Salesforce Futures Magazine
Q1 2024
ISSUE 1 Q1 2024 Personal AI
SALESFORCE FUTURES Welcome By Salesforce Futures In 1966, MIT computer scientist life and in fiction, it seems we can’t Joseph Weizenbaum created help but attribute personhood to smart ELIZA, an early chatbot machines. And now, as personal agents are designed to facilitate the perhaps on the cusp of becoming a new study of interactions between paradigm for computing, we still struggle humans and machines. to avoid using human terms to describe them—we speak of assistants, copilots, By today’s standards, the program was sidekicks, companions, and agents. crude—mirroring its responses around The potential of this yet-to-be-named new a single keyword—but the reactions of paradigm excited us at Salesforce Futures users shocked Weizenbaum. Instead as we planned the inaugural issue of our of seeing the shallow state of human- quarterly magazine. Our team’s mission machine interactions as expected, is to help Salesforce’s leaders and our users ascribed both intelligence and customers anticipate, imagine, and shape agency to the program. In other words, the future, and it is in this spirit that we they easily believed that there was a dedicate our whole issue to personal thinking intelligence behind the screen. AI futures. To be clear, we’re focusing Weizenbaum’s own secretary asked to here on the AIs that will permeate our be left alone with the terminal so she personal lives, as distinct from the types could have a more private conversation. of tools we and other enterprise software It’s no accident that the 2013 Spike Jonze companies are building into our platforms. film that provided perhaps the most While technologists in Silicon Valley and sophisticated depiction yet of a personal elsewhere are researching and developing AI was called “Her” and not “It.” In real agents, too few outside relatively narrow ISSUE 1: PERSONAL AI | 2
SALESFORCE FUTURES Welcome circles have a sense of either what it will In subsequent quarterly issues, we’ll take to deliver them or the impact they examine other possible impacts of might have on our lives, both as individuals AI on business, work, and society. and for the businesses that serve us. Oh and back to the yet-to-be-named As personal AI becomes more mainstream, paradigm: in our main article, we use we believe digital intermediaries will the term agents as it’s most widely remake the relationship between used in the technical literature. In customers and companies. How do you the speculative futures section of the advertise your offerings in such a world? magazine, we conjure other possible Will future users instruct their agents to ways of identifying personal AIs. block all forms of sponsored interruption, The Salesforce Futures Magazine is our bar a few trusted brands? Will personal newest way to help you take an organized agents negotiate with company agents like look at plausible and possible futures. those being built by Bland.ai and Parcha? We hope it inspires ideas, questions, and What does a customer relationship conversations, and we look forward to strategy look like in this kind of future? engaging with you on it. Drop us a line at It’s a good time to ask such questions. [email protected]. Imagine if, in 2006, you’d known how the combination of search, mobile apps and The Salesforce Futures team social media would revolutionize how businesses connected to their customers. We may be on the precipice of a similar moment of change. Now is the time to start to understand the contours and implications of this shift, before it arrives. In this first issue, we dive into these new personal AI creations, take a careful look at what needs to happen to make them a reality, and imagine a range of plausible near futures shaped by their presence. The goal is to help you explore personal AI futures so you can start asking better questions, and perhaps even making better decisions, today. ISSUE 1: PERSONAL AI | 3
SALESFORCE FUTURES Contents 6 Orienting 7 Anticipating 27 Imagining Personal AI Personal Agents Personal A brief history Three questions AI Futures and de昀椀nitons about personal Looking out AI futures 3-5 years 8 How does your 30 Will personal AI level agent work? the playing 昀椀eld or widen the gap? 14 Why would you want 38 Who pays for a personal agent? personal AI? 19 Do agents represent 46 Will personal AI come a viable new business from new or existing category? providers? 51 Credits ISSUE 1: PERSONAL AI | 4
SALESFORSALESFORCE FUTURESCE FUTURES Significant moments from science fiction to non-fiction in the development and portrayal of personal AI. 2013 HER Science Explores the complex 1982 relationships between Fiction BLADE RUNNER humans and AI Advanced AI beings are known 2001 2014 2017 1968 1970 as “replicants” ARTIFICIAL INTERSTELLAR BLADE RUNNER 2001: SPACE COLOSSUS INTELLIGENCE Introduces a 2049 ODYSSEY Supercomputer, 1984 Explores the multi-functional Expands on the Arthur C. Colossus, manages THE emotional journey robot, TARS original exploration Clarke’s film US nuclear TERMINATOR of a childlike of AI and what it series featuring defenses an AI defense android in a means to be human the AI character network becomes future society HAL 9000 self–aware 1966 1972 1997 2012 2014 ELIZA PERRY DEEP GOOGLE ALEXA Chatbot Chatbot developed BLUE NOW Amazon developed by Kenneth Colby IBM 2022 by Joseph GPT 4 Weizenbaum 2011 2013 OpenAI SIRI ATLAS Boston 2020 Non-昀椀ction Apple Dynamics GPT 3 OpenAI 2015 CORTANA Microsoft ISSUE 1: PERSONAL AI | 5
SALESFORCE FUTURES “Agents [are] so昀琀ware entities What capable of performing tasks on their own, ultimately in the even service of a goal, rather than simply responding to queries from human users.” is an SILVIO SAVARESE CHIEF SCIENTIST agent? SALESFORCE “Agent” is the term most “Agents are personalized, commonly used in technical persistent, and proactive.” literature to describe functions closely associated with personal PHIL MUI AND ITAI ASSEO AI R&D AI futures, but what does SALESFORCE it actually mean? The term, rooted in economics, refers to the ability of actors to make decisions in specific contexts. “The AI agent can do all the This fits with how we’re same things as an assistant beginning to conceptualize or concierge — but it also our personal AI tools. can perform actions without being asked.” KATHY BAXTER AND Our Salesforce AI colleagues YOAV SCHLESINGER are exploring agents from OFFICE OF ETHICAL multiple dimensions, and their AND HUMANE USE definitions foreshadow themes SALESFORCE we’ll explore in this issue: ISSUE 1: PERSONAL AI | 6
SALESFORSALESFORCE FUTURESCE FUTURES Anticipating Personal Agents Three questions about personal AI futures We think personal AI agents might just change the world, in time. It’s a big claim, we know. To get to that point personal agents will need to meet three criteria for breakthrough innovation — feasibility, desirability, and viability. In other words, they’ll need to be able to technically deliver certain capabilities, those capabilities will need to enable offerings that people want, and it’ll need to be worth someone’s while to deliver that, i.e. there’s a business model. None of those things are certain. In this article, we’ll lay out possible paths from where we are today, which we’ll then selectively bring to life in the speculative futures that follow. With the rapid pace of change in AI today, it’s really hard to identify the key drivers of change, let alone make accurate predictions. Our hope is that the framework we lay out helps readers sense-make their way through the torrent of new developments we’ll see in the coming months and years. ISSUE 1: PERSONAL AI | 7
SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: FEASIBILITY How does your agent work? First, we’ll examine feasibility and the question of how agents actually work, diving into knowing, remembering, reasoning, and doing to help you establish baseline capabilities agents need to deliver real world utility. ISSUE 1: PERSONAL AI | 8 Image: Art collage with the concept of thought process, ingenuity and new creative ideas
1 SALESFORCE FUTURES ANTICIPATING PERSONAL AGENTS: FEASIBILITY KNOWING Unlike earlier assistants —think of today’s Siri or Alexa—agents will know a lot about the people who use them. Users won’t need to state their context and what they’re trying to accomplish at every turn—the agent will already possess sufficient information. Think of the degrees of contextual knowledge required for this to be true. It’s a spectrum that runs from the most basic level—who I am, where I am, what I’m doing, what I’m wearing, etc.—to much deeper, far more nuanced emotional levels referred to as “sentiment analysis”—how I feel, why I feel LILIAN WENG it, etc. For an agent to provide true contextual OPENAI intelligence and pair that intelligence with action, a full range of inputs is needed. Simply put: the more an agent knows about us, the better it will work. But assembling the right data and insights to support the work of an agent operating in real time is a non-trivial challenge. Current LLMs come equipped with knowledge of the baseline data that informed their training. For most powerful LLMs today, training data includes content available on the Internet and proprietary data licensed from original content providers. As impressive as this knowledge is, it still lacks the kind of contextual understanding of the user required to make a world of personal agents a reality. What’s needed is enough contextual information to enable persistent representation; knowledge of what to remember…and what to forget; and the ability to safeguard privacy and trust through all that. ISSUE 1: PERSONAL AI | 9
1 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: FEASIBILITY REMEMBERING For an autonomous agent to be useful, it needs to have memory: the capability to store and retrieve information. Memory in AI is not a single entity, but rather a collection of layers. One layer is the ability for the personal AI to store and retrieve the prompts that the user initiates across multiple sessions. We refer to the ability for a personal AI to “stay” with a user over a longer time horizon (i.e. across multiple interactions) as persistence. The second layer is what’s already trained into the LLM and how the relationships are built between the pieces of data that inform the training. There are limits to this layer such as the cutoff dates for the training data. The third layer of memory is the ability of personal AI to retrieve data (both structured and unstructured) from other data stores by using methods such as RAG (retrieval augmented generation). Currently, context is provided by the user to the LLM, but utility is limited by the LLM’s inability to store the information provided (context length), and by the accuracy with which the user states their preferences. The user’s actual context may also include data the agent has no access to, thereby increasing the likelihood of “off” responses. Conversely, an agent might one day have access to the user’s entire recorded history, which would seemingly offer unlimited context (not to mention near unlimited data storage). But even if that’s the case, how does the agent know which parts of a user’s history are most important, or which parts are more sensitive? The challenges with memory and meaning for LLMs echo the challenges of understanding how humans assign meaning to memory. ISSUE 1: PERSONAL AI | 10
1 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: FEASIBILITY In a recent New York Times REASONING interview, the neuroscientist Charan Ragnarath points out, If you’ve seen Her, “we [humans] have a longer you’ll remember timeline in which we can the OS1 onboarding integrate information than many scene where Joaquin other species.” Humans bring an Phoenix’s character, incredible amount of contextual Theodore Twombly, knowledge to any situation, is guided through a and it’s a significant challenge short series of personal for agents to replicate that. questions to con昀椀gure Innovative companies his AI. are already considering workarounds. For example, Memorably, he’s asked, the ability to curate a personal “How would you describe your AI’s memory may become a relationship with your mother?” critical feature. OpenAI recently Theodore begins a stuttering announced memory control answer, but as he’s just getting features in ChatGPT, explicitly going, the onboarding process giving users control over which surprises him by cutting him off. interactions are remembered The OS has extracted enough and which are forgotten. information to make the design Apple’s new Journal App choices it needs to start off with. prompts users to add rich Over the course of the film, written or audio notes to Theodore’s AI, Samantha, contextual information the continues to learn and adapt. device already has access to, The ability suggested by this such as photos that were taken, onboarding scene to extract or locations where time was meaning and determine spent, or calendar appointments. complex choices from limited This is a potentially powerful information evokes a powerful step towards assembling a machine intelligence with full contextual understanding; sophisticated reasoning backwards-looking to begin capabilities. In reality, how such with but perhaps capable of reasoning might work informing a personal agent in at scale is an open question. the future. ISSUE 1: PERSONAL AI | 11
1 SALESFORCE FUTURES ANTICIPATING PERSONAL AGENTS: FEASIBILITY Consider the core function of the An opposing view suggests LLMs, “planning skills” Lilian Weng and others due to their lack of an internal world refer to when it comes to the agents of model to predict states, may never today. Heuristic search and reinforcement achieve a level of reasoning sufficient learning have limitations in dynamic, to power autonomous agents. real-life environments. For agents to Those who believe this are pursuing truly deliver Samantha-like capabilities, different approaches. For example, they’ll need to move from heuristics to some researchers are developing models inference, reasoning, and metacognition. that train AIs to learn about the world One can argue that the latest LLMs more the way humans do, which may already exhibit emergent behaviors that provide an alternative path towards sometimes look like reasoning. According generalized reasoning and planning. In to a survey by Huang & Chang, a number the future, we expect autonomous agents of past research papers have already made to have more sophisticated reasoning claims for the reasoning capabilities in and problem solving capabilities. some models. Prompting to generate intermediate steps to solve problems (e.g. Daniel Kahneman popularized Chain of Thought and Tree of Thought) a simplified model of how has produced promising results. However, thinking works in the brain, LLMs still struggle with tasks that are easy using the idea of two for humans, such as performing complex different modes of thinking: math and generating action plans to complete tasks in given environments. System 1, “thinking fast”, is intuitive and Engineers are attempting to overcome unconscious. System 2, “thinking slow”, is these limitations by incorporating planners deliberate and rational. This model offers a that break down tasks into steps and useful way of thinking about agentive chaining multiple systems together. capabilities. These systems coordinate so Early tools like BabyAGI and Auto GPT that when a problem is too complex for triggered questions about trust, privacy, System 1, System 2 kicks in with more accuracy, and reliability, but but they sophisticated logic and reasoning. inspired an innovation race to deliver Characteristics of System 1 are analogous to greater autonomy using similar methods. the strengths of current deep learning models, while System 2 maps to the weaknesses of the existing LLMs. ISSUE 1: PERSONAL AI | 12
