Transcript: Moving Fast Inside a Giant: Rami Segal on Product, AI, and Agentforce at Salesforce

Eyal David: [00:03] A directive comes down saying, okay, we need to start thinking in terms of agents. Come, analyze your use cases across the whole company — in the worlds of sales, marketing, service, analytics. And start thinking about how we integrate an agent experience — where it adds value, where it's relevant — and we're right before launch. Okay? Like, the final stretch, the last version, the release is already locked. You know, it was a blitz. We just locked ourselves in a room like startup people know how to do. What problem can we solve the fastest? Is it campaign optimization? What is campaign optimization? Maybe it's moving budget to campaigns that work? Shutting down campaigns that don't work? How do we identify campaigns that don't work? Campaigns for... that's how I'm kind of running you through it fast. Hi, this is Eyal. This time my guest is Rami Segal from Salesforce. Rami is a veteran, smart product person, and he came for an open conversation about Salesforce, a giant platform that keeps moving fast and innovating, especially in the current period. Rami shares from the experience he's accumulated as a product leader at the company over the last six years, with really fascinating stories, I really enjoyed it, and I hope you will too. Let's go, we're starting. Hey Rami, how's it going? Hey, what's up? Great. With us is Rami Segal, who comes from Salesforce. Tell us a bit about yourself — how you came to Salesforce. Where does it start? Okay, so, this is a bit — it's interesting, but my background — I started 25 years ago, in academic studies of sociology and political science, that's what interested me, that's what I went to study. And we're talking about 25 years ago — that's roughly when Salesforce was born. So you also saw mobile, right? That's the revolution that happened 25 years ago. That came a bit later; the internet came earlier. Right, and then... then it began... I saw several revolutions, and this internet part — you can't escape it, it's like today telling someone entering the tech world that they don't understand what an LLM is. That was exactly the topic — so everyone got into the worlds of development, frontend development, it's an amazing interface, you reach everywhere, there's internet, anyone in the world can access it, and it was very very attractive, so I got in from the technological side — the backend, developing databases, Oracle, Java, frontend development, and that's it. Who's there — it's a whole saga that... now, let me take us along the timeline, so basically... Pass forward? Fast forward — five years in the ERP world, I worked as a consultant implementing ERP and CRM systems, I moved to London with my wife, we came back to Israel following all kinds of personal things, meaning, we decided we want to build a family here and not abroad. Was mobile already around at that stage? It was... mobile was beginning. Mobile began, games began, Nokia, very very fast, Android, iPhone, app development, just wildfire, everyone's developing apps. What made money were the worlds of advertising and the worlds of games, and I got into that field for ten years — mobile game development, and then... the first buds of machine learning arrived — like, let's now build software with statistical models, okay? I identify a lot of language, right? Prediction, natural language processing, I somehow rolled into the worlds of chatbots — chatbots and natural language processing, at a company that doesn't actually specialize in it, at Outbrain, which set up an innovation division and decided to build a chatbot based on Messenger, meaning, Meta was the OpenAI of those times. 2016, Zuckerberg announces that Messenger is going to be the next interface, that it'll hold a lot of things, right? And everyone wants and builds chatbots, okay? So Outbrain too built a news chatbot, that Zuckerberg even announces at the launch, and from there I moved a bit to Lemonade, which was also the first business whose whole business, to this day, is founded on a conversational interface, much deeper, much more complex, both on the marketing side and the operations side, and long before we had OpenAI, they did amazing things, to this day, and fast forward I rolled into Bonobo, whose platform I had already used at Outbrain, as chatbot analytics, and at that time they built a system that knows how to analyze sales calls — meaning, to process calls, to do the full transcription, speech to text, take phonemes — it was Sisyphean work, hard, inaccurate, but in the end it worked, it took speech and turned it into text, and it knew how to identify in the text, keywords and sentiments and sentences, and to take context, like the context of sales calls or customer service calls — which is what we did — and find the insights in them. Is that like something important to Salesforce? So what happened then? Salesforce's development center in Israel, an innovation center that specialized specifically in the worlds of sales, more in lead scoring, in opportunity scoring — all kinds of evaluation, essentially, of leads and opportunities, forecasting for the sales world. Let's take a breath and explain for a second what Salesforce actually does, and the customers, an example like that. Okay, Salesforce is a software company, I think it's the fifth largest in the world, 70,000 employees, its stock symbol is CRM, customer relationship management. What it does is develop a suite of solutions for the worlds of marketing, sales, and customer service — Sales Cloud, Marketing Cloud, and Service Cloud at its core. Why do you need it? Because every business in the world that manages a sales cycle, or any business, B2C or B2B, whatever it is, needs operational systems to manage all its accounts, its sales cycles, its customer service operation, its marketing system, that brings customers into the system — and Salesforce handled this end to end. At the time, its innovation was that it was the first to build this platform in the cloud, when