From Slide Rules to AI Agents | 50 years in 60 minutes with Peter Coffee (Ep 010)

In this episode of the DATA SECRETS Podcast, I sit down with Peter Coffee—legendary technology researcher, prolific writer, and longtime Salesforce evangelist—to unpack five decades of digital transformation, the dawn of AI, and the future of intelligent business.

Peter shares hard-won lessons from the days of slide rules and room-sized mainframes through the PC revolution, cloud-first innovation, and today’s AI-powered landscape. We explore:

• Why using AI just to do old tasks “better, faster, cheaper” leaves a ton of value on the table
• The difference between incremental improvement and breakthrough, unimagined possibilities with data
• What the “AOL dial-up phase of AI” means for businesses and how AI soon becomes like the air we breathe
• How Salesforce’s strategy (via acquisitions, platforms, and partnerships) reflects the relentless march toward seamless intelligence
• Why the tools of tech should “disappear into their function” and what it takes for leaders to build truly adaptive organizations
• The new world of AI agents, what Minsky got right about “minds,” and how complex digital systems will soon self-orchestrate
• Practical pitfalls and mindset shifts for execs competing in a world where your real rival is the “best digital experience” a customer has... anywhere!

Peter also shares what’s next for him: his work with the Foundation for Intelligent Life on Earth and his urgent call for data-driven action on climate change.

From AI’s hidden potential to the cultural (and personal!) changes driving decades of innovation, this episode is packed with stories, strategies, and provocations for anyone building the future of business.

If you’re ready to move beyond the “easy wins” in AI and data, and want to unlock transformational opportunities, this conversation with Peter Coffee is a must-listen.

Connect with Peter:
🔗 Foundation for Intelligent Life on Earth: http://intelligentlifeonearth.com/
🔗 LinkedIn: https://www.linkedin.com/in/petercoffee/

Follow the DATA SECRETS Podcast

📬 Get episode recaps & bonus insights at allegroanalytics.com/podcast
🕺🏻 Connect with Nathan on LinkedIn
📺 Watch every episode on our YouTube Channel

Transcript

Peter Coffee [00:00:00]:
If you're using AI to do the things you've been doing, better, faster, cheaper, you're leaving an enormous amount of value on the table. The applications everyone uses every day and says, oh yeah, couldn't this be better? Those are easy to identify, the things that no one is doing at all. Because it's so hard to do with the old methods where you say, you know, you could have this. Now that's a much more valuable conversation, but it takes a significantly greater effort to get that conversation started.

Nathan Settembrini [00:00:33]:
Your data has secrets, secrets that could change everything if you only knew where to look. Welcome to the Data Secrets Podcast. Welcome to the Data Secrets podcast where we uncover how business leaders crack the code, find truth in their data, and turn insight into, into action. Today we have a very special guest, Peter Coffey. He's a technology researcher, writer, evangelist. He spent the last 19 years at Salesforce. I heard him give a keynote recently and just had to have him on the podcast. So, Peter, welcome.

Peter Coffee [00:01:13]:
It's good to be with you.

Nathan Settembrini [00:01:15]:
So we'll jump right in. I heard, you know, I've heard this moment that we're in, in the AI revolution referred to as kind of the AOL dial up phase of AI. Would you agree? What are your thoughts there?

Peter Coffee [00:01:28]:
Well, it's even possible that I'm the guy you're quoting to. Call it the AOL dial up phase might sound like a criticism. It's really not. I mean, if you think about the degree to which pre AOL being on the Internet was something that implied a level of technical comfort, maybe even access to privileged resources in academia or even in the national defense community and AOL, all of a sudden you had, you know, CDs in the mail and everyone was being invited to come in and discover this resource. But at the same time, you had to make an event out of it. You had to, you know, I'm. People would say, well, I'm going to be logging on now, right? Yes. If you remember the movie the Social Network, there's one moment when a young undergraduate is talking about how freakishly addictive Facebook is. She's on it like five times a day. And that sounds so anachronistic now because in a sense, when is any of us not on it? On Facebook, on X, on whatever, it's just always there. It's like the air we're breathing now. And so the AOL dial up phase of AI to me means, you know, when you're using it, you're making a decision to use it, you're going and logging in on a ChatGPT or a Gemini or an Anthropic. You're highly aware of it and it's an addition to the things you do instead of just being part of the things you do. The parallel I've drawn is that today Siri doesn't even have to make a call out to that server in your pocket, called your phone. It just runs on the watch. The migration of the capability to the edges of the network addresses a lot of people's concerns about privacy if the data never even leaves your device. And this is the Apple intelligence model. I think they're talking about it in very much the right way. They're taking a little longer to deploy, but that's because they're doing it right. Instead of starting by building something with massive, highly capital intensive, highly energy intensive, very general purpose capability, that they're building something that begins life being personalized, tailored the data that's relevant to you and that's where we want to wind up. And they're starting there, even if it takes them a little longer instead of starting with the wrong thing and then having to migrate to that right thing. So saying it's the AOL dial up phase doesn't mean it sucks. It just means the manner in which we engage it today is kind of lumpy and kind of discontinuous from our other everyday experience. And it's going to be merging into everyday experience quite quickly. It already has, I think, almost unrecognizably in the last two years. I saw someone the other day ask an audience, how many of you use AI daily? And a fairly large number of hands went up. And I was waiting for the punchline with, well, the rest of you just don't realize it. Because I don't think there's very many people living in any developed country today who aren't interacting with some kind of AI capability daily, whether they realize that, whether they're making a conscious effort to do that or not.

Nathan Settembrini [00:04:54]:
Yeah, I loved your metaphor in that presentation you gave about the home motor. Could you just give us a little more?