1 SALESFORCE FUTURES ANTICIPATING PERSONAL AGENTS: FEASIBILITY DOING Research and Insights team recently So far, we’ve outlined how knowledge, argued convincingly for the importance context, memory, persistence, and of keeping a human in the loop with new reasoning might combine to deliver a truly AI innovations— particularly in the short- transformative personal agent. The copilot term to ensure trust. Salesforce’s Office concept is well established, and we see early of Ethical and Humane Use (OEHU) has examples making their way into the market, argued repeatedly for “humans at the especially the enterprise software space. helm” until we’ve cultivated trust and enabled users to learn the technology. Arguably, the most Now, let’s consider some potential future groundbreaking result of all use cases for these action-oriented agents, these capabilities coming how we could oversee their functions, together will be, per Ninja AI, and why, if our predictions prove accurate, “autopilot”, or the ability of they will revolutionize our work lives, our agents to take action and get personal lives, and customer experiences. things done without requiring the user’s attention or even explicit command. Salesforce Chief Scientist Silvio Savarese is pioneering the concept of Large-Action Models (LAMs) which will have the ability to “automate entire processes” by combining “the linguistic fluency of an LLM with the ability to accomplish tasks and make decisions independently.” Savarese believes these changes will elevate AI “from a passive tool…to an active partner in getting work done in real time.” Autonomous autopilots are the source of a lot of interest and fascination, but they’re currently operating at a much more experimental level that underscores the difficulty of jumping from human in the loop to true autonomy. There is good reason to take this transition slowly. Holly Prouty from Salesforce’s cross-cloud ISSUE 1: PERSONAL AI | 13
Salesforce Futures SALESFORCE FUTURES ANTICIPATING PERSONAL AGENTS: DESIRABILITY Why would you want a personal agent? ISSUE 1: PERSONAL AI | 14 Image: African American businessman in glasses with robot face
2 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: DESIRABILITY Last year our team, in collaboration with Salesforce Studios and Salesforce Design, imagined a plausible future of how agents might change the shape of work in Salesforce 2030: A Glimpse of an AI Future. Jordan, the trailblazer hero, interacts with her own “chief of staff” agent and a range of more specialized agents to make faster, better decisions in collaboration with her human team at Koa Coffee. What we chose not to illustrate were the personal agents that might support end-consumers “LLMs will graduate to on the other side of these transactions. becoming agents with The transformative promise of agents lies more powerful tools, in their ability to compensate for human search, APIs, coding limitations. Anyone who has struggled to abilities, clicking finish one task, let alone juggle many of them, on the web, etc. In can understand the value proposition. particular, AI assistants for search are helpful INFINITE INTERNS + PATIENCE enough and will Imagine an army of infinitely patient interns who replace Google stand ready to work on your behalf. These interns for many young can either collaborate alongside you in the people, students and flow of work, or they can labor independently, knowledge workers.” periodically checking in to ensure they’re on the right track. The more they work with you, the smarter they get, learning how you work RICHARD SOCHER and think across modes and contexts. Think YOU.COM CEO about a world where all of us have the kind of expert staff currently enjoyed by CEOs: gifted helpers who learn our preferences, understand our goals, engineer outcomes, and specialize in doing all of the things we don’t want to do. Tech blogger and consultant Venkatesh Rao encourages us to think about machine intelligence as fundamentally different from human intelligence, particularly when it comes to “attention”. ISSUE 1: PERSONAL AI | 15
2 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: DESIRABILITY Our personal agents will have endless patience for tedious, detailed tasks (think taxes, paperwork, applications, and other) of the sort that sap our attention and, occasionally, our desire to endure the human condition. Agents have the potential to remove this burden. PERSONALIZATION Greater contextual intelligence and more persistent memory suggest agents will provide personalization that’s far greater than what we “I use a similar see today. Itai Asseo and Phil Mui, who work de昀椀nition of ‘agent’ on AI R&D for Salesforce, encourage us to think to economists; that about personalization in three categories: is, a narrow AI that “know me”, “inform me”, and “empower me”. acts on behalf of its In the “know me” category, agents keep your principal (or user) goals in mind, analyze your performance, and out of sight, while adjust to your unique style. Because every that principal’s interaction with personal AI will be remembered— attention is on or stored as a state—and factored into future other things.” use cases, a flywheel effect takes hold: the more you use your agent, the better it gets at anticipating your needs in an intuitive way. CHRIS NOESSEL In the “inform me” category, we consider the AUTHOR possibility that agents could use their contextual intelligence to help guide and prioritize our attention and separate signal from noise. This has clear applications in the personal productivity space, but the implications on the consumer side are no less significant. Imagine an agent who helps you switch on “Zen Mode” and other filters so you can better tune your environment, and even remove things from your plate by acting on your behalf. Finally, the “empower me” function speaks to personal AI’s ability to serve as a coach/mentor, or even a manager. ISSUE 1: PERSONAL AI | 16