they built the cloud, and it was a breakthrough. The second thing — it was a first mover. At the time it was founded, in 2000, it was essentially the first to build these systems in the cloud, and since then it has developed phenomenally, meaning, hundreds of thousands of customers — the software company that has acquired the most companies in the world, in terms of acquisitions, that was its main growth engine. Among them, you can mention Slack, the largest acquisition in software history, I think, Tableau, and Bonobo. Back then, Salesforce probably identified some need that you came and solved, right? At a time when there's still no AI out there, and what did it look like? Yes, what they did in the worlds of sales, which is really the core — the guy who founded Salesforce was a salesperson in his past. Sales is a lot of conversations, and a lot of emotion, but it's mainly based on conversations, with tons and tons of people we call, essentially, there are the influencers, the decision makers — a lot of people are involved in the process. Closing the sale is closed through all kinds of conversations, all kinds of signals, that come as a result of many many touch points during the sale, and these things couldn't be captured — it was like some unknown, before they digitized it, so to speak, where someone had to take these conversations, it's not just sending an email here and an email there, it's phone calls — a phone call is the level where the SDR does qualification for you, whoever does qualification for you, the salesperson whose job it is — how he asks you, what he asks you, how he brings you to the table, and then the demo, which is what every SaaS company does today. Okay? So capturing this data, and not losing it, and out of it, like that, within the whole deal, putting it into a system that knows how to find patterns — that's exactly the... They identified that you compete on the bottom line, which makes sense for the whole business. Now I'll ask, where was Gong all along? Like any good business — Israelis smell an opportunity — there were many companies in this space, including Gong, Chorus AI, which was also sold at the time, and other companies. To this day that's Gong's challenge, like, what does Salesforce do, right? It's basically the thing. Salesforce invests in Gong? Okay. So they're covered on that front, yes. Great. So essentially, at this point in time, you were about to be acquired, right? We jumped there for a second. Yes. Meaning, it's not... you know, it's... in the DNA of a startup, it's first of all to bring value to customers, to crack the technology, and after that, it's scale, right? And in the end to find the right opportunity, whether it's to merge, whether it's to be acquired, whether it's to IPO — depending on what's happening in the market. In this case, there was a center in Israel that was really looking for exactly this kind of technology, to complete the portfolio of solutions for the sales world, in the field of AI. How complex was it? How complex, actually, was the transition — beyond managing the acquisition and everything — you told me in our prep call that it wasn't too complex a process, like, to integrate that. How was it? It was a fast process — we were actually the acquisition that did the fastest integration in Salesforce's history, but that's no great feat, since we were very small. And we basically built all our technology on the core — we built it inside the code of Salesforce, inside that sales cloud, we adapted the insights we produce to the context of the sales world — something like six months, and then from there we went straight to market. And you're working at Salesforce during this time — what year are we talking about? 2019. And basically, how did it look, at that point in time — where was the world in terms of learning this tool? Probably really easy if it's something inside Salesforce, no? So first of all, even inside Salesforce there's tons of internal competition for visibility, there's no problem reaching customers, meaning, it's not like a startup where you really have to beg customers to come in the door. At Salesforce I can walk in tomorrow morning to the whole force — that doesn't mean they're ready to adopt me yet, that's just the beginning. Customers have their own work plan, especially enterprise, they have an annual plan, and there are many solutions, they have many problems to solve in parallel, you're not their highest priority. The fact that you think you're the shiny thing in the locomotive means nothing, there's an order. So, we did work, we learned to work with Salesforce's salespeople to reach customers, we learned how to manage an opportunity, how to manage ourselves, how to create a pipeline for ourselves. So you literally used Salesforce to manage the whole thing. To learn the process. Yes, and you also told me in the prep call — something we kind of skipped over — that you built the whole thing from scratch, right, at that stage? in order to get inside. Yes, we built a lot of things, really... I also came from something much more, at that point in time. Right, yes, we needed to build... the technology we brought and moved, that's something we brought with us. The integration into the system itself — that was our challenge. Meaning, coming in, there are tons and tons of dependencies, it's not something you work on by yourself, it's essentially dependence on the platform, on security, on all kinds of variables inside the system itself, which also have some roadmap of their own, and you're not exactly top of mind for them, but okay, it's something that accompanies us to this day, we'll surely talk about it later — the topic of dependencies. In any case, it's a matter of determination, and how relevant the technology you bring is. Now, you asked a good question, because 2019 was before COVID, okay? And after we did the entry into Salesforce, and went through all the birth pangs, and we already managed to bring a product — a product means you actually have an app in production, and you're already working in versions, you've