(lively music) >> Announcer: It's the CUBE, covering SAPPHIRE NOW 2017, brought to you by SAP cloud platform and HANA Enterprise Cloud. >> Welcome back, everybody. Jeff Frick here with the CUBE with our ongoing coverage of SAP SAPPHIRE 2017 down in Orlando. Really exciting day today, day two, 'cause we got to see Hasso Plattner. Got up and gave his keynote. Joined by George Gilbert. George, great to see you. I know you've known Hasso for years and years and years. Impressions of the kfeynote. God, there is so much stuff that we can dig into. I'm looking forward to it. >> Hasso almost never disappoints, 'cause he's just got %a richness of history and of vision that goes all the way back to the beginning. He was probably the technical visionary from the very beginning. He was the guy who took them from the first super integrated mainframe ERP package all the way to the client server age with R3, and now beyond into sort of in-memory, cloud ready, and with machine learning and iOT baked in. >> But he really speaks like a developer. You can really tell that he likes the technology, he understands the technology, he's kind of a no-BS guy. Some of the Q&A afterwards, people were trying to trip him up and challenge him on stuff. And he would either say, "I don't know," or, "I don't believe that," or, "Here's our impression." Really you could tell he's a humble guy, smart guy, and really has a grasp of what the heck is going on here. Let's jump into it. So many themes we could talk about. But the one that started out early in the conversation was, he literally said, "We need to get as quickly "to the cloud as possible." This is coming from a guy who built the company based on on prem ERP heavy lifting. And even he said today, 2017, "We need to get to the cloud as quickly as possible." >> I think there are a few things going on behind there, when you unpack it. One is, they did start building for the cloud in the early 2000's. It was meant to be a product for the mid-market. In fact, actually its first objective wasn't to be cloud-ready. The first objective was to be highly configurable so that you could bend it to the needs of many customers without customizing it, because typically with the customizations, it made it very difficult to upgrade. In making it configurable first and cloud-read second, they kind of accomplished neither. But they learned a lot. So they started on this next version, which was, okay, we're going to take an in-memory database which we're building from the ground up, 'cause Oracle wasn't building it at the time, and then we're going to build SAP ERP from scratch on top of this new database, 'cause database was so high performance that they didn't have to sepyarate analytics from transactions the way traditionally you do, you had to do in all applications. So they could simplify the app. Then, in simplifying it, they could make it easier to run in the cloud. And now, just like Oracle, just like Microsoft, they now build cloud first and on-prem second, because by building it cloud first, it sort of simplifies the assumptions that you have to make. >> Right, and he talked quite a bit about so much effort now is around integration connectors, to get stuff in and out of this thing. And that's a big focus, he said. It's not that we're ignoring it, it's just a big, hard, hairy problem that we're attacking. >> Yeah, and this is interesting and there's a lot of history behind this. Oracle, in the 90s, up until about the late 90s, their greatest success was in their industry-specific applications, where they took different modules from different vendors and stitched them together. That was how they built, like, a special solution for a consumer package goods company. But it turned out that that wasn't really workable because the different modules for the different vendors6 upgraded at different rates. So there was no way coherently to integrate them and tie them together. And SAP had said that all along. They were, like, this wasn't going to work. Fast forward to the last five-plus years, SAP started buying products from a bunch of different vendors, Ariba, SuccessFactors, Concur, Hybris. So you're, like, "Aren't they doing the same thing "Oracle did 10 year, 15 years before?" But no, and this is what Hasso was talking about today, which was, once those apps are in the cloud, you only have to build the integration points once. It's not like when it's on every customer's data center, you have to build integrations that work for every version that every customer has. So I think that's what he was talking about. You put it all in the cloud, you integrate it once. >> Another thing that he talked, he really, he spoke in tweets. (mumbles) goes to buy Twitter feed, I was basically, like, bang, bang, bang as he was talking. He talked about databases, and databases in the cloud. Nobody cares, right? It's a classic theme we hear over and over. "We presume it works. "We just want it to work." You know, it should just work. Nobody really cares what the underlying database is. >> But he was, in those cases, referring to these purchased apps, Concur, SuccessFactors, Ariba, Hybris. He was, like, "Some of them work on SQLServer, "some of 'em work on Oracle. "But you know what? "Until we get around to upgrading them to HANA, "it doesn't matter because you, the customer, "don't know that." If they were on prem and you had to support all those different databases, it might be a different story. But he's, like, "We'd rather give you the functionality "that's baked into them now "and get around to upgrading the databases later." >> Another thing that came up, and he actually reference the conversation with Michael Dell from yesterday's keynote, about the evolution of compute horsepower. You know, you had CPUs and CPUs kind of topped out. Then you had multicore CPUs. Now we have GPUs that he