Podcast: Salesforce CEO Marc Benioff on AI, Microsoft, Seattle tech, and the future of humanity

Our visitor this week on the GeekWire Podcast is Salesforce CEO Marc Benioff, who says he has by no means been as enthusiastic about something in his profession as he’s concerning the newest developments in synthetic intelligence — AI brokers that may autonomously motive, plan, and take motion on behalf of companies.

“We’re going to do that in a large-scale manner and provides prospects a degree of success with AI that they’ve actually by no means had earlier than,” he mentioned. “I’m tremendous enthusiastic about it. I believe it’s going to be unimaginable.”

Benioff is sort of as robust in his destructive sentiments towards Microsoft’s Copilot. He calls Copilot the second coming of Microsoft’s much-maligned “Clippy” Workplace assistant, and asserts that the Redmond firm is giving AI a foul title by disappointing prospects with underwhelming outcomes and lax safety.

There was one factor that diverted him from these subjects throughout our dialog: his canine, Honey, sneaking up behind him in his dwelling workplace to eat his bagel, tipping over a tray and breaking a plate within the course of.

Nevertheless it didn’t take lengthy for him to get again on subject: “We’ll clear that mess up,” he mentioned. “That’s, by the way in which, very metaphorical for the mess that Microsoft has made regarding the AI business.”

I spoke with Benioff, the CEO and co-founder of cloud computing pioneer and enterprise purposes big, upfront of the overall availability of the San Francisco firm’s Agentforce AI know-how for gross sales and repair.

We additionally talked about Benioff’s possession of Time journal, and what meaning for his political involvement; and the standing of his previous pledge to show Seattle into Salesforce’s HQ2.

Pay attention under, and proceed studying for a transcript, edited for readability. Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you hear.

Todd Bishop: You appear to be in all places these days. You’re speaking concerning the potential for AI brokers to rework enterprise. You’ve been slapping Microsoft upside the pinnacle, which we are able to discuss. You even talked concerning the potential to rename the corporate Agentforce. It’s just like the Marc Benioff of outdated, in case you don’t thoughts me saying. So let me simply begin by asking you, what’s obtained you so fired up about AI brokers, this entire notion of autonomous synthetic intelligence?

Marc Benioff: Properly, you’re proper. I’m fired up. I might say I’ve by no means been extra enthusiastic about something in my total profession. For the final 45 years that I’ve been writing software program and publishing software program and dealing with prospects to achieve success with software program, effectively, that is our dream, what’s taking place now, that we are able to actually see that software program can do issues that it’s by no means been capable of do earlier than, that it could actually make us extra productive, actually extra productive. It will possibly increase our workers. It will possibly enhance our margins, develop our revenues. It will possibly enhance our KPIs.

Basically bettering our total enterprise efficiency, that is our dream of why all of us obtained into the software program, enterprise as a result of we thought that is the place software program’s going and we’re at this second proper now. And that’s why I’m actually completely excited.

TB: Once you make the pitch to prospects, I can think about that a few of them are nonetheless saying, “Wait a second, Marc. I’m nonetheless getting used to the notion of simply working with AI as a companion. This may look like a step too far for 2024, to have autonomous AI brokers engaged on behalf of my enterprise.” What do you say to people once they convey that up?

Benioff: Properly, I believe you’re taking a reasonably large soar into the abyss there, and earlier than we try this, I believe you must take a step again. Look, we’ve been working with AI for a very long time. For Salesforce, we’ve been doing AI for fairly some time and critically AI for greater than a decade, particularly with our Einstein platform. We’re the most important supplier of enterprise AI on the planet. We do about 2 trillion enterprise AI transactions this week alone, and that’s throughout all of our merchandise and platforms.

And it’s been a implausible journey — machine intelligence, machine studying, deep studying. I imply, it’s been superior. And this generative AI second is thrilling as a result of it’s one other step on our collective AI journey.

So whereas we’ve been engaged on AI, I believe we’ve all the time been saying, “Oh, it’s going to make us extra profitable. It’s going to make us extra productive.” However now with this introduction of this idea of those brokers — it’s software program that’s reasoning, planning, taking motion in your behalf — and utilizing the generative capabilities that we see in these LLMs that we love. After which grounding that, connecting with our company databases, It offers us the power to essentially obtain capabilities with enterprise software program we simply haven’t been capable of do earlier than. And we’re within the zone.