already entered the system, into routine — suddenly there was COVID, which is the best thing that could have happened to us. That's clear to me. So how much did it accelerate things? Basically, if we talked about priority within the company — suddenly it flew forward? I'll tell you the story. The company is very... there's a very dominant CEO of the company, and the spirit is what drives the whole process, that's what percolates into the management and his leadership, that's what drives the company. When COVID came, everyone's in shock, okay? It's a crisis, everyone's under pressure, in panic — the guy rose with composure, and understands the situation, and understands that the world isn't being created, and you need to keep selling. Besides that, the company donates millions, if not — I don't know how much — because it's on the world's agenda, the company's agenda — it's on the company's agenda to donate a percent of the sales of projects that are social, so here he understood there's a crisis, and he flew planes with equipment and so on to all kinds of places, forget that. Now he's talking about the business, now he says, give me — first of all he thinks, how do I run my business, what am I supposed to do? I'm supposed to work remotely, do sales remotely. I know — I have a platform where people know how to do... calls, sales calls, I have the mechanism that analyzes calls, that helps, you know, understand what's going on and push things forward. So this was one... at the beginning we were called call coaching, and after that, we were called Einstein Conversational Intelligence — you know, there were two things there: one, to train salespeople, and the second, to help push the deal forward, and suddenly you have the best system in the world and you can turn it on in a second. Yes, it was already operational at that point. It was operational, and the guy gets on a call to the whole company and says, I'm setting a target of 1 million Zooms. What does that mean? That we now measure the amount of interactions between salespeople and customers, and the goal is first of all... and first of all, how do I run my company with my systems, and that way I can teach others how to work too — which is also what happens today with Agentforce, by the way. And it was an excellent opportunity. And you took it forward — and did you also notice what I did here, how we essentially connected your problem internally. Just before we continue — click the link to the Product Builder WhatsApp group in the episode description and join the community. There you can continue the discussion, ask questions, and be the first to get recipes about new Tachles episodes. We'll wait for you there — and now, back to the episode. So now we're basically at a point in time where you probably have some success. Because you basically got a lot of priority within the thing you were part of. Take us from there, and let's really also talk about Agentforce, take us forward too. But first let's touch on the size, on Datorama, right? Yes, so look, after I'd been four years in the world of Sales Cloud, and really learned every possible product in Sales Cloud, you also have to learn that each cloud has tons of solutions, it's a variety of solutions that need some integration. So we learned that, and we learned how to launch a version, how to manage a product, how to do versions, how to manage dependencies, and bring to success, meaning, generate some traction, that arrives... You basically had a very very strong crash course of four years, where you also learned the company's methodology, and the business, and how to get something into this platform. Yes, exactly, how you work inside an enterprise — it was amazing, and then there was an opportunity, basically, two years ago, Datorama was acquired before us, in 2018 — so I'll give a bit of background: all the companies Salesforce bought, some were merged inside the platform, and some, if they were big enough, if the customer base was big enough for significant technology, stayed outside — Datorama is one of them. Also the other companies Salesforce bought in the Marketing Cloud worlds, for engagement — which are systems that manage email, email campaigns, SMS, WhatsApp, and so on, and also personalization, which manages the customer experience when they arrive at websites — stayed autonomous, okay? And when I say autonomous, it means they're not in the platform, they're not in the core, so Datorama is one of them, okay? And it was doing great with thousands of customers, a product people love very very much, a very successful product, that even as an external system had all the capabilities — the ability to bring in data, the ability — what we call today a semantic layer — to model data, and visualization capability, which is really everything Tableau has. So think of a system like that, very very focused on the marketing worlds. So I want to ask you a question, actually. You say they're inside the Salesforce suite in 2018, you came in in 2019, four years, '23, everyone knows what happened in '23, there were a lot of revolutions, so another revolution came in — what really, what was the big driver to bring Datorama now into the platform, what actually? What was that CEO thinking then? Yes, it's not just Datorama — basically what's happening in the world is that for advanced AI, the kind we know today, you need data. There's no AI without data — take the data out of the AI, there's nothing, there's no air there. Since that was the situation, the buzz in the market was that Salesforce is a System of Records, meaning it's a warehouse of data, just another warehouse. Okay, if I now want to build AI, I'll take all my data, from all the places, to Snowflake, to Databricks, and there I'll build the whole AI layer, okay? What does that mean? That you become a back-end player, you're not in front. And that immediately raised a red flag for them, following which Salesforce set up something called Data Cloud. Data Cloud is essentially exactly the platform's solution, into which you bring all the data, and the goal was