said you can put 10s or 100s of 1,000s on the board at one time. Basically he's smart guy, he's down the road a few steps from delivering today's product, saying that, you know, we're basically living in a era of unlimited free compute and kind of asymptotically approaching. But that's where we are. And how does that really change the way that we look now at new application development. I thought that was a pretty interesting thing. >> And sort of big advances in software architecture come from when you have a big change in the relative cost of compute memory, network storage. So as you were saying, cost of compute is approaching zero. But the same time, the cost of memory relative to storage is coming way down. So not only do you have these really beefy clusters with lots of compute, but you also have lots of memory. He was talking about something like putting 16 terabytes of memory in a server and putting 64 servers in a cluster, and all of a sudden, I can't do that math, being that I was a humanities major, but all of a sudden, you're talking about huge databases where you can crunch through this stuff very, very fast because it's all, you have lots of processors running in parallel and you have lots of memory. >> It's pretty interesting. He made an interesting statement. He used a sailor reference. He said, "You know, we are through the big waves "and now we're in the smooth water," and really saying that all this heavy lifting and now that this cloud architecture is here and we have this phenomenal compute and store technology, that he can kind of take a breath and really refresh a look out into the future as to, how do we build modern apps that have intelligence with basically unlimited resources, and how does that change the way that we go forward? I thought that was an interesting point of view, especially 'cause he has been at it for decades. >> You know, I think he was probably looking back to some of the arrows he had in his back from having done an in-memory database essentially before anyone else did for mission critical apps. I think when he's saying we're out of the choppy water and into the smooth water, because we now have the hardware that lets us run essentially these very resource-intensive databases and the apps on 'em, so that we no longer have to worry, are we overtaxing the infrastructure? Is it too expensive to outfit the hardware for a customer? So his, when he talks about rethinking the apps, he, like, "We don't have to have separate analytical systems "from the transaction systems. "And not only that. "We can simplify because we don't have to have" what he's calling aggregates. In other words, we don't have to, we don't, let's say, take an order and all the line items in an order, and then pre-aggregate all the orders. It's, like, we do that on the fly. And that simplifies things a lot. Then, not only that. Because we have all this memory, we can do, like, machine learning very inexpensively. >> A whole another chapter in his keynote was about modern software design. A lot of really interesting things, especially in the context of SAP, which was a big, monolithic application, hard to learn, hard to understand, hard to manage. I remember a start, that were were (mumbles) using is a core V to C commerce engine. And to add 16 colors of shirts times 10 neck sizes and 10 sleeve sizes was just a nightmare. You're not going to have some merchant that works at Macy's to put that into the system. But he talked about intelligent design, which is pretty interesting. We're hearing that more and more in a lot of work done over at Stanford, intelligent design. He's talking about no manuals. He's, like, "If I can't figure it out, "I need to understand." He talked about intelligent applications that continue to learn as the applications get more data. And specifically, the fact that machines don't get bored testing 100s or 1,000s or even millions of scenarios and grinding through those things to get the intelligence to start to learn about what's going on. So a very different kind of an application, both development, delivery approach, than what we think of historically as R3. >> Yeah, like the design thinking was, they have this new UI called Fiori. I mean, if you go back 10, 15 years, let's say, when they started, 15 years, when they started trying to put browser-based user interfaces on what was a client server system, they had 10s and 10s of 1,000s of forms-based screens. They had to convert them one by one to work in a browser. I think what he's saying now is, they can mock up these prototypes in a simple tool and they can essentially recreate the UI. It's not going to be the exact same forms, but they can recreate the UI to the entire system so that it's much more accessible. On the machine learning front, he was talking about one example was, like, matching up invoices that you going to have to pay. So that you going to train the system with all these invoices. It learns how to essentially do the OCR, recognize the text. And it gets smarter to the point where it can do 95% of it without-- >> Human interaction. >> Yeah, human inter-. >> You know, it's interesting, we were at Service Now last week, as well. And they are using AI to do relatively mundane tasks that people don't want to do, that machines are good at, things like categorization and assignment and things that are relatively straightforward processes but very time-consuming and again, if you can get to a 70% solution, 80% solution, 90% solution, to free people up to do other things on the stuff that's relatively routine. Right, if the invoice matches the anticipated bill in the system, pay it. Does somebody really have to look at it? So I thought that was really interesting. Something I want to dig in with you, he talked a lot about data, where the data lives, data gravity. He even said that he fought