[With Agentforce], the early outcomes have been simply nothing wanting, effectively, it’s simply gorgeous, what is feasible. We’re going to return out of the gate actually swinging on this, and that’s what you’re feeling.

AI brokers, a sensible instance

TB: I obtained an opportunity to speak with a few your executives, Adam Evans and Brent Hayward, so I’ve a way for the reply to this query, however are you able to give me some examples of what folks will be capable of do with Agentforce that they won’t have been capable of do with only a chatbot or a generative AI resolution?

Benioff: That is about, for the primary time, people with brokers driving buyer success collectively.

Let’s take a particular business, healthcare. All of us have had this expertise. I actually have a boot proper now I’m sporting as a result of I had a foul SCUBA diving soar a pair weeks in the past, and I’m going to be going again into an MRI on Friday. …

Healthcare could be very a lot a person sport in our nation the place you’re by yourself, and sooner or later that simply shouldn’t be going to be true. As a result of what’s going to occur is, your cellphone goes to ring, otherwise you’re going to get a textual content and it’s going to be, “Hey, did you drink the water? We’re following up to your MRI and simply wish to know what’s happening with you and the way are you feeling, and are you experiencing swelling? What’s happening precisely?”

And also you’re going to say, “Sure, I did. Oh, thanks for reminding me. Oh yeah, no, it’s advantageous.” “Do you want extra assistance on understanding all of the issues happening along with your foot?” “Oh, give me extra info. Oh, that’s nice. What are the 5 issues I have to know?” “These are the 5 issues you must know. And are you positive you’re OK? Do you wish to discuss to a health care provider?” “No.” “Do it is advisable reschedule your appointment?” “No.” “Properly, we’re going to schedule a repeat MRI for you in a month, so I’ll be contacting you once more.”

That final half, it’s all with an agent, and this concept that we are able to fill within the gaps of our human workforce with a digital workforce. This can be a large thought and it is a good thing on this business, healthcare. As a result of I’d want a repeat MRI, I’d want repeat labs, I’d have to schedule an appointment with my physician.

However, look, my physician doesn’t wish to hear from me. My physician’s burnt out. My nurse is burnt out. They’re in a post-pandemic malaise. You understand that’s true. You go discuss to any healthcare group. However these digital employees, they might fill on this hole they usually can provide us some respiration house between our medical doctors, who ought to be engaged on sure essential issues.

As a result of there’s loads the AI can not do, particularly in healthcare. However there’s now loads that the AI can do, particularly in healthcare. And our job must be to place these two issues collectively. And if we are able to try this, oh my gosh, it is a second for our prospects that’s going to be completely wonderful.

Agentforce vs. Microsoft Copilot

TB: You’ve been outspoken about Microsoft in latest weeks, and I actually wish to dig into this just a little bit extra with you than you’ve talked about earlier than. Basically, your subject is that Microsoft has set expectations too excessive and never met them, along with different issues. However I don’t wish to communicate for you. What’s your drawback with Clippy?

Benioff: We’ve all had that ChatGPT second the place we kind within the question and we get the outcome and it’s super-search and we’re like, “Oh my gosh, that is wonderful. That is magic. That is superior. How is that this potential?” And our first expertise with an LLM, we’re all remembering that. That’s so essential.

However the thought for them to say, “Now right here’s Copilot, and it’s now going to make your enterprise so significantly better, and right here you go,” that’s a fantasy that has not performed out. And also you simply discuss to prospects they usually’ll inform you, Copilot has not reworked their enterprises.

Each CEO was so excited with the imaginative and prescient of, “Now that we have now LLMs, AI is now curing most cancers and local weather change has stopped and your organization’s going to be absolutely automated.” Not so quick. And after we have a look at Copilot and after we have a look at this idea that each firm has to have their very own mannequin or LLM or prepare and retrain their mannequin, once more, this isn’t how that is going to play out. And that should come to an finish.