2 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: DESIRABILITY If personal AI can deliver such control, it will In aggregate, these developments point to forever change the relationship between futures where tools are easier to use and customers and companies by raising the it’s easier than ever to get things done. bar for direct relational engagement. In NATURAL INTERACTIONS ADVANCED SKILLS AND LEARNING Recent demos by startups Humane and 2014, when Amazon launched its original Rabbit have started to tangibly articulate Echo, the device was little more than what agent-based offerings might look like. a bluetooth speaker. Adding the Alexa assistant promised to turn the Echo into Common early use cases include something else entirely: a conversational ordering food, finding gifts for loved smart home hub, particularly as the Alexa ones, planning trips and events, and Voice Service SDK toolset expanded the scheduling appointments. What these library of skills available to consumers. demos hint at is a more fluid and flexible Alas, most Echos are still used mainly collaboration between humans and as speakers and the skills revolution machines, especially when it comes to has not happened. Nevertheless, the attention. Both offerings promise less concept of a core agent and a library of tapping and scrolling and more focus. additional skills offers a preview of what we might see in an agentive world. Both products also rely on conversational interfaces to do this, but we think voice will The ability to train agents on data sets be only one of the ways people interact with (including proprietary ones) means people their agents. Many proto-agents feature can far more easily create useful personal travel in their demos, but solving a multi- AI tools for others based on specialized part travel puzzle that includes schedules, knowledge. This, in turn, empowers people price comparisons, and airline and hotel to build more agents. Already, we can see preferences using only voice commands is Open AI allowing users to rapidly build their less than ideal. Liz Trudeau of Salesforce own GPTs, enhanced with advanced skills Design encourages us to think about a based on additional “instructions, extra more practical alternative, “Think about a knowledge, and any combination of skills.” collaborative interaction, a flexible interface These advancements suggest a forthcoming that adapts to the task at hand and the Cambrian explosion of agent skills and job the user is trying to accomplish.” We capabilities, personalized for an infinite like this concept, “UI on the fly”, because it variety of tasks. We know this would emphasizes a truly responsive interface that transform how companies interact with adjusts accordingly as needs change. You can customers. What we don’t know is which get a glimpse of what this multi-modal future sectors will be transformed first. might feel like in the much-discussed Google Gemini demo depicting the planning of a birthday party. ISSUE 1: PERSONAL AI | 17
2 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: DESIRABILITY UNMET NEEDS A key element of Clayton Christensen’s theory of disruptive innovations is that early disruption will come from offerings that target the unmet needs of customers at the bottom of the market, hitherto uneconomical to serve. Often, these offerings appear inadequate to the mainstream of the already served, but they can grow and improve from tiny seeds. Looking at the emerging agent landscape, we can see an example of this dynamic in the wave of AI companions, such as those provided by Replika, character.ai, and Baidu’s Wantalk. These companions, chatbots, and avatars serve the needs of the lonely, the elderly, and those who cannot a昀昀ord costly therapy. While companions can generate an “ick factor” response among mainstream audiences, observers such as Andreesen Horowitz have identified how the “companion stack” may hold clues as to the future directions of the larger agent space. For example, we may see agents that super- empower individuals, catalyzing a new wave of entrepreneurship in developing countries by serving needs that could never have been met before, and thereby generating a similar impact to that of mobile phones a generation ago. The evolution of agents will be determined by consumer preferences and progress against technical challenges, but market dynamics and the way businesses and consumers balance trade-offs will play an equal role. In the next section, we look at how the business of personal AI might evolve. ISSUE 1: PERSONAL AI | 18
SALESFORCE FUTURES ANTICIPATING PERSONAL AGENTS: VIABILITY Do agents represent a viable new business category? Image: Cascading waves of abstract neural impulses in vivid, neon hues ISSUE 1: PERSONAL AI | 19
3 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: VIABILITY Assuming agents are technically possible and people want them, there still needs to be a compelling business case for companies to bring them to market. Which companies are positioned to win in the emerging marketplace? Who pays for these things? How does the business model for services shift in a world of agents? How can we trust them? Let’s dive into how this emerging market might evolve. WHICH COMPANIES ARE POSITIONED “Form follows TO WIN: INCUMBENTS OR STARTUPS? funding.” Historically, startups have thrived in moments of disruption. However the unique dynamics of STEWART BRAND personal agents—with their reliance on vast data, AUTHOR expensive compute, and the critical currency of trust—may tilt the playing field toward incumbents. Established players can also use entrenched hardware and closed ecosystems for further advantage. In thinking about who might win, it’s helpful to break personal AI out into a few component parts. First comes data, the life blood of any agent. In a 2023 post entitled “The Four Wars of the AI Stack,” Swyx et al. from influential AI newsletter Latent Space point to the “data war” as one of the major theaters of competition shaping the market. So we see LLM creators hammering out licensing deals with content providers, the New York Times’s legal fight against OpenAI and Microsoft, and heightened interest in synthetic data within the AI research community. The next component is the model employed, which is the motor powering a given agent’s decision-making process. We see GPU chips as a part of this category. The third component is user interface (UI), the bridge between AI’s potential and its practical utility. ISSUE 1: PERSONAL AI | 20
3 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: VIABILITY Finally, distribution plays a critical role in competition. Even if consumers love a product and it lives up to its technical promise, a lack of distribution presents significant headwinds when it comes to long-term growth (see the fate of Netscape Navigator). The aggressive acquisition strategies by incumbents point to a winner- take-all dynamic. Incumbents want lock-in an existing user base before upstarts can scale distribution. As author and analyst Sangeet Paul Choudary points out, these companies see a unique opportunity to become the primary interface for AI interactions in both work and personal life. Choudhary also posits that people don’t want an army of agents, they want one agent. If he’s right, the stakes for companies competing to provide such an agent couldn’t be higher. Winning could mean the opportunity to control a choke point on any transactions that flow through a personal agent. Network effects could further entrench winners: the more interactions that happen on your platform, the faster the learning loop gets, the more people use your products, and the better your service gets. All of this adds up to higher switching costs and more defensible moats. Incumbents have considerable advantages over startups. First, they’re more likely to already have access to relevant personal data, not to mention the compute to process it. Second, many are already leaders in building the leading LLMs or are using their market power to access them. ISSUE 1: PERSONAL AI | 21