to flip the equation, to say, wait, wait — you're a business managing sales, marketing, and service with us. Leave the whole thing here, huh? Come on, there's nowhere to go, why are you leaving? And they investigated and did all these inquiries into why people take the data out, what data they need, where they need it, how they need to consume it, what data they need to bring into that platform, and what was mainly missing? And they built Data Cloud, which is Salesforce's Snowflake — Salesforce also invests in Snowflake — and they built a system that would essentially allow customers to stay inside Salesforce, and once the data is connected and in the same place, then you can start building AI and Agentforce and all the automations. Maybe let's explain what Datorama is and what information was missing, and how it actually covered that? Yes, so Datorama is a very specific type of information, it's a type of marketing information, of analyzing digital campaigns in the worlds of ad networks — whether it's Meta, Google, TikTok — Datorama has 170 connectors, connected to every possible ad network, and it also knows how to model data of email, WhatsApp, SMS, whatever, and of course you can integrate with a CRM, but that's not its main thing, it's like just another connector, it's fine. But the data itself sits somewhere else, okay? It doesn't natively... the data doesn't actually flow, that's exactly the point. Yes, that's the point. So there was this vision, which is a pretty simple and brilliant vision, yes. I think it's mainly, like, economic, but okay, so basically, now how do you bring something like that down to the ground? Of course, you take Rami. Okay. This, doing a transformation of a product like that, is not a simple task, because it's not a system, it's a system that was built over a decade, okay? With tons and tons of functionality inside, and they simply looked for people who understand Salesforce, they looked for people who are Core Salesforce, and that's how I met this product, and the task was essentially to take Datorama and rebuild it inside the platform. The platform itself was also being built on the fly, meaning Data Cloud was being built — we're talking two years ago. What, the train is moving... you get on the train now, while it's moving. Now, on top of this, Tableau also got a task, to bring Tableau into Salesforce — Tableau Cloud — it's not the Tableau that lives natively inside the system. So they too, we have a dependency on them. Why a dependency? I'll explain. I talked earlier about three areas of data. Bringing the data in — connectors, every place needs the integrations to bring the data in, modeling the data, and querying the data, okay? Say, the logic. Yes. For each of these things, there's now a whole layer inside Salesforce that handles it. You have Data Cloud, which enables — it's a platform through which you bring in the data; there's a semantic layer that's only just starting to be built, in which you enter all the KPIs, the calculated fields, the formulas, all the things through which you measure your business metrics, and there's Tableau, which is the visualization, and then you need to build an app that also brings the unique things of the marketing worlds, like attribution modeling, and things relevant to marketers, and also to build yourself amid... with this puzzle. Now, it's not like these are Lego bricks that are ready and you just assemble. It's much deeper than that, anyone who works in the worlds of software — how does something like that even look? How do you know where to cut into the flesh? Because apparently you went and cut real deep, you scoped versions, like you say. Yes. It's totally crazy. I'll ask a question before you continue — in '23 or '22 the decision was made, that this is what you do. Yes. And up to now, what's the situation now? Like, what happened in that time span? A very fast run, at that kind of scale. Right. First of all, we built the product and launched it. Okay? We're now doing deployment to customers. That's what's happening. Meaning, we have a product live that managed to cover the whole product gap and be relevant enough within the new platform to start migrating customers and bringing in new customers. Which is like test number one. If we add on top of that a few difficulties so it won't be simple — because it'd be boring if it were simple — well, there was a war, and that definitely affected the whole market, just as it affected us. And the second thing is that Salesforce, within this whole world, also underwent a historic shift in terms of pricing. A company that went to the cloud and invented the model called SaaS of seat-based, where you sell per monthly seat, suddenly needs to start thinking in consumption. What is consumption anyway? How will you explain it to your customers now? How do you translate for them what used to be into the new world? Luckily for us it's not, like, unique to Salesforce. It's a global trend. So no matter where your customer goes, they'll encounter it. Okay? They have no choice. That's what's happening. All the bringing in of data and losing of data and visualization, querying it, and after that sending it to the LLM, and building the models. It's all consumption. Okay? So we too, essentially. It's another challenge in the process. How do you do it? Like everything, and like — you need to do what's called a steel thread. What's a backbone? What's a backbone? That I don't move right or left, and with that I go out. No matter what. No matter what. With that we have to go out. Okay? So we had to be very very very tough and not say, ah, maybe do this too. No, you can't. And my job as a product person was, first of all — my team's job was to define this thing, to give ownership to each of the people on my team to take part of the process. It's essentially the journey of the data from these networks into Salesforce, and the whole transformation it needs to go through, which is fascinating in itself. To protect the team from all kinds of external things. And that's it, in the end. We didn't say — actually, how