That’s actually the massive message that we have now for our prospects, which is, we’re the most important supplier of enterprise AI, and we’ve made you much more productive with AI. And now watch this, with Agentforce. With Agentforce, you’re going to attain the desires and visions that you simply’ve all the time had for AI. And we’re going to indicate you, similar to in my examples with healthcare, the place you’re going to have the ability to fill within the gaps along with your prospects to have a a lot increased degree of buyer intimacy.

Now, I’ll provide you with one other nice instance of an actual buyer. One of many first prospects that we turned Agentforce on is Wiley. Wiley is the textbook producer, and proper now everybody’s going again to high school in order that they’re all making an attempt to purchase extra textbooks, and they’re often surging their workforce, hiring extra gross sales and repair folks.

However not now, not presently of the 12 months, as a result of they’re utilizing Agentforce. In order that they have been capable of scale their service group, their gross sales group, their advertising group, their buyer outreach, with digital employees, not simply human employees. And that’s so essential. And this concept that Agentforce offers you the power to ship a better degree of buyer success than ever earlier than, that’s actually wonderful. And that’s the reason I’m so excited.

Microsoft’s many Copilots

TB: Microsoft itself has created some branding confusion right here and so it’s essential to level out, if you’re speaking about Copilot, you’re speaking about Copilot for Microsoft 365, which is the factor that they cost an additional $30 a person a month for. Not GitHub Copilot, which frankly has been one of many early successes within the business when it comes to its capacity to be a coding companion.

Benioff: Let’s simply take this aside.

TB: Go for it.

Benioff: As a result of I believe what you’re stepping into is one thing essential, which is that software program builders are extra productive utilizing AI, and there’s totally different instruments for doing that, together with that.

However then to say, now “My software program developer is extra productive with AI of this,” and now to say, “My entire enterprise goes to be extra profitable with this know-how” shouldn’t be true, and but it’s what they’ve been promoting.

And you may go on the market and see with the purchasers, they don’t have success with this concept. This concept {that a} essentially ungrounded AI into your enterprise goes to ship you extra productiveness shouldn’t be true, and that’s the place Microsoft has gotten themselves into hassle right here. So let’s simply put that apart for a second.

TB: Properly, however Marc, I ought to say, they might push again on that. We each know that, and they might say 60% of the Fortune 500 are utilizing this factor, they usually have particular case research the place they might present successes as effectively. I’m not going to undergo them.

Benioff: Which buyer are you going to say, let’s discuss it. Take one.

TB: I don’t have Microsoft’s case research in entrance of me. I can simply inform you that there are some.

Benioff: I’ve the product. It doesn’t work. … I imply, I’ve it proper right here. I’m like, “It’s simply ChatGPT.”

TB: What then is the distinction with Agentforce?

Benioff: What you need to understand is that LLMs are only a part of intelligence. They’re simply an ingredient. Like if we’re making a cake and we’d like sugar or salt, it’s only a part. However an LLM by itself is simply an input-output search engine. It has an amalgamated knowledge set with an algorithm which might be mixed and you set one thing in, your immediate, and out comes a response. It’s not a knowledge storage mechanism. It doesn’t retailer your company knowledge. It doesn’t have your metadata. It doesn’t have your sharing mannequin, it doesn’t have your safety mannequin. Let me provide you with one other instance.

I had a financial institution CEO come and discuss to me … they usually have this concept which you could simply dump your company knowledge into an LLM and it in some way goes to make you a better, higher firm. It’s not true.

However what’s true is you possibly can construct an structure out by way of a platform, good old style enterprise software program platform strategy, and the mix of an LLM, your company knowledge, your company metadata, your sharing mannequin, your safety mannequin, and in addition different crucial components of enterprise software program that we all know are so essential, just like the apps themselves, as a result of there must be a seamless handoff between your human employee and your digital employee. All of these items mixed can obtain that imaginative and prescient, and that’s the magic and the promise of what’s potential at present.

Gartner on enterprise AI safety

TB: Associated to this, you’ve mentioned that Microsoft Copilot is spilling knowledge in all places, and I went again and found the Gartner study that you simply have been citing there. And so simply briefly right here, I wish to quote what it says. It says, “Microsoft 365 Copilot honors person permissions. It additionally respects knowledge safety controls reminiscent of sensitivity labels the place these have been appropriately utilized. Nonetheless, if controls haven’t been appropriately utilized or permissions are too open, Microsoft 365 Copilot can enhance the chance of oversharing by retrieving, summarizing, and producing content material that the person mustn’t have entry to.”