Peter Coffee [00:05:00]:
Yeah, I love that one. Thank you for remembering it. There's an old Sears catalog in which a thing called a home electric motor is proudly offered for sale. And you would buy the home motor and then you'd buy a fan attachment for your motor, or you'd buy a connecting coupler to attach it to your sewing machine, or you would buy a blender attachment for the motor. And in that very same Sears catalog on another page is the next generation sewing machine. The Franklin electric, which where with the same kind of enthusiastic language they talk about how the motor is part of the machine, you don't have to connect, costs you only 2 cents an hour to run. And you will have, and I think the phrase is a much more satisfactory sewing experience with this, and that is exactly the cusp we're on. You can go buy a home electric motor, that is to say a great big large language model, and make a big training investment and do a whole bunch of stuff. Or you can build a great big data center full of today's processor chips that may or may not be obsolete in three years. Or you can be looking for a way to have the AI so intrinsically, inherently embedded in what you're doing that you're not even conscious of it. You're sewing. And the other connection I've often made is with something that the philosopher Martin Heidegger wrote in a book called, as translated from the German, the Origin of the Work of Art, where he talks about the if you do this right, your tool disappears into its function. And when people look at me puzzled as if what does that mean? I say, well, if you're hammering a nail and you're thinking about the hammer, it probably means it's the wrong hammer for the job, or it's a broken hammer and you're noticing that the head is wobbling or the handle is cracked. If it's the right tool and it's in the right condition and you know how to use it and you stop thinking about the tool, you're just thinking about the nail. And if you're thinking about the AI, you're doing it wrong. This is something that my 19 years at Salesforce have accustomed me to discussing with people, is that we brought AI into the products in such a way that you weren't conscious of using it. All of a sudden the product was doing intelligent lead scoring, next best action, next best content recommendation. It just became more assistive to. You didn't have a sense of having to leave your work to go up to the mountaintop and consult the Oracle at the top of the mountaintop. The Oracle came down to you and assisted you in the environment where you were already getting your work done and crucially, where your data were already at hand. Because that was the other really key issue of the early so called AI Winters, is that you would build these processing nodes with names like inference engine, and then the question would become, okay, but right now they're ignorant savants, they know nothing about my business. How do I teach them that and the first AI winter was really a crisis of coding. How do we obtain knowledge about what human expertise is? How do we obtain knowledge about what context is? And Marvin Minsky early on, like, like you're in the early 60s, was saying, oh yeah, within a generation we're going to have human equivalent intelligence. Well, less than 20 years later, he's quoted as saying, this is the hardest thing we've ever tried to do. And I asked people, what do you think he realized during that time? He realized that common sense knowledge is extraordinarily difficult to characterize and even more difficult to encode and even more difficult to apply in any meaningful way. Again, the thought experiment I ask people to indulge in is you're walking down the street, you pass the 7 11, you suddenly remember that you finished your last quart of milk that morning and you need to pick up some milk on the way home. I can almost guarantee you haven't spent the whole day thinking, got to remember the milk, got to remember the milk. And yet the contextual moment of it being available to you makes the need surface in your memory and so you can take action on it. Imagine trying to build systems that have a complete understanding of the world around them, but don't spend most of their time deciding what they can safely ignore at the moment. Your brain does this with about 20 watts of power. We don't even know how to encode it, let alone how to run it in hardware that won't melt the ice caps. And that is among the real challenges of so called general Artificial intelligence is building things that can know everything and ignore almost all of it at any given time.

Nathan Settembrini [00:10:01]:
Yeah, like the reticular activation center that we have in our brain that helps us spot the thing that we're looking for, like the red car, you know, that whole thing.

Peter Coffee [00:10:12]:
Yeah, yeah. Well, if you walk into the Stata center at MIT and turn in one direction, you'll find people trying to figure out how the brain does what it does, and maybe building algorithms and hardware models to test their theories of the brain. And other people who don't think they care how the brain does it. They're just trying to build artifacts that do things that require what we call intelligence when they're done by a person. And there's a reason why they need to be overlapping because if you have no understanding of what the brain does and how it does it, you may build clumsy and unscalable technical solutions because you've ignored what our brains have figured out how to do. There's also another really great quote from Minsky where he says, you need to understand something. You don't have a mind, you have a brain. Mind is something your brain does. What we call the human mind is a behavior of the meat machine between your ears. If you look at someone, if you look at a positron scan of someone doing a task like reading, the vision center is involved, the speech center is involved, many different parts of the brain. And interestingly, back in the 1980s, Minsky actually used the word agents to describe the pieces of the brain that are each doing part of the task in ways that are not directly aligned with each other, but somehow collaborating, and that's where we need to go with AI agents is building things that each have a piece of the job, that manage to collaborate and gain more than the sum of their parts out of their interaction without a human needing to exercise conscious and deliberate supervision of that process.

Nathan Settembrini [00:12:02]:
Wow. Wow. Well, Peter, let's go on to our next segment here. I'm calling it Data Secrets through the Decades. Okay, so this is the Data Secrets podcast. It's all about uncovering insight and data, turning it into action. You've been analyzing trends and sharing your perspectives on technology for four plus decades over the most profound leaps in technology that humanity's ever experienced. Right. From, yep, PCs that sat on your desk now to AI and everything in between that.

Peter Coffee [00:12:38]:
Oh, I go back to slide rules. I really do. Yeah.

Nathan Settembrini [00:12:42]:
Yeah. So would love to basically walk through decade by decade and just talk about, like, your thoughts and stories, things that you saw, insights that maybe you caught that others missed, the data that maybe brought some of those insights into your brain. And then some of the implications that came about for business leaders, because now we're not sending letters, we're sending emails. Right. Like that kind of stuff.

Peter Coffee [00:13:11]:
Well, I'd love to do that. And I will begin actually with my first year as an undergraduate.

Nathan Settembrini [00:13:17]:
Okay.