big is your team? It keeps growing. We're six product people, spread between Israel and the United States, and sixty developers. Nice. So that's also a nice piece to play on. Yes. And you're also basically operating against other things happening, other forces. I want to dwell, for a second — we'll come back to the story. I'm opening a parenthesis — I know you can, I can see it. Let's talk, for a second, about the pricing model in CVI. What do you think about where this thing is going? I mean, will it stay in terms of consumption, like you say, because it's a hard and a bit complicated pricing model. So at first, it was very technical. They talked about the type of data and the action you do on the data, and by its complexity, they decided on a rate card — where something costs less because it's less complicated, and there's something that costs more because it's more complicated. For example, just bringing in data, that's no problem. So that's a low rate card, but if I want to take part of the data and run a model on it, even a statistical one, then it's more complicated. If I now need data to do grounding for an LLM, then I need to look at broader data, and send it to the LLM — and you remember, with OpenAI, they played with you with 32-bit and 64-bit. Well, I'll even narrow it down — these are really easy decisions, how much can it be, at the organizational level, these are crazy leaps — how do you explain them? And it's at scale. An organization that processes — that has tons of calls to send, tons of data to process, it's a big scale, and it also depends on the complexity of the output itself, how much, what you want to get. At Bonobo we did call summaries — you want a call summary, which today you produce in a second and a quarter, we had to think about how much data the output would include, so as not to exceed, so it wouldn't be too heavy. Today nobody cares about that, like. So the pricing was like that, it was very technical. Now I think pricing is moving more to outcome-based, meaning — let's say the one who contacts customer service. I don't know how many ping-pongs I'll need with the customer to solve the problem. Maybe complicated, maybe easy. Okay, so what, for how many do I... every time I send to the LLM — so they said, okay, if I solve the problem I can close the case, if I closed the case, then I charge you. Okay, that's a certain model. Yeah, yeah, I'm like — thanks, where's this going, because it sounds really complicated to me, it's already really complicated today. It's not simple, also if you have multi-agents, and it's an agent that needs another agent, it's really complicated. Yes, okay, so you basically came in as Marketing Cloud, and from here you started launching, and now what's actually the big challenge now? Now the challenge is to expand. Meaning, once you set up a framework like this — we set it up only for the worlds of paid. But within the world of Marketing Cloud, as I mentioned at the beginning, there are a few more units that did did exactly what we do, rewrote themselves — the worlds of engagement, which are systems that let you write email briefs and so on, SMS, set up content, campaigns in the field of owned media, and personalization, which also have their own analytics, but they built their analytics on a different framework, which isn't as advanced as ours, and also when you look at the whole customer journey, from awareness to consideration to purchase, to service, and you want to see all the data in one place, and also afterward to do planning to target a segment, based on historical data, you need everything to be on the same framework. So we built a framework that's super generic, super modular, like we know how to build, we had the luxury of building from scratch, planned right, and we're essentially doing expansion — we take this thing and say, we're now building — expanding to be Marketing Cloud Intelligence for all of Marketing Cloud, not just some one niche. And I'll ask, if you look back now really at the process you went through with Datorama, and ran there with a knife between your teeth, that's what I understand between the lines — what really was the thing you took on that was hard and dealt with? Look, first of all, I have to admit I didn't know Datorama's system, and first of all, I took the people who understand. Meaning, the technology people also came, there were excellent technology people there at a very very high level, without whom this thing would have been no match, you understand? We connected Salesforce people and Datorama people together, people who understand marketing data, and people who understand the platform, and we were very very focused on this thing, so first of all, the first thing is to know how to choose the right core team, that would know how to weave the vision, and would also know how to make the right decisions in execution. That's the first thing. What did I take away? That — it's like, it's an amazing software project, it's like, a bit, I went into it, for me, I was always at the frontier — always at the frontier of the internet, at the frontier of mobile, at the frontier of AI, and for me it was not an easy decision, to go and do a traditional project like that, for me, something that's classic software, also a transformation of existing software. I said, wow wow, it seems to me, it's like crafting, it's art, it's simply art, it's the thing the most beautiful thing that was in this project — to really take an amazing system, a good system, with a challenge like how do I bring this value into a world that's existed here already 25 years, and is also undergoing a major transformation — it's really like bringing this value and also fighting for it on the critical path, like you said, and such. Yes, in... in the final stretch, I told you, like, that if we needed to make sure no one disturbs us, then Agentforce was born. It, like, it was born. And when something like that happens, it also... Let's tell it in a word. Agentforce is a layer in Salesforce that handles the topic of agents, naturally. In the past, it