against data warehousing in the 90s and lost. A lot of real passionate conversation about where is data and how should apps interact with data, and he's really against this data replication and a data lake and moving this stuff all around, but having it kind of central. Want to just get your thoughts on that history. What do you think he means now, and where's that going? >> It's a great question. There's a lot of history behind that. Not everyone would remember, but there was an article in Fortune Magazine in the late 90s, where it described him getting up in a small conference of software CEOs, enterprise software CEOs, and he said basically, "We're going to grind you into dust, "because everything comes in our system integrated. "And if you leave it up to the customer "to try and stitch all this stuff together, "it's going to be a nightmare." And that was back when everyone was thinking, "One company can't do it all." And the reality was, that was the point in time where we really had given go past go, collect $200, to every best-of-breed little software vendor. It did prove out over the next decade that the fewer integration points there were, that it meant much lower cost of ownership for the customer. Not only lower cost of ownership, but better business process integration, 'cause you had the (mumbles) integration. I bring this up because, well, actually, I was there when he said it. (laughs) But I bring it up because he's essentially saying the same thing now, which is, "We'll put all the machine learning technology, "the building blocks, in SAP. "If you need any contextual data, "bring it into our system. "You don't want to take our data out "and put it into all these other machine learning programs "'cause there's security issues, "there's, again, the breakdown "in the business process integration." He did acknowledge that with data warehouses, if you have 100s of other sources, yes, you may need a external data warehouse. But I think that he's going to find with machine learning the greatest value with the data that you use in machine learning is when you're always adding richer and richer contextual data. That contextual data means you're getting it from other sources. I don't think he's going to win this battle in terms of keeping most of it within SAP. >> It kind of bring up this other intersection that he talked about. In now delivering SAP as a cloud application, he said, "Now we have to learn how to run our application, "not our customers," a very different way of looking at the world. The other thing that piggybacks off of what you just said is, we've seen this trend towards configuration, not customization. It used to be probably, back in the days, if you had the big SI's, they loved customization, 'cause it's a huge project, multi-years. I used to talk to one of our center partners, like, "How do you manage a multi-year SAP project "when most the people that started it "probably aren't even there the day you finish it?" But he had a specific quote I wanted to call out now, what you just said, is that he said, "Only our customers have the data, "the desire, and the domain knowledge "to make the most out of it." So it's a really interesting recognition that yes, you want customers to have this configuration option. But we keep hearing more and more, it's config, not-- >> Both: Customization. >> For upgrades and all these other things, which now when you go to a cloud-based application, that becomes significant. You don't want customizations, 'cause that's just complicates everything. >> You can't. I don't know if he said this today. I guess he must have said it today. But basically, when you're in the cloud, I forgot the terminology for the different instances. But when you're in, like, the SAP cloud, you can only configure. There's essentially a set of greater constraints on you. When you go to the other end of the spectrum, let's say you run it in your own data center, you can customize it. But when you're running it, essentially sharing the infrastructure, you're constrained. You're much more constrained. And they build it for that environment first. >> Right. But at the same time, they've got the data. Again, this has come up with other SAS companies that we've talked to, is hopefully, their out of the box business process covers 90% of the basics. I think there's been a realization on the business analyst side that we think we're special, but really most of the time, order to cash is order to cash. So if you got to tweak your own internal process to match best-of-breed, do it. You're much better off than trying to shape that computing system to fill your little corner cases. >> It's funny that you mention that, because what happened in the 90s was that by far the biggest influencers in the purchase decision and the overall lifecycle of the app were the big system integrators. They could typically collect $10 in implementation and change management fees for every dollar of license that went to the software vendors. So they had a huge incentive to tell the customer, "Well, you really should customize this "around your particular needs," because they made all the money off that. >> Right, right. Another huge theme. Again, it was such a great keynote. We watch a lot of keynotes, and I have a very high bar for what I consider a great keynote. This was a great keynote by a smart guy who knows his stuff and got history. But another theme was just really about AI. He talked a little bit, which I thought was great. Nobody talks about the fact that airplanes have been flying themselves for a very long time. So it is coming. I think he even said, maybe this is the age of AI. But there always have to be some humans involved. It's not a complete hand-over of control. But it is coming, and it's coming very, very quickly. >> I actually thought that they were a little