Peter Coffee [00:13:18]:
In which. Well, actually, I'm going to begin with high school. High school chemistry, where the last exam we had before the Christmas break at the middle of the year was not a chemistry exam, but was a test of our ability to do basic work with a slide rule. Why? Because a calculator was a $400 purchase in 1970s money that was more like a multi thousand dollar purchase. Basically, the price of a good PC today was the price of a good calculator then. And the teachers wanted to be able to give us problems where they didn't have to go to a huge amount of trouble to make the math easy. So they wanted to be sure that everyone in the class could do basic computation on a slide rule so that the rest of the year the problems didn't have to be baby talk problems. So slide rule. Fast forward to freshman year college where there's a class called Differential Equations that included, because you need it for the non trivial work, some, I think one lecture on so called numerical methods where you do rapid iterative approximation calculations because you can't do a perfect solution. And to do the problems for that lecture you went to the lab and signed up for an hour on a programmable calculator because you couldn't possibly assume that every student would own a programmable calculator. One year later that class was renamed from Differential Equations to Differential Equations and Numerical Methods because in one year the cost of owning a programmable calculator had been cut in half and the new calculators were way better than the old ones. And fast forward again another decade or two when my son is taking the same class and it's assumed that every student in the class has a laptop that can run a piece of software called Matlab that can do this kind of thing on steroids. That transition is not merely the ability to do the old stuff faster. The scope of the class substantially increased its acquaintance and reliance on the so called numerical methods, because now those were something everyone could afford to do. And that's what we have to understand about AI. If you're using AI to do the things you've been doing, better, faster, cheaper, you're leaving an enormous amount of value on the table. The really interesting opportunities are to do the things you've been dismissing and maybe even unconsciously dismissing because they were just too difficult to do. And this is something Stephen Wolfram has said about this. If you have something that you're getting done today with, I'll call it traditional software at maybe an 80% level of what you'd like to be doing, and AI gets you up to 85 or 90%, that might be a nice return on investment. But what about the things that you haven't been doing at all because the problem was considered to be too poorly structured or the data was considered to be too voluminous and rapidly changing? What about things that you're currently doing at a 0 to 10% level? If you can get those up to 40 or 50%, that's a much bigger gain. But you have to look harder for that opportunity. The applications everyone uses every day and says, oh yeah, couldn't this be better? Those are easy to identify. The things that no one is doing at all. Because it's so hard to do with the old methods where you say, you know, you could have this now that's a much more valuable conversation. But it takes a significantly greater effort to get that conversation started. And that's again the difference between an AOL dial up era of AI and what it's going to be when people think about the availability of that capability up front instead of just thinking of it as the lipstick you put on the pig that you've had snorting around your office for years or decades.

Nathan Settembrini [00:17:18]:
You were an engineer and then you joined PC Week as an editor columnist.

Peter Coffee [00:17:22]:
I did, right. Well, it wasn't an obvious transition. I mean, I was an engineer. I was an engineer in big oil. When the first desktop machines started to become available. I mean, when I started, we were on timeshare terminals, we were on green screens, we sometimes had to deal with punch cards. When I started out and made a transition from oil to aerospace, we were submitting card decks to the big control data machine in the basement. If things went well, you got back a five page printout. If something went wrong, and it could be anything as trivial as leaving out a piece of punctuation in your input, you got back a 1 inch thick output because the machine had done a core dump on you and you were. And this was what a waste of dead trees, right? That was the early days. And then these PC things came along. And in the early days, before everything was assumed to be an IBM compatible, there were five or six viable candidates out there that differed in some pretty key areas. So you would buy the Texas Instruments version of your spreadsheet package or the Digital Equipment Corporation version of your spreadsheet package. We didn't really have a universal standard. And the company management said, well, we're not going to have five or six incompatible versions of desktop computing in the company. They actually had a VP go to a lunchtime meeting of the IBM PC user group, write down the names of the four people who seemed to be answering most of the questions and, and tell us, you guys are going to be the PC standardization group here. You're going to have to herd the cats and get this company full of engineers and PhDs, not merely to accept, but actually to prefer the company standard. So it was a very interesting point of view. We were not allowed to mandate very much of anything. We had to create a desire to adopt the standard rather than imposing the standard because it was just that kind of a community. So we had to invent a lot of things we invented newsletters, we invented real time user support. And the process of doing that led me to have an opportunity to be on some panel discussions about creating this discipline of desktop computing management in a large organization. That was a new thing. It used to be the computers were the hulking giants in the data center and everybody just had terminals. And now we have this radically decentralized distributed capability. And that's how I wound up making that transition from an engineer to an engineer who was an informal expert on these PC things, to someone who did that for a living, to someone who wrote about doing that for a living. And then eventually to 18 years at Ziff Davis Media PC Week when I started, and at that time networks were a little rare. Most PCs weren't on a network. You had to go buy a modem to get on a network or string ethernet cable. While I was there, about one quarter of the readers got the special supplement called, I think it was called Netweek. And then gradually all PCs could be assumed to be networked. And you remember when the imac first came out and they said there's the three step process of using an imac plug in the power plug in the network. Oh wait, there is no step three. That was a huge, huge reduction of the perceived complexity. You used to have to buy TCP IP as a piece of software that you installed on your machine. It was like buying a car and having to install the transmission as an after effect. The idea that every machine would now come with a built in modem and a built in ethernet port and everyone was assumed to be connected again. Many massive, massive change in behavior. And so PC Week turned into E Week. And we would say for a while we made it into eWeek before we changed the name because now it was about the dot com enterprise instead of what we actually were called. When I started, which was the national newspaper of IBM standard microcomputing. And I remember having a conversation very early on with when we created that group, the so called Gang of four that turned into the PC Standards group. I said, you realize at some point having a magazine about computers is going to seem like having a magazine about office copiers. You're going to have a computing column in Women's Day, you're going to have a computing column in your stamp collecting magazine. Everything that people do, they will. Among the things they'll want to know is well, how can I use my computer to help me do this better? Because the computer just becomes part of the background. And being a computer hobbyist today, most of the people I know. Who behave the way computer hobbyists behaved then are gamers who are really getting in there. They actually still care about stuff like video card performance and things. They're a relatively specialized audience now compared to a time when PC magazine was the largest volume newsstand publication in the country, when the postal service was complaining that their postal carriers were getting back problems from delivering these half inch thick magazines once a month. And so they moved to a higher frequency and people said, oh, you'll never be able to sell ads in that. Yeah, they could. It was the largest revenue newsstand publication. And I don't think that's the case anymore because the tool has disappeared into the function. As Heidegger said, we just assume that you'll have this capability in your pocket, on your wrist, available anywhere. Try to go to a museum anymore without having them assume that you're carrying a smartphone. So things that they used to do by handing out dedicated devices like the audio tour of the museum, well now that's scan a QR code and find out about the painting you're looking at. It's become so wired in to everyday life that trying to get through the day without a smartphone is getting to be kind of retro.

