.000
<v Speaker 2>go to a hyperscaler, they have to get a database,

0:00:36.440 --> 0:00:39.360
<v Speaker 2>they have to buy a security technology, they have to

0:00:39.440 --> 0:00:42.559
<v Speaker 2>choose a chip, they have to hook it all together themselves,

0:00:42.600 --> 0:00:47.120
<v Speaker 2>they have to deiy their AI. It's not true. You

0:00:47.200 --> 0:00:50.479
<v Speaker 2>do not have to diy r AI. And when we

0:00:50.520 --> 0:00:53.680
<v Speaker 2>started the show, they didn't really believe it. So that's

0:00:53.720 --> 0:00:56.680
<v Speaker 2>why we created a whole factory across the street as

0:00:56.680 --> 0:00:59.760
<v Speaker 2>you saw in Moscowne West, where we have put more

0:00:59.760 --> 0:01:03.320
<v Speaker 2>than ten thousand customers through already to show them that

0:01:03.720 --> 0:01:06.880
<v Speaker 2>on the Salesforce platform, which they've already been committed to

0:01:07.600 --> 0:01:11.480
<v Speaker 2>for more than twenty years, had great success in sales

0:01:11.520 --> 0:01:15.760
<v Speaker 2>and service and marketing and communications and commerce and analytics

0:01:15.800 --> 0:01:20.560
<v Speaker 2>and Slack and all of these technologies that now they're

0:01:20.600 --> 0:01:25.000
<v Speaker 2>also building AI. And this is a shock to them,

0:01:25.319 --> 0:01:28.080
<v Speaker 2>and it's exciting for us because we're showing them how

0:01:28.280 --> 0:01:32.200
<v Speaker 2>once again we're future proofing their company with the Salesforce platform.

0:01:32.240 --> 0:01:37.360
<v Speaker 1>So you've basically di wied it for them, software company.

0:01:37.440 --> 0:01:38.600
<v Speaker 1>What models you've built it on?

0:01:38.720 --> 0:01:41.080
<v Speaker 2>Is it? We have our own? We have many and

0:01:41.120 --> 0:01:44.160
<v Speaker 2>they don't need to know. And they don't Why what

0:01:44.360 --> 0:01:48.120
<v Speaker 2>chip am I using? You know, what hyperscal are? They

0:01:48.120 --> 0:01:51.400
<v Speaker 2>don't care. They don't care what I'm using. You know,

0:01:51.640 --> 0:01:55.360
<v Speaker 2>it's like what type of flowers in the cake? What

0:01:55.480 --> 0:01:57.960
<v Speaker 2>kind of sugar did I buy at the store? Why

0:01:58.000 --> 0:02:00.040
<v Speaker 2>does it matter? Do you like the cake? Is this

0:02:00.160 --> 0:02:01.640
<v Speaker 2>is the final product that you like?

0:02:01.880 --> 0:02:02.240
<v Speaker 1>All right?

0:02:02.240 --> 0:02:04.240
<v Speaker 2>And I think that for our customers, they don't want

0:02:04.280 --> 0:02:07.040
<v Speaker 2>to know all the pieces. They want to know does

0:02:07.080 --> 0:02:11.680
<v Speaker 2>this work? And can they make a greater success for

0:02:11.760 --> 0:02:13.560
<v Speaker 2>their own companies? And so when you see like a

0:02:13.600 --> 0:02:17.440
<v Speaker 2>company like you know, Wiley, who's here and they're like

0:02:17.880 --> 0:02:22.280
<v Speaker 2>lower their cost by fifty percent day one, right, And

0:02:22.360 --> 0:02:25.480
<v Speaker 2>that is the power of technology that we're going to

0:02:25.480 --> 0:02:28.440
<v Speaker 2>help them make it easier for them, lower their cost,

0:02:28.560 --> 0:02:32.040
<v Speaker 2>deliver a high ROI and the Salesforce Platform just delivered

0:02:32.080 --> 0:02:32.480
<v Speaker 2>that FROMIL.

0:02:32.600 --> 0:02:35.000
<v Speaker 1>So let's talk about how it works. You know a

0:02:35.000 --> 0:02:38.520
<v Speaker 1>lot of companies are rolling out agents now service now, Microsoft,

0:02:39.200 --> 0:02:42.320
<v Speaker 1>how do we know this isn't just another buzzword like copilot?