That’s a part of it. I gotta say, if I have been modifying a narrative {that a} reporter turned in, and that was the textual content, I might have a tough time placing the headline on there that it was “spilling knowledge in all places.” Are you able to get into extra element about what you imply by that?

Benioff: Properly, I can. I believe you truly simply mentioned it, which is that prospects — and you may learn that Gartner report your self — prospects principally have reported on to me that they’re seeing ends in queries that they shouldn’t be seeing, that they don’t have that authorization as a person.

Within the instance of banking, which we simply went into with the know-your-customer strategy, my banker and your banker have totally different sharing fashions. Your banker can not see my account stability and my banker can not see your account stability. However on this instance, Copilot is displaying up, giving individuals who shouldn’t be seeing that knowledge, that knowledge. That’s all that’s happening.

TB: Proper, and I believe the purpose of the Gartner report is that it’s taking place in some circumstances, and it’s truly stopping some firms from adopting Microsoft 365 Copilot, however it looks as if it’s extra concerning the permissions that these firms themselves have placed on the information and the protections that they’ve applied.

Benioff: My level is that’s by no means going to occur with a Salesforce platform since you’re defining these person roles up entrance, your sharing mannequin. You’re defining your safety mannequin. And after we’re speaking about regulated industries like healthcare, monetary companies, the federal government, and different items, there is no such thing as a margin for error on these items.

My healthcare knowledge, my banking knowledge, I don’t need that spilling all around the flooring of those enterprises, which is what’s taking place. And also you’re citing the report truly that I learn, and there’s additionally a very good Computerworld article on it, as effectively, however it’s not simply that. It’s this false premise that Microsoft has bought prospects that that is what AI was meant to be, and it’s not true.

What AI was meant to be is the concept we’re going to have a extra productive enterprise with augmented workers, with increased margins, with increased revenues, and that is how we’re going to attain it. And in my instance with healthcare, or my instance with banking now, effectively, that’s what Agentforce does.

Like within the instance with Wiley. They’re attaining all of these issues which you can’t obtain with Copilot. And this concept that we’ve been bought for the final two years that Copilot can do that, it’s not true.

So we for 25 years have helped our prospects construct their gross sales forces, their service forces, their advertising forces, their commerce forces, they’re analytics forces, they’re knowledge forces. Now, we’re serving to to construct their Agentforce.

And that concept that workforces are essentially overwhelmed, that they’ve low-value duties, stalled productiveness, fastened capability, burning out, like I discussed, in healthcare, effectively, prospects need extra. They wish to get right down to that fast, zero-hold-time thought, private and empathetic service, working with specialists, scheduling issues, taking actions instantly, have an prompt response. None of these issues occur with Copilot.

And that larger thought we talked about originally: what if workforces had no limits? What if workforces had no limits? What does that imply to your firm? That’s not what we’re getting with Copilot, my pal. And I believe you understand that’s true. That concept, which is the massive thought when it comes to AI, radically bettering our capabilities, what if workforce has no limits, that isn’t what’s happening, and that’s what we’re going to ship with Agentforce.

And that concept that our prospects can use a platform that’s trusted and safe, scalable and correct, simple to customise, that has the AI in-built, that has your apps in-built, that makes use of a metadata framework, that has an open ecosystem, all of those primary capabilities. That concept which you could construct radically new forms of purposes to automate your organization? That is AI.

That’s what AI was meant to be and that’s the distinction that I’m enjoying round this fantasy of not solely Copilot, but additionally this false premise that I hear from numerous prospects that they’ve been bought by Microsoft, that they should construct their very own fashions and retrain their very own fashions. And that’s why we’d like these rising ranges of hyperscaler capability. And people issues will not be true. This may be executed in a extra environment friendly software program infrastructure simply as we have now for many years because the confirmed pc science strategy.

Microsoft’s new AI brokers

TB: I don’t imply to place myself within the place of Microsoft proxy right here, however I do wish to level out …

Benioff: No, it’s good. It’s nice. You’re doing advantageous.