3 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: VIABILITY Finally, and perhaps most crucially, OpenAI’s ambitious efforts to link their incumbents already possess the model with UI and distribution seems distribution channels, including the more promising. For an upstart to win in most-used hardware, apps, and websites the personal AI space, they will need to through which to deploy personal AI solve the distribution problem and become solutions. These established companies, a platform. This is no easy feat. A look at while far from perfect (some further the history of technological innovation than others) also have established reveals a pattern where excitement about practices around privacy, content decentralization and openness runs high moderation, security, and ecosystem early in the innovation cycle, only to give governance—including the power to shape way to more centralized options once interoperability standards—that may give companies figure out ways to make money them an edge grabbing market share in off centralization. It’s also notable that a complex, multi-agent world. according to reporting by the Financial Times, Amazon, Microsoft and Google The Humane and Rabbit demos accounted for two-thirds of the $27bn mentioned in the previous section raised by AI startups in 2023. capture our imagination because they hint at the possibility new upstarts may shake Any new platform would have to get big up an ossified consumer tech landscape. enough, fast enough to fend off acquisition We imagine new hardware, new platforms, and establish its position. None of this is and new modes of interaction. News such to say that upstarts can’t emerge and knock as the rumored hardware collaboration the tech giants off their pedestals. It’s between OpenAI and former Apple Head early yet. Perhaps personal AI disrupts of Design Jony Ive only adds to our sense some sectors and categories while of anticipation. sustaining others. Before we get too carried away, in our WHO PAYS FOR THESE THINGS: opinion, Humane and Rabbit compete SERVICES > ADVERTISING? primarily on UI, betting they can deliver The business model of the internet is something so compelling consumers will powered by advertising, but we don’t really abandon their existing devices and jump know how ad-supported revenue will evolve on board. Less certain is the quality of in a personal AI world. Could advertising data these companies have access to, how as we know it today disappear entirely? effective their models are, and how much Because personal agents are so their distribution can scale. Less certain is intimately connected with their users, the quality of data these companies have the value exchange will be more access to, how effective their models are, explicit and examined than it is with and how much their distribution can scale. today’s applications. ISSUE 1: PERSONAL AI | 22
3 SALESFORCE FUTURES ANTICIPATING PERSONAL AGENTS: VIABILITY A note on privacy “Consider this thought experiment We’re already sharing a lot of about personal agents and privacy. our lives online and are under Imagine being constantly surrounded pressure to share more. Recent by cameras that track your location in ads for the Meta/Ray Ban smart space, sensors that track your emotions glasses emphasize the benefit and thoughts, and devices that record of live streaming and recording every word you say. This is the kind of while continuing to, in the future we foretold in Minority Report, specific ad in question, dance and it was not a future people wanted through the streets wearing to live in. a yellow jumpsuit. New AI tools, such as Google Gemini, How many people would put up with default to requiring access to that in exchange for an agent that all your personal messages could give you answers grounded and permission to retain your in the best intelligence? Who would conversations for years. actually be willing to give their agent Peter’s thought experiment enough context to ensure its actions are highlights the complex trade- as accurate as possible? I know a few offs around privacy and utility, people who would make that trade, but what we’ve referred to for more I know a lot more who wouldn’t. than a decade as “the creepy or cool equation.” If sharing your data creates sufficient value for you in return, it’s cool. But if PETER SCHWARTZ you perceive that the company CHIEF FUTURES OFFICER AND asking for your data gets most ADVISOR, OFFICE OF THE CEO SALESFORCE of the value, it’s creepy. ISSUE 1: PERSONAL AI | 23
3 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: VIABILITY Agent gatekeepers with access to vast stores of knowledge might radically diminish the habits of consumer research and comparison, which is where discovery-focused advertising lives. Consumers may also turn on those filters we mentioned earlier. If all this happens, the rise of personal agents could deflate the advertising revenues that power the internet the way Craigslist did to print media. Users will use something, and pay for it—either with time, attention, data, or their hard-earned dollars—if they find it useful. Those who subscribe to ChatGPT or Google Gemini today pay for access to a baseline set of skills, but it remains an open question if people will continue to pay for intelligence as competition grows fiercer. When personal AI requires additional skills to accomplish tasks, a plausible outcome may be the proliferation of pay-per- skill marketplaces where consumers pay to upgrade their agents in an incremental and case-by-case way. If personal AI agents are able to accomplish high-order tasks and goals, one can also imagine people paying for outcomes in the same way we see now in some B2B sectors. In the multi-agent world, where agents are interacting with other agents at the speed of light, we will likely see an emergence of a new economy where deals and transactions are completed with personal agents acting on our behalf. Consider how personal agents could deliver growth through machine-to-machine, or M2M transactions. Gartner and others have been exploring these ideas recently, but they were anticipated as far back as 2011 by economist W. Brian Arthur in “The Second Economy,” a classic article foreshadowing the rise of “autonomous economy” where machines communicating with each other drive economic value independent of humans. Chris Noessel’s concept of “post-attention value,” Kevin Ashton’s “Internet of Things,” and other ideas also point us to new opportunities this world presents. ISSUE 1: PERSONAL AI | 24
3 SALESFORSALESFORCE FUTURESCE FUTURES ANTICIPATING PERSONAL AGENTS: VIABILITY Given all of this economic potential In their recent HBR article on how to build and how much advantage agents will trustworthy assistants, Kathy Baxter and confer on their users, one additional Yoav Schlesinger of the OEHU point out possibility is that state-sponsored four key issues that will undermine trust agents step in to democratize access. in agents: integration and amplification, In this world, access to powerful AI, as a lack of a human in the loop, poor Salesforce CEO Marc Benioff recently data quality, and responsibility gaps. put it, is seen as a right, not a privilege, Technical concerns, such as security, and the line between the public and reliable grounding, and protocols for private sector begins to blur. In a later cross-ecosystem access challenge section of this magazine, we imagine some of the loftier fever dreams of one way this future might play out. a future dominated by agents. Early incidents like Air Canada being forced to acknowledge a policy invented by a TRUST WILL BE ESSENTIAL FOR chatbot show how far we still need to go EFFECTIVE PERSONAL AGENTS before companies and consumers can Trust, Salesforce’s number one value, trust personal AI across all of the use will play a fundamental role in shaping cases mentioned here and elsewhere. personal AI futures. Agent relationships are, by default, intimate and ongoing. Recent research by Salesforce points to Agents’ need for context pushes trust to the a gap between customers’ overall trust forefront, suggesting a world where aligning in companies and their belief those with human goals becomes the primary companies will use AI responsibly. For measuring stick for personal AI and a must- example, while 76% of customers have ingredient for long-term customer trust companies to make honest claims relationships. Trust concerns about agents about their products and services, only stem from ethical and technical issues. 57% trust them to use AI ethically. This trust gap points to the importance of Salesforce’s Office of Ethical and Humane designing trustworthy AI from the outset, Use (OEHU) is a leading voice on ethical particularly when it comes to considering concerns, and an excellent source of the role humans will play in AI systems. responsible AI frameworks, guidelines, and best practices that help us anticipate and guard against trust failures. The OEHU has implemented consequence scanning in how Salesforce designs and builds products. ISSUE 1: PERSONAL AI | 25