Nathan Settembrini [00:23:47]:
Yeah, totally.

Peter Coffee [00:23:49]:
It's like trying not to. It's like trying to live without credit cards.

Nathan Settembrini [00:23:52]:
Yeah, yeah, for sure.

Peter Coffee [00:23:54]:
You just can't participate.

Nathan Settembrini [00:23:57]:
So you've taken us from kind of 70s, 80s and kind of getting into the 90s, right.

Peter Coffee [00:24:02]:
Pretty much got us to right around the year 2000 at that point.

Nathan Settembrini [00:24:05]:
Okay, yeah, so we're, we're on the.

Peter Coffee [00:24:10]:
Verge of my joining Salesforce, aren't we?

Nathan Settembrini [00:24:12]:
Yeah, yeah, we're getting there because that was 2007.

Peter Coffee [00:24:16]:
Yeah, I remember.

Nathan Settembrini [00:24:17]:
So my experience, I was born in 84, um, and my grandfather was a tinkerer. He was a navy guy and had computers and, and so he would, he actually got us our first home computer that, you know, didn't live, you know, in your backpack. It lived in the computer room, you know.

Peter Coffee [00:24:42]:
Oh yeah, it was a box. It's like the old color TVs that were a piece of furniture.

Nathan Settembrini [00:24:47]:
Right, right, exactly.

Peter Coffee [00:24:48]:
Right, exactly.

Nathan Settembrini [00:24:49]:
And he would give us these little floppy disks with games and you know, you'd go to the, the green screen and type in like directory and then launch, whatever, all of that stuff. Yeah, yeah. And then in the beginning was the.

Peter Coffee [00:25:01]:
Command line as Neal Stephenson wrote. Yeah, that's right.

Nathan Settembrini [00:25:04]:
And, and to get on the Internet, I'd have to say, hey mom, like don't, don't get on the Phone for the next.

Peter Coffee [00:25:09]:
Don't pick up the phone. Right, right.

Nathan Settembrini [00:25:11]:
I'm going to be on Instant messenger talking to a lot of.

Peter Coffee [00:25:13]:
A lot of painful family moments on that one. Right?

Nathan Settembrini [00:25:16]:
Yes. Yeah. And so, so we moved to a.

Peter Coffee [00:25:18]:
Period when literally it went from being dial up to being online. And you can see it within my own family. Our oldest child grew up in a household where you walked across the room and pulled the encyclopedia off the shelf to look up something. And even his own brothers, let alone different generations, even his own younger brothers at that age were in a always on DSL household where it was faster to look it up online than to walk across the room and pull the encyclopedia off the shelf. I mean, within a matter of single digit years, that transition took place. Which brings us to that moment of epiphany that Marc Benioff had those years ago when he said, why are we still selling office software as if it's still like the floppy disk era where you have to buy the software and install it on your own machine. Everything else you get over an Internet connection. Why aren't you getting capabilities like CRM over the Internet? We weren't really calling it the cloud back then. We called it on demand and IBM called it utility computing. The phrase the cloud hadn't really emerged quite yet. And in a lot of ways, I have told people the biggest achievement of the first 10 years of Salesforce was not so much cloud computing. It was rehabilitating the idea of CRM from being a complicated, technically demanding infrastructure, investment in a system of record that people would log in on Friday afternoon because otherwise you update their call sheets, because otherwise they wouldn't get paid to turn that instead into something that you started using between your first and second cups of coffee on Monday morning, because it was easy to get on, easy to understand, updated itself over the weekend without you having to have a system administrator do that. And it was going to improve your sales performance for the coming week instead of just recording your sales activity for the week that had just ended. And the transition of the way people engaged with the system was made possible by the cloud, but the cloud was not the product. And you have to remind people, if you think the thing you've gone to all this trouble to design and build is the product, you're probably wrong. No one actually wants your product. What they want is the benefit and the outcome that it delivers. And if you focus on that, then technical change stops being a threat and becomes an opportunity. Yeah.

Nathan Settembrini [00:27:52]:
So he took an existing thing that people were used to Doing like, I need a place to store my customer information and my interactions with my customers. And you change the mechanism. Like you wouldn't be able to performance.

Peter Coffee [00:28:06]:
Improvement tool instead of just a system of record artifact. Yeah, yeah.

Nathan Settembrini [00:28:12]:
So was the. Do you think the hardest part was getting people to understand the whole, like, hey, you just log in here, you don't have to install anything because like, I imagine a lot of the IT administrator type folks were like, oh no, that's my job is maintaining the software.