was called Einstein, Einstein Copilots or Einstein AI, and once we moved to the new worlds of LLMs, the concept of an agent was coined, and in Salesforce too, by its nature being a low-code no-code platform, a whole system was set up that lets you set up agents, where each agent represents some action, okay? It has actions associated with it, and it has logic and data standing behind it to complete this action. And it can also, of course, get help from additional agents. It's a thing — it's a conversational interface that assists the people who use Salesforce, or the customers of Salesforce's customers, to perform some action — whether it's answering customer service calls, whether it's sales, that's basically the system. And a directive comes down saying, okay, we need to start thinking in terms of agents, come analyze your use cases in the company, in the worlds of sales, marketing, service, analytics, and start thinking about how we integrate an agent experience, where it adds value, where it's relevant. And we're before launch, okay, like, the final stretch, the last version, the release is already locked. You know, it was a blitz — we just locked ourselves in a room like startup people know how to do, and we said, well, what can we do, what can the platform... what problem can we solve, man, fast? Which is campaign optimization — what is campaign optimization? Maybe moving budget from campaigns that don't work, shutting down campaigns that don't work — how do we shake up the campaigns that don't work — that's how I'm kind of running you through it fast. No, that's great. I'll join you on this — you're drawing on the board, you say, wait, okay, this I can do, this I can't do, cool, so let's build a conversation where we give the marketer, by a certain objective — meaning if their campaigns target awareness or lead generation, according to the customer journey, take the group that isn't working, by what, by metrics. Okay, he needs to define metrics by which he measures himself, and once he runs this conversation, we let the agent run on it, and tell him, okay, do you now want me to stop the campaigns that don't work, and move the budget to campaigns that do work? And we said, okay, come on, let's see if it works, and it's just like that. You also made it a mechanism that actually asks the marketer, so that way you basically removed the initial fear, maybe, if you push something like this. By the way, that was one of the first questions — will a marketer adopt the system? So we simply put in human in the loop, like you do when you need to gain trust, and even though the system can work alone, right? But it's exactly like you say, you have to have him believe in it. You have to have him believe, you have to have him press the button so he feels he's doing something, and sometimes you need that eye — obviously you need it, okay? You can't at the start, you know, and it refines itself, meaning, it's ultimately a system that learns, because, you know, it removes things and can add things, but bottom line, bottom line, that we did the... we did this shift and launched with the agent, and it was such an excellent decision. Critical, because the visibility — I go back to the matter of prioritization and how much you get into the organization, basically. Yes, exactly. The people who... it's exactly one of the things that whoever hasn't done this move and hasn't entered an organization with a startup, needs to understand that visibility is very important. You have to be relevant. You have to be relevant and have to keep running to be on the wave, and keep reinventing yourself and innovating and not rest on your laurels. What are you doing today in these worlds? What other use cases are you actually focusing on today? Of agents. Yes. So there's a world that's really really wild, of course — the whole world of content. The world of textual, visual, and video content is amazing. There's basically creation of content and personalization of content for segments, okay? And A/B testing, meaning to do tests there. So that's a direction that's... we're researching it and we already have a collaboration with a company called Typeface, a very large company that knows how to do generative content, meaning personalized ads by segment, for all kinds of products. What, really at the video level? Or how does it look? Wow, stunning? And no human is involved here? How does it look? No, no, of course. It's simply, you know, instead of going to an external studio or external agency, that produces tons of materials for you and you start running them. My wife's a photographer — I know the deal, but... yes. So we have an integration with a company like that, and it's one of the things that most interest us, it came from customers, mainly customers, you know, customers who are big brands, B2C, who want to lower cost, basically. They mainly want to push the process, to reach results faster. Also, of course, to lower cost, but they won't do it because of that — they'll do it because they now have the ability to do multivariate what's called testing, for multiple segments, multiple channels, personalized, because it's possible, because it's possible. Great. Tell me, and at the level of — maybe tell a bit, we also talked about it a little — at the level of Salesforce's vision regarding the new world, how does it look? It's, like you said, you have the matter of dogfooding, dogfooding, where you basically eat it first. Always. Right, so tell about the... how does it look at your place, eating it, how do you see it. Yes, so look, everything, everything we develop, we adopt within the company. And the company is, as you see — even though it's, like, a company that's old in its bones — you say, Salesforce, like, what innovation can it bring, it's like, cutting edge, it's so on top of things, it's amazing. Amazing technology people, visionaries, and all the time, you kind of say, okay, come on, we reached the end, we reached the top, and you discover there's something new, all the time — oh, okay, the next thing, the next thing, the next