TB: … they did come out this week, seemingly in a preemptive transfer ..

Benioff: Did you utilize the phrase panic?

TB: I don’t know who’s panicking. I learn half the headlines they usually say you’re panicking. I learn the opposite half they usually say Microsoft’s panicking. … However Microsoft did announce their own AI agents, about 10 of them, for Dynamics 365. Did you see that as validation?

Benioff: No. You must perceive, primary, Copilot has failed, and this fantasy that each firm has to go construct and rebuild their very own mannequin has failed, and prospects are on the lookout for a brand new alternative for this degree of automation. That’s what is thrilling. And we are able to do that. We are able to provide firms an unimaginable new functionality, and we see that with our prospects throughout the board.

Let’s discuss Disney. Disney is one other nice instance the place they’ve a really advanced product line and for his or her workers, they name them solid members, to get this very clever response for a buyer is extremely troublesome.

Now, we automate each Disney buyer contact level from Disney+ to the service heart, to promoting the true property, to this concept that you simply’re within the park with a Disney information. And this is a chance for his or her workers to have a fair heightened buyer expertise.

And I’ll provide you with a really concrete instance. I’m strolling by way of the park, one thing I truly do 4 occasions a 12 months, and I’m with a Disney information, and we’re on our approach to Rise of the Resistance, my favourite journey, it’s at Galaxy’s Edge, which is Star Wars Land, and now the information is notified the journey is down, as a result of it’s a really advanced journey to keep up. What’s the information to do? Does the information, A, know circulation management throughout the park in order that it is aware of each journey how they’re doing and what’s busy and what’s not, and, B, my journey historical past, all the pieces I’ve executed and never executed? As a result of these two issues want to return collectively right away to make a advice to me, “Marc, Rise of the Resistance is down. We have to go to this new location.”

However the agent at that second can say, with now greater than 90% accuracy, “Sure, Marc must go to Toon City.” … There’s no manner the information can try this on their very own. It’s not potential. However the AI and the agent can, and now the agent is working facet by facet with the information. That’s brokers with people driving a higher degree of buyer success to higher.

‘Trailblazer’ and human connections

TB: You have been speaking earlier about these “aha” AI moments, and I can guarantee you that you simply’re talking with an actual human right here on the opposite facet, as you possibly can inform, and never an AI agent but. However I had a kind of “aha ” moments with Gogle Pocket book LM.

Benioff: That could be a nice product.

So I learn Trailblazer when it got here out in 2019, and I’ve a duplicate on Kindle, and I took the Kindle copy and was capable of fiddle with it and get it into Pocket book LM and add it with all types of background concerning the AI stuff and all the pieces. And I requested it for inquiries to ask you. And one of many questions blew me away.

“You write in your e book Trailblazer that you simply’ve by no means forgotten what you discovered out of your dad. “Nothing can take the place of human relationships, the bedrock of any enterprise.’ How do you stability your perception within the energy of AI with the significance of human connection and relationships in enterprise?”

Benioff: Properly, that is I believe going to be the seminal second, not only for me, however for all of us. As a result of there’s no query that we have now this unimaginable capacity to reinforce our workers for the primary time with digital employees. That concept is so thrilling. That concept that we’re going to work facet by facet with AI, we’re all going to get our head round that. We’ve seen it within the films. All of us noticed it in “Minority Report.” We noticed a “Terminator.” We noticed “Her.” We noticed “House Odyssey.” We noticed all of them. Properly, we’re going to dwell it and there’s no query it’s all unfolding in entrance of us.

And that’s going to be the existential query of our time, which is, you and I are going to work facet by facet with brokers. I’m positive we’re already doing that at some degree already. For instance, what number of occasions a day do you already go to Gemini or ChatGPT or Perplexity otherwise you.com or any of those techniques to get a solution? Most likely fairly a bit.

However now let’s simply return a second. Earlier than that, you have been going to Google and also you have been doing Google Search, and Google Search is also AI. And now you might have “tremendous search,” is what I’d name with ChatGPT. After which after we mix that know-how and we floor it into your organization knowledge, add your sharing mannequin, which I’ve now made that time exhaustively, integrat it along with your apps, it’s time for some magic to occur. And I believe that’s what’s actually cool.