Peter Coffee [00:28:30]:
There was some of that concern that we were, you know, displacing people. But, you know, Peter Drucker said there were seven sources of innovation operations opportunity. And there are three in particular that I've considered to be the trifecta of Salesforce. The first one is what he calls a moment of incongruity. And with Salesforce, it was having people suddenly look at the way they were doing their jobs and say, this is stupid. This is cumbersome, expensive, technically demanding. Why am I putting up with this? Because once they're asking that question, you can go to the second source of innovation opportunity, which is new knowledge. The new knowledge is now Internet connection is everywhere. Software can be delivered to you as a subscription service instead of as a product. But then the third change of perception, the initial reaction that people had was put my sensitive deal closing data on a system that isn't even in my own building. Why would I take that risk? The change of perception was when people had pointed out to them, sometimes by me, that the data breaches that they saw in the news were not breaches of Salesforce or Amazon or Google or other professionally administered, large scale, state of the art, vigorously tested and maintained facilities. They were data breaches of systems that were in your basement that hadn't had patches done for the last year that were still using default passwords that everyone knew because no one had changed the password when the product was installed. Things like that. That perception that DIY was the expensive and dangerous way to do it took some time to achieve. And again, that was a large function of my, of my job for my first several years at Salesforce was to help people have that moment of recognition. You know, the assumptions I've been making about what's risky and what's safe, the assumptions I've been making about what's expensive and what's hard. Some people would say, well, I don't want to pay you guys a subscription fee every month. Once I buy software, I own it. We have to say, look at the three year cycle here. Look at what you pay to administer it, what you pay to do updates, what you pay to do infrastructure upgrades. All of a sudden our monthly subscription fee looks like a pretty painless alternative to that. And again, this took some time. It was, it really was about helping people achieve and understanding. And you know, when that happened, that happened when people started using the term the cloud, which is right around, you know, maybe the end of 2007, 2008, you do your Google Trends search. That phrase cloud computing really comes bubbling up out of the, out of the world because people used to draw on a whiteboard to do this to this, to this and they would, you know, sketch a little cloud. And then this happens over here in the Internet. The metaphor of the cloud is that I don't have to care about the physicality of it. We had a wonderful video once where a Hollywood post production company, one of their people pointed at their WI fi router hanging on the wall and said, well, that's our data center. Then we had a conversation with Trunk Club men's apparel retail, where they talked about touring a new office space. And their realtor said, and over here is a great room for your data center. And they just kind of looked at her and said, we're not going to waste prime downtown office square footage on a data center. It all happens in Salesforce. We're going to use that room for private fittings. And that idea that suddenly you had the ability to let your processing be taken care of. The non value adding tasks shouldn't be done by your most important value adding people and your most important value center resources like your downtown office square footage. Those things should be someone else's problem. That's what cloud should mean.

Nathan Settembrini [00:32:31]:
Yeah. It's been abstracted away from the modem that you hook up to the.

Peter Coffee [00:32:36]:
Oh yeah, yeah. You don't shift gears by opening a little door in the bottom of your car and rebuilding the transmission on the fly. You push a button that says drive.

Nathan Settembrini [00:32:45]:
Right.

Peter Coffee [00:32:45]:
And you need to achieve that level of abstraction so you can focus on value adding differentiation instead of being bogged down in non value adding, non differentiating maintenance tasks.

Nathan Settembrini [00:32:59]:
Yeah. What led you from like the editorial side of technology to want to join Salesforce?

Peter Coffee [00:33:07]:
Well, you know, cloud computing happened to print media too. We had a conversation once where our editor in chief said, you know that whiteboard I used to keep next to my office door where we would record our scoops? I'm taking it down. We're never again going to have a scoop in print. We used to be able to pride ourselves that if we got something in the office on Friday morning, we could have it installed, tested, written up, reviewed, and on 440,000 desktops, ink on paper on Monday morning, well, that was impressive. But now we could have it on people's screens Friday night instead of Monday morning. And you had this fantastic inversion where we used to have a website that was basically highlights of the print publication. And in the space of a year or two, that turned around to where now the print publication was. Follow ups and in depth analyses of stories that had broken online. And we did that to ourselves. We participated in that change. And so at some point, being in print media really wasn't a place where I could do cutting edge work. I could have the opportunity to help people grasp a new opportunity and to put those opportunities to work more fully. At Salesforce, where I was in something of an evangelistic role on helping people grasp the opportunities of the cloud, I was able to do that more effectively at Salesforce than I was able to do it in a publishing industry that many people regarded as being a legacy eyeballs business. And fortunately, I had gotten to know Marc Benioff when he was at Oracle and I was at eWeek. I was coordinating product evaluations where he was representing Oracle's interests at the time. And so we already knew each other. And when they began to think, you know, we've got this CRM thing that we sell, what if we offered people the chance to take the CRM off the top and just buy the platform capability? Now, we would not just be selling to sales professionals, we'd be selling to systems integrators. We'd be selling to independent software vendors, we'd be selling to IT managers. We don't know how to talk to those people. We need a whole new piece of our competitive intelligence business that talks to that audience. Hey, I know this guy named Peter Coffee. Maybe he'd be interested in the job. And that was my first job at Salesforce was to, as it were, grow another lobe on our brain that understood the competitive environment, the buying criteria, and the customer base for a platform business that the company had never had before. So I joined the company just at the time that we put Apex Code into Preview and just at the time when we were getting ready to go to market with a platform license. And my first job was called head of Platform Research because that was just a new space for us in 2007. Fast forward to a couple of years later when one of the websites, TechRadar, I think it was, headlined an article salesforce is a platform company, period. Because for some time the majority of the work that our systems do, services, applications that we did not write, that we do not sell. In a lot of ways, the products, the things that look like products is what I usually call them like sales cloud or now agent for sales. Those are really demonstrations of the platform. They are not developed with secret capability. We are the alpha user of our own platform and the things that we offer to people are really someone actually the CIO at Caterpillar Tractor said this. Salesforce doesn't have products, it has serving suggestions. And when we brought out our, when we started to go vertical with finance cloud and healthcare cloud and so on, John Wilke at the time was running that operation. He said any customer who had sales cloud and service cloud could have built our financial services cloud, could have built our medical cloud. We've gone ahead and done that as a demonstration of what best practices would be. But any piece of the process that you want to tailor or extend or replace, you would have the full ability to do that. And that's what a platform really is, don't you think, is when you can repurpose the underlying capability, it comes ready to use to do something that many people find useful. But anything you want to add, you won't have to dismantle it. I tell people it's like a Rubik's cube. If I hand you a fully solved Rubik's Cube, it seems like a very simple object. It's got, you know, a white face, a red face, a green face, a yellow face. Okay, I understand this. The moment you want it to do something different, you don't have to take it apart and you don't even have to understand the mechanisms on the inside. You just twist it into the shape you'd prefer. And that's what I want people to understand about the architecture of a cloud native metadata based technology portfolio like the Salesforce platform is whatever you don't like, you can change.