thing. Now, they don't want — on one hand, look, it's not that the technology is a tool, it's not, it's not — it's a means to an end, so to speak. First and foremost, at the center of Salesforce stands sales. Always, always it's sales. And in terms of market competition and relevance, then you have the technological side that says, look, this is what's happening, this is what's possible, you have a huge research lab that's constantly running experiments and developing and works, and there's the connection to the market — it's not just a connection, it's a connection to the market, but it's a connection of years to the market. Okay, you're living these companies that have been on Salesforce 15-20 years, you understand their business so well, or there are people in the company who understand their business so well, and it's per industry, meaning, you know how automotive works, how finance works, how health works, how manufacturing works. You simply understand it, understand it in depth. So this connection is a tremendous connection. Now, that's what lets you, as a company, look ahead and really put people in the right places. So okay, so let's say we have some thing in the world, AI, agent team, multi-agent team — let's now imagine I can take this capability and let's see how it looks in the worlds of banking. What happens there? There's regulation, there's compliance, there are these challenges, these are the products, these are services, and a think tank sits and says, okay, how can I now adapt my capabilities to this sector? That's more or less how it works, and that's how the company's roadmap is born. I want to challenge a bit, because you're right, like — you summarized it very neatly, it's nice. Banking and worlds that are regulated, then it's really easier to envision them, but if we go for a second to the world of software development, say, which is much tougher. What do you imagine there? What do you imagine there? Look, I see our company — people use Cursor and write code using all kinds of dev tools that help them both write the code, and test the code, also merge it, also do deployment for it — I think I don't see developers disappearing. Like, forget it. I haven't yet seen a business that would go — I'm careful not to throw out empty words, but — to platforms that know how to produce code and apps on their own, and say, what, we'll build my business like that, and I'll let it run, me alone, one person in the world, and if some bug happens then I don't know, someone will fix it. I agree, there's going to be — there's going to be a correlation here, that's clear, right? I don't know, I look, there are certain departments in organizations that probably will cut SDR, so I don't know, but there are ones you can look at and say — but an organization is going to go through some process, which is clearly going to happen, which is that it'll be human alongside machine, right? So how do you imagine it? Yes, right. It's clear there'll be people alongside machines, like, it's funny to call them machines. Right? Right? Yes, but it's, look, what makes it not be, like, a machine, is the conversational interface. Everything begins and ends with the conversational interface in the end. We talk with some digital entity, that you can give a name and a visual, but that's the key, and by chance, at Salesforce, there's the most dominant Enterprise AI system in the world, for a conversational interface, which is Slack. It started from development — after all, Slack was born from people, from developers, from geeks who simply used Slack for their needs. It kind of reminds me of where Discord is today — it's already my son's age, and I'm totally hooked on it, but that's Slack, okay? Slack is built with a conversational interface, today we can't imagine ourselves working without Slack, we work most of the day, like we don't have — we don't look, I don't look at email at all. I work only with Slack, both internal communication and communication with salespeople, also communication with customers, with Slack Connect, also automations and workflows, to get information, to send information, you know, all the apps that run on this interface, and naturally, we'll also see there, once it becomes the main interface, okay? for work — let's say, a work OS, meaning, an operating system for the business itself, what Salesforce used to be, of nice tabs and screens and everything, will probably, probably, become a conversational interface, and naturally, it'll be very easy to integrate agents there, because it's simply an interface. I think it'll stay text — how do you imagine it? Yes, you know, it'll stay text, voice — the output will be much richer. Now the war is over the output, over unstructured data, you can see it by the rumblings in the field. Maybe expand on that a bit. So getting textual information is very scalable, it's good up to a certain level, and losing it, and then suddenly they offered summarization, because who has the energy to read all this information? So already now in Tableau we have Concierge, which is an agent that generates visual data, charts on demand, and you'll probably have these interfaces that create the ad displays, which is already worlds of visual and video, and it may be that in the end we'll simply get someone you can talk to, and he talks back to you. That's it — the best is for meetings, right? I went really futuristic here, I don't do this at all in episodes, but you're here, so I allow myself. But meetings, for example — today it's really in a room, so suddenly I'll now have, like there used to be that iPad that rides on... on wheels. I'm really very wrong about these things. Like, I've seen it already happening, because when we did the... at Bonobo the call coaching, we already saw the... you can take sales calls and ask them to simulate a call. Meaning, you'd think that in front of you stands, like, someone that... a figure, 3D, that simulates the customer, he asks you questions, you talk, and there's some script that afterward they check — they check that you said how you handle