8 Forecasts & Implications for the Years Ahead :: 2024+
This special edition features eight forecasts and implications - some new and some reformulations that stimulated the most reaction and debate. All are crucial things to consider for the year ahead...
Edition #14 of Implications.
I hope you’ve had a chance to clear your mental cache over the holidays, and perhaps a force quit and reset. As we shift our gaze to 2024+, let’s consider the implications of some seismic shifts ahead.
If you’re new, here’s what to expect. This ~monthly analysis is written for founders + investors I work with, colleagues, and a small group of subscribers. We don’t cover news; we explore the implications of what’s happening. My goal is to ignite discussion, socialize edges that may someday become the center, and help us connect dots. I am curious which one resonates most, and welcome your riffs or debates on social. These are intended to be catalysts for discussion.
Here’s a quick rundown of eight insights and implications that made the cut:
The future will be personalized to your preferences.
Local OS-native AI models change our everyday lives, and win consumer AI.
Small brands will be 10x more competitive with big brands, and all brands will compete with our objective hyper-personalized AI agents that increasingly make purchase decisions on our behalf.
Digital products continue a rapid path to commoditization and go-to-market advantages compound as AI makes (and replicates) software effortlessly.
Education undergoes an overdue refactoring that completely transforms K-12 curriculums, higher-education, government budgets, and the role of teachers for the next generation.
The talent stack collapses across functions in organizations, refactoring operations for both quality and efficiency.
We will realize our sensitivity to algorithms and our vulnerability to being polarized.
Traditional business models are disrupted and reimagined in the age of AI.
(…and then some final thoughts at the end) ;-)
1. The future will be personalized to your preferences.
Of all my convictions, this is my strongest. Every digital experience should be personalized for you based on the information you want brands to know about you - always in your control. E-commerce websites should welcome you by name, use your preferred language, know your gender, preferred sizes, colors, and ages of your kids. Restaurants should know your allergies, favorite dishes, and dietary restrictions. Hotels should know your preferred room temperature, and whether you have rewards status at other hotels or airlines, or if you are an influencer. Future AR and mixed reality experiences should be hyper-personalized, including your favorite colors and personal tastes and interests. Your preferences should be managed by you and shared purposefully, not deduced or purchased from others. What implications can we expect in the coming years?
Step one, sync yourself. On the consumer side, you’ll have the opportunity to sync yourself (a personalized model trained off your data aggregated from a variety of sources) with any application or experience you engage with. The killer app here will be an intelligent arbiter - an “agent” of sorts - that acts on your behalf to selectively share your data and personalize as appropriate.
“Recommendations” kill “Favorites”: No doubt, AI has reached the point in some verticals where it knows our taste better than we do. In such a world, the model’s recommendations will please us more than our own favorites. Will your personal travel history suggest recommendations that transcend a Google search or a travel agent? We’ll look back and realize that “favorites” were always quite limiting, as was discovering your favorite places and creations by accident.
The benefits of personalization compound over time. I am thinking a lot about the compounding benefits of a relationship with a personalized ML model after years of working together. Does the model discover our quirks and biases? Does it develop a degree of sarcasm in order to engage us on our own terms? And what does this mean for marketing and the world’s largest brands?
Non-portable data will be the ultimate moat. Three vectors of differentiation matter most for new products: (1) is there a unique dataset - or proprietary understanding OF a dataset - that yields an advantage?, (2) is there a significant distribution advantage or a final mile advantage (like critical adjacent tools in a given workflow)?, or (3) is there a revolutionary interface advantage that, for some reason, is 10x better?
2. Local OS-native AI models change our everyday lives, and win consumer AI.
Many people seem to assume that just a few massive LLM companies will win the consumer AI application game, much like Google Maps and a couple others became the ultimate winners (and underlying APIs) for all location-based consumer applications that enhance our everyday lives. But it is increasingly clear to me that consumer AI will be all about personalization, trust and privacy, and cost efficiency.
The LLMs running locally on our phones will be 10x better for everyday consumer use-cases. Technology ultimately thrives when it meets us where we are and brings us where we want to be. Personally, I want a highly personalized agent that knows everything about me, without compromising my privacy. I want my agent to sync with every data source in my life, but live “locally” (on my device, without adding any privacy risk to the equation). In such a world, this agent will tell me what to buy and where to buy it with the greatest and most objective quality and price comparison engine imaginable. My AI agent will block every spammy text for me, will sort through my email and deliver a daily briefing synthesizing everything it knows I care about, and will help me make decisions better than any person could. It will truly know me. For this world to happen, the LLM’s-powering such capabilities will need to live on my device and be fully enriched and interconnected to everything else on my device. Once this happens, I suspect LLMs in the cloud will serve a different and higher order purpose than answering random questions and conducting the glorified internet search.