Nathan Settembrini [00:38:21]:
Yeah. I've always wondered why Salesforce never rebranded as something genericized from the CRM. Right. Like there was around the time you joined the Forest.com concept.

Peter Coffee [00:38:34]:
When I joined this is a funny story. The domain force.com was owned by a family named Force. They'd had it for their family business for some time. So it took us a little while to get that. And there were times when I would have an account team say, could I just get business cards that say force.com and not salesforce.com because if I show up with the salesforce.com brand. They show me down the hall to talk to the VP of sales. Who else could I be there to see? And I actually had GE on stage with me at one point at Dreamforce. GE, as you know, makes jet engines. They haven't made light bulbs in a long time. And I said, you know, my ambition for Salesforce is that we'll have our GE moment when no one thinks of the sales syllable in the name as restrictive. It's a brand, it's got fantastic equity. We've got Salesforce towers in many of the major cities of the world. It's still occasionally a challenge to get people to see us through a lens of sales capability is a demonstration, not a limitation of what we do. But at this point, I think it is pretty clear to me that changing the name of the company is probably not a net advantage.

Nathan Settembrini [00:39:50]:
Yeah, yeah. It's like Google tried it with Alphabet, but everyone still just calls them Google. Right.

Peter Coffee [00:39:56]:
It's like being the artist formerly known as Prince, you know, and the same thing with the social media platform that changed from a cute little blue bird to a menacing looking letter as its brand. Everyone still refers to it as, you know, the company that used to be so and so. I think that's a pretty good litmus test for whether that rebranding was a useful thing to do. I think the answer is no. Now, there are examples where that has worked. You know what a Taser is, right?

Nathan Settembrini [00:40:26]:
Yes.

Peter Coffee [00:40:27]:
Okay. Taser used to be the name of the company. Taser is now a brand of a company that has been relabeled as Axon. And if that name sounds at all familiar is because when you see police body cam footage on the evening news, I'll bet the word Axon is somewhere in the margin of that body cam frame. They now sell an overall package of public safety and criminal justice capability in which the non lethal device, that is to say the Taser, is part of the package. But Axon is now the umbrella brand that includes all of that. That's been a successful rebranding for them, I think. But they're among the exceptions to a company taking a really well known brand and somehow managing to rise above it. It doesn't happen often.

Nathan Settembrini [00:41:13]:
Yeah, well, so let's jump through maybe the next 10 years as Salesforce. There's a combination of things happening right. Where there's a lot of data acquisitions of Mulesoft, Tableau and Informatica as recently as this past week.

Peter Coffee [00:41:33]:
Well, it all makes sense if you think about it. If you think about it, in an Evolutionary way. Step one, get people to log into their CRM on Monday morning instead of Friday afternoon.

Nathan Settembrini [00:41:42]:
Yep.

Peter Coffee [00:41:43]:
Step two, now that they're living in that system, make sure that it knows what's happening outside the sales department. Step three, as more and more of your customer creation and curation starts to happen out in the world of social media, make sure data can come in from outside your walls as easily as you get it from inside your walls. So now you're starting to talk about things like the, you know, the mulesoft acquisition, but now you've got this enormous volume of largely unstructured data that you didn't have the opportunity to organize to meet your own needs. You're going to need to repackage and retransform that data to make it meet your needs. Okay, let's go buy Tableau. Well, now you know a lot about what's going on. No sparrow shall fall, not just about the customers you have, but about the customers you previously almost had, and now you can take action to get them the rest of the way. Now we need to form ecosystem networks and not just internal conversations about that. So now you go buy Slack. Now you've got this fabric of collaborative action informed by an entire planet's worth of data, tailored into well defined selling, servicing and marketing processes. This is all great, but now the complexity and the speed of change have reached the point where you really need the assistance of machine intelligence to do that kind of thing quickly, effectively, and at scale. When I frame it that way, the acquisitions over the last 20 years really make a lot of sense, don't they? At every point we've been saying, what's the current boundary of people's ability to get value from what they have? Do we partner for that? Do we acquire that? Do we build for that? And we've done some of each. We've built some things. The AI capability is largely native and has a really tremendous portfolio of patents and published research to do things like developing focused, tailored models instead of general purpose models that do a worse job at greater cost. There have been the partnerships. The partnerships are enormous, including partnerships with companies like Informatica. And at some point you ask, well, is there a reason why it would make more sense for that to be part of the family instead of just a very close partner? The mulesoft acquisition would be an example of that. Bringing them in allowed us to do some things more quickly than we could have done merely working with them at arm's length like a partner and bringing in Informatica. I'm Very pleased that they're closer to us now. We had a great relationship before, too. We weren't taking a competitor off the table. We were just further integrating an existing partner.