We simply did this convention in San Francisco referred to as Dreamforce, and we had 45,000 folks come. It was superior. I don’t suppose you got here, however possibly subsequent 12 months. And whereas we have been there, we took 10,000 of our prospects, as a result of I introduced in 4,000 of my workers simply to give attention to one thing referred to as Launchpad, which was assist our prospects construct brokers in actual time.

I’ve a dream that I’m going to attempt to get a billion brokers going within the subsequent 12 months. So I needed to essentially use it as a kickstart. I did one at present in Tokyo, by the way in which. I did one final week within the UK and in Australia. I’m making an attempt to essentially get this going globally. And sat down with these prospects and mentioned, “We’re going to construct this. Let’s go.” And that concept that we are able to do it and add and lengthen their system, that’s actually essential.

Saks Fifth Avenue got here to the convention, and they’re an amazing Salesforce buyer already, we automate all their buyer contact factors, and in an hour they constructed a system to handle their returns and their customer support. And in case you go to their web site at present, it’s operating.

In order that’s a giant thought, that they’ll lengthen their human workforce right away, in a second. And I believe all of us understand our workforces are constrained emotionally, bodily, spiritually, mentally, and we simply don’t have sufficient folks to do the roles that must get executed.

Even in my very own firm, I really feel that manner. Even proper now as I’m rolling out Agentforce, I want I may very well be doing a lot extra quicker and extra scaled. Brokers are going to supply that functionality, and that’s what it’s all about.

Salesforce within the Seattle space

TB: You talked about when Salesforce acquired Tableau that Seattle was going to be HQ2. Do you continue to see it that manner?

Benioff: We’ve got an enormous workforce in Seattle. We’ve got simply large people up there. We’ve got so many teams all around the world now. I believe what I didn’t anticipate was, we had this factor, I don’t know in case you heard about it, referred to as the pandemic. And after we had the pandemic and all of us went dwelling and all of us discovered the right way to use Zoom and be snug with Zoom, that modified all the pieces for me.

I believe HQ2 most likely appears to be like just a little bit extra like our properties and never a lot like a particular metropolis. Possibly that’s HQ1 for lots of firms. I’ve mates of mine who solely have distant workforces. So I believe all the pieces modified at that second.

It’s not that Seattle isn’t actually essential to us. We’ve got an enormous facility there. We’ve got a model new tower in Chicago, in Sydney, in Tokyo, in New York, in Atlanta, in Paris, all around the world. However on the finish of the day, we’re extra versatile, extra digital employees than ever earlier than, and we’re going to proceed to be that manner.

Time journal and the media enterprise

TB: In 2020, as a part of your possession of Time journal, you declared, I believe rightly so from a journalistic standpoint and ethically that you’d now not be making political donations and (private) endorsements. As we strategy November fifth, any regrets about that call?

Benioff: No. I believe it’s been an incredible determination. To begin with, I’ve actually loved my stewardship of Time. Right here’s this subject on inspiring ladies, I’m holding it up, and that is one thing that we introduced truly into the model, which isn’t solely the idea of Man of the 12 months, however Lady of the 12 months.

I might say that this has been a second within the final six years the place we’ve tried to focus extra on the center manner, and that’s proudly owning Time and being within the media business. It’s, by the way in which, very totally different than the tech business. It’s a lot more durable. The media business could be very troublesome. It’s very difficult. The tech business, software program, I’ve been doing for 45 years, it’s a lot simpler.

TB: What’s the hardest factor about media?

Benioff: It’s the fluidity of the fashions, and I believe that with Time, what we’ve executed is, we have now the print, which I simply confirmed you. We even have the digital. We’ve got occasions. We’ve got different kinds of digital merchandise. Even our NFTs, in case you bear in mind these. We’ve got our Time Studios. We’ve got a film studio now. We’ve got many various kinds of merchandise inside Time. And that concept that we’re continuously constructing and rising and increasing manufacturers, sub-brands, new merchandise, it’s so fluid and so fast-paced. I might say that in some methods it’s much more so than the tech business.

Subscribe to GeekWire in Apple Podcasts, Spotify, or wherever you hear.

Audio modifying by Curt Milton.