objections, that you said all the things you need to say in the conversation about the product, like it's already there. It's already totally there. But yeah, what you're saying — what a thing, I think the story is really funny. Go ahead. When they launched this topic of AI and chatbots, and it was under Einstein, then we came to Dreamforce, which is Salesforce's biggest event in the world, and Marc was on stage, he got up, and said, come, I'll show you what the future looks like, a table like the one you imagine, an executives' table, and in the middle there's like an Einstein doll, which is like a designed speaker, really pretty, with a voice output, and he says to it, tell me about our sales pipeline, what can you say about it? Will we hit target, won't we hit target? And it talks to you. Everything's fine. It talks, it takes the data, it looks at the statistical models, without an LLM and without anything, with models of the old days. So I totally connect to what you're saying, meaning, it's also logical, like, the question is, how far will people want to take it? Like, how essential is it? The fact that it's possible isn't... I agree, especially traditional organizations — and come, let's for a minute go to our little angle, how do you imagine the role of product people, really, now in this era of AI and something that'll arrive in the coming years? Look, I'm sure all product people in the world today already use AI for all kinds of needs, we use it, of course, to do research, to write requirements, to create mockups, to do simulations, it very very much pushes the process, by the way. Also to build pages that... you no longer need someone to build you UX and all those things, so okay, from our standpoint, we already know how to work with these tools, all product people. In terms of building products forward, you simply need to take it, take it into account, that this is the interface. It's going to be the interface. A conversational interface, you say? Everywhere, probably, where it'll be possible to push this conversational interface, it'll probably... even if you have agents that are headless, meaning, that do the work in the background, and it's a kind of... kind of automation with reasoning capabilities and decision-making capabilities, that will need to, probably, probably, be integrated into every future product. That's for sure, that's for sure. I think in this aspect, how do you imagine — will there be more product people, will there be fewer product people, in this era? Because if you go and manage the process, yes, you say, well, you need less head count, because I can do things with the help of AI, for instance. On one hand. On the other hand, there are a lot more problems here that we need to solve, and there are things you need to do, but, also here, like, there are a lot of ideas that are just there, and to do validation, that's supposed to be easier. So it's really interesting, where this is going. I'll tell you what I think. As long as people are involved in the decision-making process of purchasing software, or any other product, and people needed to talk with people, then product people will be there, inside the process. Because product people, for a long time now, are no longer product people who sit with their legs crossed and write requirements. I'm telling you that today I'm half a salesperson — I get on almost tons of sales calls, with customers, because someone needs to help, someone needs to do discovery — which use case, whoever has a product, you need to adapt it to your use case — you need to know how to speak this product. We do deployment, we get on and do hands-on deployment, there are tons of things that... okay, so we have, okay, so at Salesforce we have a concept called Forward Deployment Engineer, I think that's what it's called. It's a new invention that says, it's like a solutions engineer, broadly — these are people who know the products, do discovery for the customer, and they're at the site itself, even if needed, and implement it, get it out to the customer. As long as this chain exists, there'll continue to be product people. Probably, as long as you sell to organizations, and probably the product's price tag is high, then it makes sense. Yes. No, because I took it to what you described earlier, from the other angle — you talked about, basically, a marketing system, that basically runs tons of variations, and doesn't stop — so probably here it'll actually be much easier to bring, I don't know if easier, but less human interaction, in terms of marketing, to bring in B2C products, or ones with a low price point, this whole aspect of B2C products, in my opinion — quantity, you also came to the world of, sorry, of gaming, so you can also comment on that, but it seems to me that here a lot is going to drop, basically how much direction of product, and also of UX, and also of this. Like, it seems to me it's going in that kind of direction, in short, I don't know. There's no knowing. Stunning. Okay, Rami, really interesting, a really open conversation — I love hosting episodes like this, usually, usually, I wanted to know, what do you wish for yourself a year from now, in this world you live in. I have to tell you, that I wish for all of us, that there won't be war — that's the only thing I wish for. Everything's small, next to what's happening here, and that there'll be peace and there'll be quiet, and that we'll have fun. Amen, amen and amen. Okay, Rami, I had a lot of fun, so come on, bye for now, bye bye. Bye. Hey, friends, thanks for listening. If you found this podcast valuable, you can subscribe, follow us, of course, for more episodes, on Spotify, Apple Podcasts, or any other app, of course, and if you didn't find us on some app, I'd be glad if you write to us. We'd be very glad — for five stars, on every platform, and follow us, so that more listeners can be exposed to us and find the podcast. You can also find the previous episodes, on every app, or on the YouTube channel, we have links in the description. Until next time, come on, stay efficient, and bye bye.