Nathan Settembrini [00:44:30]:
Gotcha. Yeah. When that was announced, it took me a little bit to try to understand the why behind that.

Peter Coffee [00:44:39]:
I always have to understand the why. I always say, tell me why we need to acquire this company instead of just having a strong partnership with it. And one of the great examples of that would be the exact target acquisition, which people may have forgotten. Exact target was huge at the time. 2.6 billion was a new record for us. Fast forward, where Slack is 10 times that. But. And my first thought was, we're spending $2.6 billion so we can check the box and say, now we've got an email capability. My first reaction was skeptical. Then we videoed into the exact target version of Dreamforce from one of our own management meetings. We had our 600 most senior managers all watching this video and. And we were seeing customer journeys. We were seeing the ability to instrument an email and know when it was read, how many times it was reread, to whom was it forwarded? And the next thing you know, the word journey has just proliferated through the rest of Salesforce. That was a fantastic example of an acquisition that did more than add to us. It infiltrated us. And I mean that in the most positive possible way to create a new point of view on how we did what we do.

Nathan Settembrini [00:45:55]:
That's amazing. Yeah. I was actually personally impacted, kind of third order effects by the exact target acquisition, because Exact Target had purchased Pardot, which was founded by David Cummings, who is based in Atlanta. He took the proceeds from that and invested it into Atlanta and built the Atlanta Tech Village. And that's where I really got kind of a spark around technology and startups and started going to all these meetups, and it really changed the trajectory of my career.

Peter Coffee [00:46:28]:
Well, that's an interesting example of another thing that the acquisitions have done for us over the years, because our presence in Indianapolis as a side effect of the. I think that was the exact target acquisition.

Nathan Settembrini [00:46:38]:
I think that's right. Yeah.

Peter Coffee [00:46:39]:
Our presence in Atlanta, our presence in New York. Well, I think that kind of began when we bought Buddy Media for all of $690 million back in the day and started really to think of our business as helping our customers curate their social media audience and really target their messaging to the people who were going to be most interested in what they had to say. And in the process, we became much more a presence in that New York media and publishing community. And so the acquisitions have done more than just buy stuff. They've also extended the nervous system and the awareness of the different cultures that Salesforce serves.

Nathan Settembrini [00:47:22]:
Yeah, that kind of brings us to the present with Agent Force and all that. I'm sure just your maybe high level thoughts on maybe the next 10 years.

Peter Coffee [00:47:34]:
Yeah, well, the thing to understand about agents is that if you just take an AI and let it do things, you're not really getting it. And I forget who I'm stealing this from, but I'm combining this with a couple of things that I've heard from different people where someone said, you know, every time you think you're looking at a web page on Amazon, you're not. You're not seeing an HTML page, you're seeing something that's been turned into browser consumable view of. I think the most recent number I've heard is five to 600 different microservices that are all collaborating in real time to compose what you're most likely to want to see and what they would most like you to see. And agents need to be visualized in terms of, like I said before about Marvin Minsky's comment about the way pieces of your brain collaborate. If you aren't thinking in terms of dozens or at some point, many dozens or even hundreds of agents that are each responsible for access to different and specific data sets, capability, portfolios. You notice how on Alexa they don't even call them applications, they call them skills. You add skills to your Amazon Echo capability. If you're not thinking that way, then you won't build systems that can scale to that complexity. The funny thing is, this is really old stuff. In 1994, Rosenschein and Zlatkin at MIT wrote a book called, I think it was called Rules of Encounter, about this idea of what happens when the agents are talking to each other and carrying out actions on behalf of their people. This is a discipline that's gotten a lot of thought for literally decades. And now we're able to deliver on these thoughts. We're now able to deliver this at scale, at a low price, with a high level of abstraction to your earlier point where people don't have to be an expert on agents. If you've been through the experience of crafting an agent on AgentForce, you discover, wow, I can point it at a website, tell it what I want to do, and it will go out and explore that site and pull things together and create something that can do something that today would require me to know four or five different pages within that site and be able to move back and forth between them and copy and paste maybe and make that happen. For me, it's a marvelous moment. The first time you realize the technology really can do that for you now already, it can do that for you now. This is not prospective. This is existing capability.

Nathan Settembrini [00:50:13]:
Yeah. What do you think business leaders misunderstand about this present moment and the future?

Peter Coffee [00:50:19]:
Well, the tendency is to want to dance with the one that brung you. People have investments in their perceived competence and expertise and they don't always give up readily. The idea that what they knew that got them where they are might not be what's going to take them where they want to go. And being prepared to say, okay, this is how we got here, doing more of this better might have a superficial return, but we'd leave an enormous amount of money on the table. And getting people who are ready to step back from their perceived expertise and leave their comfort zone say, remember, all that matters is what does the customer expect? And someone said the other day, you're not competing with your competitors. You're competing with the best experience people have had all day from anyone. You think you sell shoes, but you're competing against pizza delivery. You think that you are in the rent a car business, but you're competing against Netflix and Amazon. And the reason that Amazon just seems to be taking over the world is because they continually take something people are used to doing and say, yeah, but how would you do this if you'd always known that people were going to have Internet everywhere and tomorrow? How would you do this if you'd always known people would have the assistance of machine intelligence in bringing the complexity together and simply asking and answering the question that's on your mind right now.

Nathan Settembrini [00:52:02]:
Awesome. Well, Peter, what's next for you? Tell us a little bit about what you're up to now.