0:02:43.720 --> 0:02:45.520
<v Speaker 2>Well, I think you know that, I think and we

0:02:45.600 --> 0:02:48.640
<v Speaker 2>all know now that Microsoft Copilot is basically the new

0:02:48.680 --> 0:02:52.400
<v Speaker 2>Microsoft Clippy that customers have not gotten value from it.

0:02:52.840 --> 0:02:57.400
<v Speaker 2>You know, these llms, this is like we're selling science

0:02:57.440 --> 0:03:00.280
<v Speaker 2>projects to companies and they're tired of it. They have

0:03:00.400 --> 0:03:03.200
<v Speaker 2>not gotten the value, which is why you see these

0:03:03.200 --> 0:03:06.680
<v Speaker 2>customers so excited. They're coming here and they're getting immediate

0:03:06.800 --> 0:03:12.240
<v Speaker 2>value and they're able to use this next generation platform

0:03:12.680 --> 0:03:16.880
<v Speaker 2>to do everything that they've heard that's possible. Create an

0:03:16.880 --> 0:03:18.960
<v Speaker 2>agent is a great example. How do I find tune

0:03:18.960 --> 0:03:21.600
<v Speaker 2>an agent? How do I create guardrails for an agent?

0:03:21.960 --> 0:03:24.160
<v Speaker 2>How do I ground an agent? And data? Well, all

0:03:24.160 --> 0:03:27.600
<v Speaker 2>the data is in the Salesforce platform. All of the

0:03:27.639 --> 0:03:31.160
<v Speaker 2>metadata is there, the logic is there, the workflow is there.

0:03:31.200 --> 0:03:35.840
<v Speaker 2>They're just adding agents into their existing workflow and no

0:03:35.880 --> 0:03:38.200
<v Speaker 2>one knew it was possible before the show, so we

0:03:38.280 --> 0:03:42.160
<v Speaker 2>knew that because we hadn't created the product. And now

0:03:43.120 --> 0:03:46.200
<v Speaker 2>it's so exciting because we've onboarded tens of thousands of

0:03:46.240 --> 0:03:50.440
<v Speaker 2>customers and we've convinced them you don't need a DIY

0:03:50.520 --> 0:03:53.640
<v Speaker 2>your AI. You can just use and you're going to

0:03:53.680 --> 0:03:58.520
<v Speaker 2>get the best AI, the highest accuracy, the lowest hallucination rate.

0:03:59.320 --> 0:04:02.520
<v Speaker 2>Our customer are saying in many cases it's two x

0:04:03.000 --> 0:04:05.560
<v Speaker 2>the accuracy of the best models.

0:04:06.040 --> 0:04:09.440
<v Speaker 1>So look, you know it's not all ways going to

0:04:09.440 --> 0:04:12.800
<v Speaker 1>be smooth sailing. Disney actually just announced they've dropped Slack

0:04:12.920 --> 0:04:16.039
<v Speaker 1>because they got hacked and said they're switching to an

0:04:16.160 --> 0:04:20.479
<v Speaker 1>enterprise wide collaboration tool, which sounds maybe like Microsoft, what

0:04:20.520 --> 0:04:23.400
<v Speaker 1>are you doing to improve Slack security and what's your

0:04:24.200 --> 0:04:24.920
<v Speaker 1>response to that?

0:04:25.640 --> 0:04:28.560
<v Speaker 2>Well, listen, I think our security is rock solid. This

0:04:28.720 --> 0:04:32.039
<v Speaker 2>is really important, and also there's no finish line when

0:04:32.040 --> 0:04:34.800
<v Speaker 2>it comes to security. But companies have to also take

0:04:34.839 --> 0:04:39.120
<v Speaker 2>the right measures to prevent phishing attacks and to lock

0:04:39.160 --> 0:04:42.520
<v Speaker 2>down their employees from social engineering. So we can do

0:04:42.640 --> 0:04:45.160
<v Speaker 2>our part, but our customers also have to do their part.

0:04:45.200 --> 0:04:48.479
<v Speaker 2>That's extremely important. But the second thing that's important is

0:04:48.960 --> 0:04:51.640
<v Speaker 2>over twenty five years, we've shown that this model is

0:04:51.680 --> 0:04:55.400
<v Speaker 2>the most trusted, most secure, most capable model. Disney's a

0:04:55.400 --> 0:04:58.400
<v Speaker 2>great example. They run the Disney Store on Salesforce, the

0:04:58.480 --> 0:05:01.640
<v Speaker 2>Disney guides are on Salesforce, all of the sales and

0:05:01.680 --> 0:05:06.920
<v Speaker 2>service operations Disney plus call center and contact centers on Salesforce.