Where is the technology to make this happen? Apple has published research on using flash storage and a multitude of efficiencies for running powerful ChatGPT-like models locally. Microsoft has already announced devices capable of running pretty powerful LLMs locally. And I think we’re just getting started, especially with the contributions of META’s open source models and the development of new mobile chips tailor made for this new world. Suffice to say, it is all coming sooner than we think.
3. Small brands will be 10x more competitive with big brands, and all brands will compete with our objective hyper-personalized AI agents that increasingly make purchase decisions on our behalf.
Small brands will gain the operational benefits of big brands without the overhead of a big company. From customer service and marketing execution to procurement and internationalization, the next generation of AI-infused SaaS products will level the playing field for small businesses, allowing them to compete with larger brands in unexpected ways.
AI will threaten subjectivity in purchase decisions, and with it the sway of brand and marketing. As we gain trust in the guidance of agent-assisted experiences, will the impact of brand, referral, and relationships in purchase decisions be diminished? Whether we’re buying batteries, sneakers, potato chips, or kitchen appliances, we are more influenced than we care to admit by brand perceptions as opposed to factual comparisons. However, as your “AI Agent” gets to know you better - infused by every personal preference and previous purchase as well as every online review and consumer reports determination - you’ll start trusting the guidance of your agent more than any other signal. Perhaps the stakes are even more pronounced in the enterprise, where a procurement process tainted by human emotions, laziness, and previous relationships is the persistent fear of any CFO. AI threatens subjective decision-making tainted by human error and bias and will usher in an era where the best product at the best value may in fact win.
Modern brands need the plumbing to “think and act in real-time.” In the modern digital world, wit and timing matter. Brands must be able to create content and reply within social feeds in real-time. This is “agile marketing” as opposed to the traditional “macro marketing” that involves a long supply chain of participants to produce marketing content. What must happen under the hood for brands to empower brands to pull off such marketing feats in real-time? Artificial intelligence is the greatest unlock, and I’ve had a front row seat watching brands of all sizes start using Adobe’s Firefly AI models and especially Adobe Express as a content suite that enables AI-content creation while also connecting social media marketers at big companies with their colleagues using products like Photoshop and Adobe’s marketing products. Such plumbing allows for anyone within a brand to instantly whip together content with the right brand assets and logos while adhering to brand guidelines. But beyond the plumbing, big companies must modernize their marketing practices and embrace the real-time nature of brand building.
The future of creativity and digital marketing will be fundamentally different. More empowering, less dependent. Less generalized, more personalized. Less time-constrained, more imagination constrained. Consider the shifts my strategy team and I find most profound for the era ahead:
4. Digital products continue a rapid path to commoditization and go-to-market advantages compound as AI makes (and replicates) software effortlessly.
Software will increasingly be developed using prompts as opposed to software development as we know it. If you consider what % of functionality in any software we use has existed in any software that came before it, then the “commoditization of software via AI” argument makes a lot of sense. Models are currently being trained on all software that came before them. Ultimately, just by describing a set of capabilities, you’ll get functional apps. The implications are mind-boggling. Will all internal apps be made without engineers? Will providers of such apps become irrelevant? When any screenshot can be turned into code, and any application conjured up with a text prompt, software becomes as accessible as electricity. In such a world, what makes software defensible? What key technologies will we use?
In such a world, our data and personalized experience is the ultimate form of retention. When consumers have superpowers deeply integrated into everyday workflows (and can conjure up pretty much any capability with natural language) we’re less likely to endure switching costs to alternative tools. The AI superpowers emerging across every discipline are generally built as APIs on day one, which makes integration into market leading tools pretty effortless. And companies that know their customer better than anyone else (and have the data advantage that comes with such relationships) are in the best position to build the models that would otherwise disrupt them in the platform shift to AI. As such, the go-to-market advantage of those with customers will only increase as the AI era unfolds (and product features will be less defensible).
The last few decades will become known as “The Graph Years,” as it becomes increasingly difficult to build any NEW social or professional graph in a privacy-centric, increasingly personalized, regulated world. Those of us around in the early 2000’s know that the hacks and tactics to build social and professional graphs like Facebook and LinkedIn don’t fly anymore. We aren’t willing to blast all of our friends with a click of a button or quickly sync of our entire address book to a startup’s server. So, for new products that require network effects and social graphs, the old meritocratic days of “launch a great product and users will follow” are over. Instead, a startup team needs to be GTM driven, embracive of partnerships, build far superior interfaces, and make better use of the data they get than incumbents.
5. Education has an overdue refactoring that transforms K-12 curriculums, higher-education, government budgets, and the role of teachers for the next generation.