Peter Coffee [00:52:09]:
Well, I ended my time at Salesforce in November after literally a couple of years of thinking, yeah, at some point I should declare victory and move on. And every time there was either a new capability that I wanted to help us deliver or a particular challenge that I wanted to help us deal with. And the time just came when I said, you know, there are some things that just need full time attention that I can't give them when I'm still doing what I do at Salesforce. And so I just had to make that decision. Six years ago, my wife and I formed a private foundation to focus our concerns and support for things in the area of climate change. Mitigation and adaptation, environmental conservation, education, exploration. So I made a running start and said it's time for that to stop being my company approved volunteer activity and just become my full time job. I'm not paying myself anything, so I'm a full time volunteer. But now I'm full time with our 501C3, which is called the foundation for Intelligent Life on Earth. Funny story, I was out to dinner one evening maybe 10 years ago, and for some reason I thought, I wonder if intelligentlifeonearth.com is a domain that anybody has. And no one did. So I just went ahead and bought it. And I think I had the faintest idea in my mind that it would be cool someday to have a thing called foundation for Intelligent Life on Earth. And so that's where the brand came from. And in a matter of the last few weeks, as I've ended my time at Salesforce, we've taken things that were nascent ideas all along. Building a blog for the organization, strengthening our partnerships with groups like the Nature Conservancy, with mit, with ucla, with some other environmentally oriented groups, and really being able to say, look, this is what I do. Now. These are problems that I say there are two kinds of problem in the world. There's the problem that you can, when you notice it, fix it. If you get a headache, you take an aspirin. The other problems where if you wait until it's causing you real discomfort, it's gonna be much, much worse. And you know, not to get too graphic, but headaches, you wait till you have the headache. You don't wait until the cancer tumor is too big to ignore. As soon as you detect it, you start doing something about it. Climate change is kind of like that now. In fact, it's been that way probably for decades. But I'm really focused now on helping people decide to treat things with urgency, even if they're still easy to ignore compared to other things. Because if you don't deal with these things now, it's going to get difficult. And you know, there's nothing like having grandchildren, which I do now, to make something that's going to happen 40 years from now seem a little more urgent than it would if you were thinking, well, I'll be dead by then, maybe I will, but my great grandchildren are still going to be needing to know, how come you didn't do something about this? Great grandpa?

Nathan Settembrini [00:55:16]:
Yeah. So this is you doing something about it. I liked your blog post about the two types of numbers around climate change.

Peter Coffee [00:55:23]:
That was just this week, right? The two kinds of numbers, there's the measures of size and there's the measures of, of rate. When you start multiplying them together and look at things like the melting of the Greenland ice sheet or the deforestation of the Amazon, you realize what might not seem like a very scary trend. When you multiply it by the scale and factor in the momentum of the process, you start to realize, you know, this actually is something I ought to be doing something about today and not, you know, not waiting until next year or next decade to get around to it.

Nathan Settembrini [00:55:55]:
The, the metric that hit me in that post was football fields per second of deforestation.

Peter Coffee [00:56:03]:
Yeah. I saw the other day that deforestation in the world today is about three football fields per second. I think that was the number.

Nathan Settembrini [00:56:10]:
It's great.

Peter Coffee [00:56:10]:
Kind of shocking, huh?

Nathan Settembrini [00:56:12]:
Yeah, yeah, yeah.

Peter Coffee [00:56:13]:
Hectares. Hectares per year is kind of a big fuzzy number. Football fields per second makes it a little bit more, I think, tangible to people.

Nathan Settembrini [00:56:24]:
Yeah, that's wild.

Peter Coffee [00:56:26]:
Yeah. Well, well good. Then, then, then we're head. Then this conversation is having the effect that, that we intend the foundation to have, which is to get people thinking as, as Mark did 20 years ago. Get people thinking differently about the world around them and the capabilities and the challenges that it presents.

Nathan Settembrini [00:56:45]:
Yeah, yeah. And so if any listeners want to learn more about, connect with you, or help out, even participate, like what we're out there at.

Peter Coffee [00:56:55]:
I'm sorry, it's a long domain to type, but www.intelligentlifeonearth.com will take you to where I'm writing and sharing. And we don't solicit outside funding at this time. The paperwork is brutal when you start to do that. Right now, this is all self funded, but we are happy to speak with people. I am still doing keynotes for our community. I'm still writing and speaking and I'm not disappearing and I'm not detaching from the ecosystem of which you and I are a part, because this ecosystem has to be an active support facility for making this stuff happen.

Nathan Settembrini [00:57:38]:
Yeah, sure. Awesome.

Peter Coffee [00:57:41]:
Thank you for the chance to share.

Nathan Settembrini [00:57:42]:
Yeah. Thank you for joining and for all the data sleuths out there. Keep searching and you'll find some answers.

Peter Coffee [00:57:51]:
Oh, and I'm going to share one last comment on data. I heard this the other day from someone in the Federal Reserve System. He said, you got to be a data dog if something's on the floor, you taste it, you sniff it, and maybe even you chew it up because it might be food. And you gotta have that attitude toward data. One of the funny things about the word data, you might think that its root is some kind of idea, like a fact or an event. The root of the word data is a thing given and isn't that interesting until I have collected it, filtered out the junk. Yeah, until I've given it to somebody, it has not become data. It may be truth, but until I've given it to someone who's gotten value from arguably doesn't actually earn the label of data. And so asking that question all the time is I don't want to give someone a dump of fact and claim that I provided them with data. In the military, they say it's not intelligence until it's been focused and presented to a decision maker in a way that makes it easy to make a good decision and helping your facts about the world become truths that you validated, become data that you've shared, become intelligence. Because you focused that sharing is an evolution that's worth some effort to achieve.

Nathan Settembrini [00:59:25]:
Amazing, Peter. Thank you so much. This has been fantastic.

Peter Coffee [00:59:30]:
I appreciate the time. Thank you. You've been generous with your time and I appreciate the chance to share. Thank you.

Nathan Settembrini [00:59:35]:
Awesome. Thank you. That's the PO.