0:05:07.320 --> 0:05:11.160
<v Speaker 2>We are the customer service capability and they're the proving

0:05:11.200 --> 0:05:14.520
<v Speaker 2>ground actually for our new AI system. So our next

0:05:14.560 --> 0:05:19.279
<v Speaker 2>generation AI was actually proven at Disney. It's amazing. They've

0:05:19.320 --> 0:05:23.200
<v Speaker 2>shown us incredible metrics and I'm really looking forward to,

0:05:23.240 --> 0:05:27.160
<v Speaker 2>you know, future success with Disney. But it's incredible that

0:05:27.240 --> 0:05:30.080
<v Speaker 2>customers are able to see this value from these platforms.

0:05:30.200 --> 0:05:31.839
<v Speaker 2>This is that's what's really exciting.

0:05:31.880 --> 0:05:34.279
<v Speaker 1>And let's talk about the money, because I'm curious about

0:05:34.320 --> 0:05:36.560
<v Speaker 1>how you got to this two dollars per conversation and

0:05:36.640 --> 0:05:40.080
<v Speaker 1>just how much additional revenue you think agent Force is

0:05:40.160 --> 0:05:40.880
<v Speaker 1>going to drive.

0:05:41.440 --> 0:05:43.520
<v Speaker 2>Well, Look, our goal is very simple. We want to

0:05:43.560 --> 0:05:46.800
<v Speaker 2>get what we call a billion agents you know, in

0:05:46.880 --> 0:05:49.960
<v Speaker 2>place for our customers within the next twelve months. And

0:05:50.000 --> 0:05:51.960
<v Speaker 2>that means we're going to have to deploy it in

0:05:53.120 --> 0:05:56.000
<v Speaker 2>one hundreds of thousands of customers, okay, and get them

0:05:56.000 --> 0:05:59.159
<v Speaker 2>all to roll this out to build this huge blanket

0:05:59.200 --> 0:06:02.240
<v Speaker 2>with all of their conc and of course the funny

0:06:02.240 --> 0:06:04.159
<v Speaker 2>thing is, we say a billion agents, but you know

0:06:04.200 --> 0:06:07.480
<v Speaker 2>it's only one piece of code. The Salesforce platform is

0:06:07.920 --> 0:06:12.039
<v Speaker 2>automating all your customer touch points, building a huge data

0:06:12.040 --> 0:06:15.000
<v Speaker 2>cloud and all your agents, but it's all one piece

0:06:15.000 --> 0:06:18.680
<v Speaker 2>of code that you're just connected to. And that idea

0:06:19.279 --> 0:06:21.960
<v Speaker 2>that it's one piece of code, well then we're of course,

0:06:22.000 --> 0:06:24.760
<v Speaker 2>you know, we've always charged you know, per email or

0:06:25.080 --> 0:06:29.760
<v Speaker 2>per commerce transaction, and now per conversation. It's consumption pricing.

0:06:30.240 --> 0:06:32.760
<v Speaker 2>In some cases it's per user pricing, like in our

0:06:32.800 --> 0:06:35.480
<v Speaker 2>sales cloud or our service cloud, so it's a mix

0:06:35.920 --> 0:06:39.159
<v Speaker 2>of two different pricing models. But we're of course work

0:06:39.200 --> 0:06:41.720
<v Speaker 2>with our customers, and you know, we have tens of

0:06:41.720 --> 0:06:45.279
<v Speaker 2>thousands of customer facing professionals who work with them to

0:06:45.320 --> 0:06:48.400
<v Speaker 2>make the right agreement with our customers every single time.

0:06:48.520 --> 0:06:50.040
<v Speaker 1>All right, Well, it's great to be with you here

0:06:50.040 --> 0:06:51.080
<v Speaker 1>in person, and.

0:06:51.040 --> 0:06:53.960
<v Speaker 2>To always fun to be with you, Agent Thrill, you

0:06:54.000 --> 0:06:54.640
<v Speaker 2>could come back.

0:06:55.520 --> 0:06:59.200
<v Speaker 1>Salesforce Chair CEO and founder Mark Benioff, good to have

0:06:59.320 --> 0:06:59.719
<v Speaker 1>you