further behind than might expected, considering that it's been years now that people in software have seen this coming. But they have in the dozens of applications or functions right now that are machine learning enabled. But if you look out at their roadmap, where they get to predictive accounting, customer behavior segmentation, profile completeness for in sales, solution recommenders, model training infrastructure for the base software foundation, they have a pretty rich roadmap. But I guess I would have thought it'd be a little farther along. But then Oracle isn't really any farther along. (mumbles) has done some work for HR. For whatever reason, I think that enterprise application vendors, I think they found this challenging for two reasons. On the technical side, machine learning is very different from the traditional analytics they did, which was really essentially OLAP, you know, business intelligence. This requires the data scientists and the white lab coats and instead of backward-looking business intelligence this forward-looking predictive analytics. The other thing is, I think you sell this stuff differently, which is, when it was business intelligence, you're basically selling reporting on what happened to department heads or function leaders, whereas when you're selling predictive capabilities, it's a little more transformative and you're not selling efficiency, which is what these applications have always, that's been their value preposition. You're selling transformational outcomes, which is a different sort of selling motion. >> It's funny, I heard a funny quote the other day. We used to look backwards for the sample of the data. (laughs thinly) Now we're in real time with-- >> Both: All the data. >> Very different situation-- >> And forward-looking. >> And forward-looking as well, with the predictive. >> That's a great quote, yeah. >> Again, he touched on so many things. But one of the things he brought up is Tesla. He actually said he has two Teslas, or he has a second Tesla. And there was question and answer afterwards really about the Tesla, not as the technology platform. And he poked fun at Germans. He said Germans have problems with simplicity. He referenced, I presume, a Mercedes or a Porsche, you know, the perfectly ergonomically placed buttons and switches. He goes, "You sit in a Tesla "and it just all comes up on the touch screen. "And if you want to do an update overnight, "they update your software, "and now you have the newer version of the car," versus the Mercedes, where it takes 'em three years to redesign the buttons and switches. I thought that was interesting. Then one of the Q&A people said, "But what about the buying experience? "If you (mumbles) ever bought a Tesla, "it's a very different experience "than buying a car." How does that really apply to selling software? It was pretty interesting. He said we're not there yet. But he has clearly grasped on, it's a new world and it's a new way to interact with the customers, kind of like his no manuals comment, that Tesla is defining a new way to buy a car, experience a car, upgrade a car. >> Operate it. >> At the same time, he got the crazy mode, fanatical mode, like, ludicrous mode, so that he could stop and tell the Porsche guys that you're falling behind further every single day. So I thought, really interesting, bringing that kind of consumer play and kind of a cutting edge automotive example into what was historically pretty stodgy enterprise software space. >> You know, it's funny, I listened when you're saying that. That was almost like the day one objective from SalesForce, which was, we want an enterprise app like Sebol, but we want an eBay-like, or Yahoo-like experience. And that did change the experience for buying it and for operating it. I think that was almost 20 years ago, where that was Marc Benioff's objective and he's saying it's easier to do that for CRM, but it's now time to bring that to ERP. >> The other thing he brought in which I was happy, being a Bay Area resident, is the Sharks. Because he's a part owner of San Josey Sharks, obviously it's SAP Center now, also known as the Shark Tank. It used to be owned by another technology company. But he made just a funny thing. "I like hockey, so I should like SAP," and he was talking about the analysis of how often the logos come up on the telecast et cetera. But the thing that struck me is, he said the analysis is actually now faster than the game. Pretty interesting way to think about this data in flow, in that the analysis coming out of the game that feeds Vegas, it feeds all these stat lines, it feeds fantasy, it feeds all this stuff, it feeds the advertising purchase and the ROI on my logo, is it in the corner, is it on the ice, is it in the middle, is actually moving faster than the hockey game. And hockey is a pretty fast game. Very different world in which we live, even on the mar-tech side. >> That was an example of one of the machine learning-type apps, because I think in their case, they were using, I think, Google image recognition technology to parse out essentially all the logos and see what type of impact your brand made relative to your purchase. >> I mean, I could go on and on. I've so many notes. Again, I live tweeted a lot of it, you know, he's just such a humble guy. He's a smart guy. He comes at it with a technology background, but he said we're a little bit slower than we'd like, he talked about some things taking longer than he thought they would. But he also now sees around the corner, that we are very quickly going to be in this age of infinite compute, and we are already in an age of, no one's reading manuals. Just seemed very kind of customer-centric, we're no longer the super-smart Germans that, "We'll do it our way or the highway, "and you will adapt your process to us," but really customer-centric