SALESFORSALESFORCE FUTURESCE FUTURES IN CONCLUSION This framework allows us to imagine with greater clarity a set of plausible, How will relevant, and challenging futures, with agents shape a particular focus on how personal AI may revolutionize the relationship our world? between customers and companies. Next, we explore three pairs of scenarios Our goal has been to equip around the following questions: Will agents you with a framework for level the playing field or widen the gap? understanding where personal Who pays for agents? Will we get our agents from new or existing providers? AI agents are today and the most critical uncertainties that will Turn to the following section, Imagining need to resolve for the category Personal AI Futures, to see how different to truly emerge. But the question answers to these questions might result of how feasible, desirable, and in very different futures, experience what those futures might look and feel like, viable personal agents will and learn the key questions and strategic only be answered over time. implications businesses should consider. ISSUE 1: PERSONAL AI | 26
SALESFORSALESFORCE FUTURESCE FUTURES Imagining Personal AI Futures Looking out 3-5 Years
SALESFORCE FUTURES At Salesforce, our mission is to anticipate, imagine, and shape the future. In this section, we imagine some of the ways things might play for personal AI over the next 3-5 years. For each of our three big questions, we consider a pair of possible futures, bring these futures to life with speculative artifacts like ads, news articles, and product pages, and call out the most important questions and implications for businesses to consider. metadata Each future is calibrated with three potential measures of impact. SURPRISE FACTOR How unexpected is this future? TIMELINE How far out is this future: near-term, long-term, far-out? SCALE Could this future scale from a local to global level?
SALESFORSALESFORCE FUTURESCE FUTURES SPECULATIVE FUTURES Will personal AI level the playing 昀椀eld or widen the gap? In a future 昀椀lled with personal AIs, access to powerful, latest- gen tools will be critical to competition. Who gets access, and what they get access to, will determine whether opportunity expands or agents exacerbate existing inequalities and serve the privileged few.
SPECULATIVE FUTURE #1 Personal AIs turn out to be great equalizers, What if AI delivering more access to education, career levels the coaching, and business services for those with the drive to access them. Broader availability of better tools playing 昀椀eld? expands opportunity and bridges the achievement gap. Accessibility and inclusion improvements 昀氀ow to all. Companies and society also bene昀椀t from access to a broader talent pool as people learn to use personal AI tools to 昀椀ll in knowledge gaps. How an in昀氀uencer started her company with AI In 4 steps
SPECULATIVE FUTURE #1 With no business or tech background, Jordan turned her highly opinionated, insanely popular headphone reviews into an empire. How did she turn likes into buys? Everyone knows, I have strong opinions on headphones. I want sustainable, durable products that look good, hold 1 up, and sound so good they make you cry. I started doing the review videos just for fun. Suddenly, I had an audience. Then, before I knew it, I had a really big audience. My fans kept reaching out to me with ideas. It was, “You should design your own headphones,” or, 2 “You should charge for personal advice.” These were not things I knew how to even consider, let alone do. I started asking some of the new AI tools for help. These tools became my digital team. For me, the first winning strategy, in addition to building my audience, was partnering with companies I admired to help them 3 refine designs before they went to market. My digital team members helped me learn how to incorporate my business, pitch to investors, design my brand, set up payments, refine my offering, get it to market, and run my first marketing campaign. Basically everything. 4 Eventually, I realized I no longer needed to partner with these companies. I had relationships with designers, manufacturers, and distributors. I had loyal customers. It was time to make my own products. I used digital team members to vet partnership opportunities, draw up contracts, and optimize my audience outreach. I just had my weekly meeting with my new digital CFO. The way I think of it is, I have six human team members and hundreds of digital ones.
SPECULATIVE FUTURE #1 surprise timeline scale Questions and Implications If personal AI levels the playing 昀椀eld... How will AI tools What sectors and How might super-empower activities will be most companies use new entrepreneurs? disrupted by a wave of AI to up-level AI-driven startups? their own talent?
SPECULATIVE FUTURE #2 The most e昀昀ective personal AI “chiefs of sta昀昀” are What if pay-for-play luxury items, used by elites to sharpen personal AI their edge and further their advantages. The best widens the chiefs of sta昀昀 deepen societal divides and reinforce gap? consolidation of power by o昀昀ering exponential skills and productivity gains to users who can a昀昀ord them. Meet Victor, the world’s best AI chief of sta昀昀
SPECULATIVE FUTURES #2 Think of Victor as the entire personal team of a Fortune 500 CEO all rolled into one supremely competent digital employee. The best part? You don’t need a CEO salary to a昀昀ord elite service. With three convenient price tiers, there’s a Victor available to everyone. Bronze Gold Platinum Pay-per-skill 5 skills Discounted skills Unlimited Skills Choose monthly; one-time fees 25% o昀昀 skill purchases any skills from our library Ad-supported 3-4 Ads weekly Ad Free 2 hours of weekly 10 hours of weekly meeting support meeting support Unlimited meeting support One size 昀椀ts all Semi-customized Fully Customized Our most popular Choose from a selection Personality adapts and personality of personalities learns based on your inputs FREE $25 / MO. $100 / MO. ACCESS
SPECULATIVE FUTURES #2 surprise timeline scale Questions and Implications If personal AI widens the gap... When employees Will AI chiefs of How will this a昀昀ect bring their personal sta昀昀 allow their income inequality AIs to work, will users to scale their and perceptions of their productivity productivity? fairness around AI? correlate with their individual ability to pay for AI skills?
SALESFORCE FUTURES SPECULATIVE FUTURES Who pays for personal AI? As the old proverb says, you don’t get something for nothing. Whose interest will the agent act in? While incentive structures lurk in the background of our current technologies, we believe the intimacy of agentive interactions makes the value exchange between users and buyers far more explicit. The way this market evolves determines who controls knowledge, who controls information, and whether access to personal AI is a privilege or a right.
SPECULATIVE FUTURES #3 “Good enough,” ad-supported products force What if users to endure sponsored messages in exchange personal AI is for new skills, tools, and better deals. How much ad-supported? would people pay to ditch these interruptions and access more premium o昀昀erings? What happens if they can’t a昀昀ord the alternatives? AD – SUPPORTED
SPECULATIVE FUTURES #3 BELLA One interface. Every application in the world. No monthly fees. Business. Entertainment. Productivity. Travel. Bella partners with the world’s best companies to connect you to the o昀昀erings that make the most sense for your needs. You win. Our partners win. Bella wins. Join millions who love accessing the world’s best concierges without expensive monthly payments.
SPECULATIVE FUTURES #3 surprise timeline scale Questions and Implications In an ad- supported AI future... How might an ad- How will consumers Might ad-supported supported agent trade o昀昀 attention models expand work as a business for personal AI skills access to personal model? and services? AI to previously underserved segments?
SPECULATIVE FUTURES #4 What if personal AI is state- supported? In another plausible STATE – future, the provision of public goods includes SUPPORTED state-sponsored aides for all citizens. It’s easy to imagine the bene昀椀ts: e昀케cient discovery of resources, community initiatives, and services. But using these services means buying into a public ethos, and there could be a worst-case scenario where some states abuse this access to reinforce control.
SPECULATIVE FUTURES #4 Activate Your AIde! The more you use your AIde, the more you do for your country. Access healthcare. Access 昀椀nancial bene昀椀ts. Pay transit fees. NO Enter state drawings. ADS! Apply for university. Apply for government jobs. Report anti-social behavior. AIDE.gov
SPECULATIVE FUTURES #4 surprise timeline scale Questions and Implications In a state- supported AI future... For what use How might Could this alter cases might a companies partner considerations highly functional, with governments about which state-supported to deliver public countries are personal AI be services? most desirable for more appealing citizenship? than private sector alternatives?
SALESFORCE FUTURES SPECULATIVE FUTURES Will personal AI come from new or existing providers? Companies with vast resources, including data, compute, and intelligence, would seem to have the advantage in a race for personal AI market share. When new upstarts like Rabbit and Humane emerge, one question you often see in the comments sections where people discuss such things is, “Great, but why can’t I just do this on my phone?” Still, disruptive technologies have a way of disrupting established paradigms. Might new providers emerge who take the personal AI market in totally unexpected directions?
SPECULSPECULAATIVE FUTURES #2TIVE FUTURES #5 What if What if agents personal widen the AI comes gap? from existing providers? How may I help you? Incumbents bundle personal AIs with existing Maryam, what products and services. would you like to do? This is the world of Siri on steroids and Alexa on amphetamines: familiar players with new, next gen personal AI interfaces that intermediate the relationships between customers and companies (and take a small piece of every transaction). This world isn’t perfect, but a lot of consumers love it. Maryam is a Hawaii-based freelance stitcher with a packed schedule. She loves time with family and friends, especially in nature. How does she do it?
SPECULATIVE FUTURES #5 These are a few of my favorite prompts…. Scans all meetings, compares them to past meetings, and adjusts allotted time accordingly. “Streamline my calendar” This kills every distraction unless there’s an emergency involving a loved one. It’s essential. Admittedly, sometimes I do miss “Go Zen my distractions, but not enough to Mode” give up time outside. I want to be with friends and family and 昀椀nd time to get outside. This command helps you schedule “Prioritize these things well in advance, which my people” improves the odds they happen. My AI is a ruthless negotiator. “Negotiate. You If I tell her “you can do better,” can do better.” she gets even sharkier. I should probably check and see if there’s a better deal somewhere, but who has the time?
SPECULATIVE FUTURES #5 surprise timeline scale Questions and Implications If personal AI comes from existing providers... Will there be Is voice really the Similar to the network e昀昀ects most dominant explosion of here, or will rent- mode of AI advertising taking confer interaction, or technology disproportionate do other modes companies that power to fewer ascend? emerged with companies? Web 2.0, will a new category of enterprise so昀琀ware emerge around personal AI?
SPECULSPECULAATIVE FUTURES #2TIVE FUTURES #6 What if What if agents personal widen the AI comes gap? from new providers? Just as people settle on 9 or 10 apps on their phone today, people choose nine or ten AI-昀椀rst products for the vast majority of their needs. Users highly value control in selecting their tools rather than letting an AI oversee these choices. As a result, we see an explosion of AI startups that target speci昀椀c use cases.
SPECULSPECULAATIVE FUTURES #2TIVE FUTURES #6 EINSTEIN SERGEANT PEPPER What if agents “My real genius is making “Last task, 3 minutes. GO!” widen the your customers happy.” Struggling with gap? Customize Einstein with the procrastination? skills and tools you need to Need some motivation? get the job done. Add pepper. Meet DIGITAL My YENTA “I’m a spicy matchmaker Crew here to save you from swiping.” LULU Tao is a Manila- Keeping your love based marketer/ life in check... product designer. He uses personal AIs for “Feeling creatively stuck productivity, production, on a design?” recreation, and everything in between. She may never give you Tao shares his favorite AIs the answer you expect, on Meet My Crew. but you might 昀椀nd the answer you need. DR. JULIA SILVA “Talk to me anytime, day or night.” The real Dr. Silva will learn from your conversations so that your next visit with her is even more useful.
SPECULATIVE FUTURES #6 surprise timeline scale Questions and Implications If personal AI comes from new providers... How might startups In a world where Will new personal AI get to context faster people see value providers also open than incumbents in having multiple the door for new without losing trust? agents, how do players in enterprise people 昀椀nd, vet, so昀琀ware? and purchase personal AIs?
SALESFORCE FUTURES coming soon in Futures Issue 2... Will AI help us build better futures?