point of view, design thinking, talked about sharing their roadmap as far out in advance as possible. I think he specifically, when he got questioned on design thinking, he's, like, "You know, the studies show that a collaborative effort "yields better results. "It's no longer, 'We're the smartest guy in the room "'and we're going to do it this way "'and you're going to adapt.'" So really progressive. >> And he talked about, with Concur, he said their UI is so easy that you really don't need a manual. In fact, if you do, you failed. And I think what he's trying to say is, we're going to take that iterative prototyping capability agile development and extend it to the rest of the ERP family. With their Fiori UI and the tools that build those screens that it'll make that possible. >> You've handled CAP. We don't spend enough investment on design in UI, 'cause it is such an important piece of the puzzle. But George, we're running out of time here. I want to give you the last word. You've been paying attention to SAP for a very long time. Hasso's terrific, but then Hasso gets off the stage and he said, "I don't run the company any more. "I only make recommendations." As you look at SAP, and Bill McDermott was yesterday, are they changing? Are they just stuck in an innovator's dilemma because they just make so much money on their historical business? Or are they really changing? What's your take as they develop, where they are now, and what do you see going forward for SAP? >> Well it's a really good question. I would say, I look at the value of the business processes that they are either augmenting or automating. I hesitate to say automate because, as he said, you still want the pilot in the cockpit. >> Jeff: In proximity to take control. >> Right. And he was, like, "Look, when we do the invoice matching, "it's not like we're going to get 100% right. "We're going to get it," I think he was saying, like, in the labs right now it's, like, 94% right. So we're going to make you more productive, we're not going to eliminate that job. But when you're doing invoice matching, that's not a super high value business process. If you're doing something where you're predicting churn and making a next best offer to a customer, that's a higher value process. Or if you have a multi-touchpoint commerce solution where you can track the customer, whether it's mobile, whether he's coming via chat, whether he's in the store, and you're able to see his history or her history and what's most appropriate given their context at any one moment, that's higher value. And then it's super high value to be able to take that back upstream towards, "Okay, here's where the inventory is. "I have some in this store. "I can't fulfill that clothing item directly from the store, "but I can fulfill it from this one," or, "I have it in another warehouse," when you have that level of awareness and integration, that's high value. >> Yeah, but I want to push back a little bit on you, George, 'cause I do think the invoice ma-, if he can automatically match 94% of the invoices, that is tremendous value. I just think it's so creative when you apply this machine learning to tasks that feel relatively mundane. But if you're speeding your cash flow along, if you get 94% of your invoices done one day faster and you're a multimillion dollar business, what is the direct dollar impact on the bottom line, like, immediately? It's huge. And then you can iterate and move into other processes. I think what's termed a low value transaction is actually a lot higher value than people give it credit. It's just like again, another one we hear about all the time, automation of password reset. Some of these service desks, password reset, I heard a stat, and one of them was 70% of the calls are password reset. So if you could automate password reset, sounds kind of silly and mundane, oh my gosh, it's like 70% of your calls. It's humongous. >> I hear what you're saying. Let me give you another counter example, which was, I think he brought this up. I don't know if it was today or when Michael Dell spoke, which was that Dell's revolution wasn't that they were more efficient than doing what Compaq did. It's that they had a different business model, which was specifically, they got paid before they even procured or assembled the components. >> Or paid for them, right? >> George: Yes, yes. >> They had no inventory carry costs. >> In fact, that meant their working capital, their working capital needs were negative. In fact, the bigger they got, the more money they collected before they had to spend it. That's a different business model. That wasn't automating the invoice matching. That was, we have such good systems that we don't even have to pay for them and then assemble the stuff until after the customer gave us their credit card. >> Right, right, right. >> I think those are the things that new types of applications can make possible. >> Right. Well, we see it time and time again. It's all about scale, it's all about finding inefficiencies, and there's a lot more inefficiencies around than people give credit, as Uber showed with a lot of cars that sit in driveways and Amazon and the public clouds are showing with a lot of inefficient, not used utilization and private data centers. So the themes go on and on, and they're pretty universal. So, exciting keynote. Any last comment before we sign off for today? >> I guess we want to take a close look at Oracle next and see how their roadmap looks like in terms of applying these new technologies, iOT, machine learning, block chain. Because all of these can remake how you build a business. >> All right, that's George Gilbert from Wikibon. I'm Jeff Frick from the CUBE. We are covering ongoing coverage of SAP SAPPHIRE 2017. Thanks for watching, we'll be back with more after this short break. Thanks. (lively music)