SALESFORCE FUTURES references and inspiration FUTURES CURATORS David Berthy, Lynn Carruthers, Mick Costigan, Marc Escobosa, Angela Gleason, Daniel Lim, Peter Schwartz SALESFORCE “HOW COMPANIES CAN BUILD TRUSTWORTHY AI ASSISTANTS” Kathy Baxter and Yoav Schlesinger | HBR “ARTIFICIAL GENERAL INTELLIGENCE: CASTING A VISION FOR A BETTER FUTURE” Silvio Savarese and Peter Schwartz | THE 360 BLOG “TOWARD ACTIONABLE GENERATIVE AI” Silvio Savarese | THE 360 BLOG “BUILDING TRUST IN AI NEEDS A HUMAN TOUCH” Holly Prouty | THE 360 BLOG “GENERATIVE AI: 5 GUIDELINES FOR RESPONSIBLE DEVELOPMENT” Paula Goldman and Kathy Baxter | THE 360 BLOG “WHAT IF SUPERINTELLIGENCE RAN YOUR BUSINESS” Itai Asseo | THE 360 BLOG “FROM COPILOT TO COORCHESTRATION” Silvio Savarese | SALESFORCE AI RESEARCH BLOG ADDITIONAL ARTICLES “THE NEW NEW MOATS” Greylock Partners | LINK “OOZY INTELLIGENCE IN SLOW TIME” Venkatesh Rao | RIBBONFARM “IT’S NOT A COMPUTER, IT’S A COMPANION” Justine Moore, Bryan Kim, Yoko Li, and Martin Casado, Andreessen Horowitz | LINK “ROBOT KEGGERS AND ROOMBA SPIES” Christopher Noessel | MEDIUM “A LEADING MEMORY RESEARCHER EXPLAINS HOW TO MAKE PRECIOUS MOMENTS LAST” David Marchese | NEW YORK TIMES “THE SECOND ECONOMY” W. Brian Arthur | MCKINSEY QUARTERLY ISSUE 1: PERSONAL AI | 51
SALESFORCE FUTURES references and inspiration “UNLOCKING POST-ATTENTION VALUE” Christopher Noessel | LINKEDIN | BOOK “LLM-POWERED AUTONOMOUS AGENTS” Lilian Weng | LIL’LOG “THE ANATOMY OF AUTONOMY: WHY AGENTS ARE THE NEXT AI KILLER APP AFTER CHATGPT” SWYX and Alessio | LATENT SPACE ACADEMIC PAPERS “LEARNING UNIVERSAL PREDICTORS” Jordi Grau-Moya, Tim Genewein | LINK “TOWARD REASONING IN LARGE LANGUAGE MODELS: A SURVEY” Jie Huang, Kevin Chen-Chuan Chang | LINK “CHAIN-OF-THOUGHT PROMPTING ELICITS REASONING IN LARGE LANGUAGE MODELS” Hason Wei, Xuezhi Wang | LINK “TREE OF THOUGHTS: DELIBERATE PROBLEM SOLVING WITH LARGE LANGUAGE MODELS” Shunyu Yao, Dian Yu | LINK “THINKING FAST AND SLOW IN AI: THE ROLE OF METACOGNITION” Marianna Bergamaschi Ganapini, Murray Campbell | LINK VIDEOS “SALESFORCE 2030: A GLIMPSE OF AN AI FUTURE” Salesforce Futures | VIDEO “THE DISAPPEARING COMPUTER” Irman Chaudhri | TED TALK “INTRODUCING R1” Rabbit r1 Demo | DEMO “PERSONALIZED AI FOR YOU” Google Gemini | DEMO “OS1 ONBOARDING SCENE” Spike Jonze 昀椀lm Her | FILM CLIP COMPANIES REFERENCED PARCHA, PACTUM, RABBIT, OLLIE, HUMANE, CHARACTER.AI ISSUE 1: PERSONAL AI | 52
SALESFORCE FUTURES ACKNOWLEDGEMENTS HUMAN SALESFORCE FUTURES IS INDEBTED TO OUR SALESFORCE COLLEAGUES AND THE FOLLOWING INDIVIDUALS FOR THEIR DIRECT AND INDIRECT CONTRIBUTIONS TO THE CONVERSATIONS THAT INSPIRED AND SHAPED THIS WORK: SILVIO SAVARESE, KAT HOLMES, LIZ TRUDEAU, SHELBY HEINECKE, ADRIAN SMALLEY, PHIL MUI, ITAI ASSEO, PAULA GOLDMAN, KATHY BAXTER, YOAV SCHLESINGER, JAYESH GOVINDARAJAN, AND ADAM EVANS. AI PERSONAL AI MAY NOT HAVE FULLY ARRIVED (YET), BUT SALESFORCE FUTURES FOUND PLENTY OF USES FOR AI TOOLS IN THE PRODUCTION OF THIS MAGAZINE. IN OUR INITIAL DRAFTS, WHILE EXPLORING CONCEPTS FOR THIS MAGAZINE, WE USED MIDJOURNEY, ADOBE FIREFLY, AND DALL-E TO PRODUCE IMAGERY TOGETHER WITH CHATGPT TO BRAINSTORM, COPY-EDIT AND TO HELP FORMAT THE FAUX ADS. THE FINAL PRODUCTION INCLUDES CUSTOM GRAPHICS PULLED FROM ADOBE STOCK (INCLUDING STOCK AI GENERATED), AND ADOBE FIREFLY (CUSTOM AI GENERATED). FRONT AND BACK COVER ART: DALL-E. IN OUR APPROACH TO THE FINAL VISUAL DESIGN PROCESS IN THIS ISSUE, WE FOUND THAT COLLAGES STRIKE THE BALANCE BETWEEN VARIOUS DIMENSIONS OF HUMAN INTELLIGENCE AND ARTIFICIAL INTELLIGENCE. CONNECTING TO THEMES ON PERSONAL AI, THERE IS A CONSCIOUS TREATMENT TO THREAD THE OCCASIONALLY INVISIBLE INTERACTION OF TECHNOLOGY WITH HUMAN BEHAVIOR, AS PERSONAL AI TAKES FORM. THE VISUAL STORY BEGINS WITH AN ABSTRACT DEPICTION OF PERSONAL AI, LEADING VIEWERS INTO IMAGINING FUTURES WHERE MORE DEFINITIVE VERSIONS OF PERSONAL AI EMERGE. OUR AI COVER ART IS A NOD TO THE COVER OF THE BOOK, THE ART OF THE LONG VIEW: PLANNING FOR THE FUTURE IN AN UNCERTAIN WORLD, BY PETER SCHWARTZ. ISSUE 1: PERSONAL AI | 53
About Salesforce Futures We help Salesforce and our customers anticipate, imagine, and shape the future, building the shared understanding required to tackle adaptive challenges. [email protected] © 2024 Salesforce, Inc.