We’re on the precipice of a fundamental shift in education, powered by AI. While special interests and legacy biases will persist, we have the opportunity to transform learning from a generalized pursuit that leaves people behind by design (all about grades and outperforming others) to a personalized pursuit designed to meet each student where they are and identify their strengths and opportunities over the course of a lifelong partnership. We can slash wasteful and outdated educational spending, we can refactor the architecture of public education and the role of teachers, and we can prepare students for a fundamentally different workplace.
Education is notoriously expensive and inefficient, and AI has the opportunity to fundamentally refactor learning as we know it starting with “lifelong tutors.” Countless attempts - from charter schools to performance-driven funding legislation - have been taken to refactor the cost basis of K-12 education in America. The staffing and spending per student has increased substantially since the 1970’s (see data source below), yet the reading, math, and science scores have remained stagnant. Something is horribly wrong. We must radically reimagine and refactor the cost and methodology of traditional education amidst the AI platform shift. Imagine a hyper-intelligent AI-powered highly personalized tutor for every single student starting in the first grade that patiently coaches, constantly encourages confidence and reinforces skill development, and informs human teachers of gaps and areas of development in real-time. Such a “lifelong tutor” would get to know each us and provide an unparalleled holistic educational support system - a true 100% available and scalable partner for each student’s educational journey. One great side-effect of AI accommodating and tailoring learning for every student: the end of old-school anxiety-inducing assessments in the forms of tests and quizzes. Such assessments are, in my opinion, the Achilles heal of education - they make students hate learning and turn the means of education into ends. After all, the purpose of learning is not to get good grades but to apply the knowledge in meaningful ways to enrich your life and the world around you.
New skills emerge. In Edition 2 of Implications, we discussed the art and science of prompt engineering and how we might expect a return of the Socratic method for teaching - at scale and on demand. First, when it comes to directing AI-powered tools, we can expect a new logic-based lexicon for how to inform, constrain, and optimize the prompts we give to AI that impact outcomes. Much like students learn Excel and calculators - and even how to use more advanced formulas and engineering calculators - how do we outfit students to make sure AI is working for them (and not the other way around)? When it comes to learning from AI-powered tools, I suspect it will resemble the Socratic Method, anchored on dialogue between teacher and students, fueled by a continuous probing stream of increasingly pointed and intelligent questions. The method is designed to explore the underlying perspectives that inform a student’s perspective and natural interests. The framework is optimized for surfacing relevance and stoking organic intrigue. Imagine history “taught” through a chat interface that allows students to interview historical figures. Imagine a philosophy major dueling with past philosophers - or even a group of philosophers with opposing viewpoints.
6. The talent stack collapses across functions in organizations, refactoring operations for both quality and efficiency.
Organizational design changes in an era where every tool is multi-player, many tools become headless services, and companies unlock higher ROI through cross-functional roles. Startups know (and big companies are learning) that the best marketing IS the product. Modern companies don’t distinguish so strongly between product and product marketing, product design and development, and user experience and copywriting. Why? Modern organizations are story first and design driven. Not only do they want to avoid the “game of operator” between functions, fiefdom building, and the organizational debt that accrues by scaling bigger matrixes, but they also want design to lead. They want the product to be an end-to-end story (in ways only empowered designers can achieve) — one that every member of the team embodies and executes. They want the product to indoctrinate the brand. These aren’t new ideas, but these ideals are finally possible in an era of multi-player tools, scalable underlying third-party components, and more granular real-time performance metrics (for both people and products). In the year ahead, consider making your product manager and product marketing leader the same person. Empower your design leader to own the interface and drive innovation…
Collapse the talent stack every chance you get. As I reflect on the teams I’ve led and hundreds of start-ups I’ve worked with, there is a consistent unfair competitive advantage i’ve witnessed when the talent stack was collapsed - when the lead designer was also the product leader, when the front-end engineer was also a designer, when the designer is also a great copywriter, when the product leader was also the founder/ceo, etc. Tighter conduits for decision making and synthesizing information are an incredible advantage when it comes to crafting products. Many start-ups enjoy the benefits of collapsed talent stacks and then undo them as they grow (and most big companies just don’t understand this). In your hiring (and your consolidating), collapse the stack whenever you can. The time has come to reimagine organizational structures, roles, and norms.
AI-powered management tools will isolate problems and decisions, and streamline decisiveness (we will be managed, in part, by AI!). I’ve talked about the notion of “organizational debt” for years, which I define as the accumulation of decisions in an organization that should have been made but weren’t. Organizational debt is not reduced by having more meetings. It is only reduced through better management and decisiveness. I’ve seen at least two products reimagining management to be AI driven. And while it sounds crazy at first that we might “work for a computer,” I firmly believe that management will become a hybrid (human-machine) discipline. One thing startups and big companies have in common is being plagued by inexperienced and bad managers who have no idea how to monitor and measure performance, have no idea how to lead without micromanaging, no idea how to run a 1:1 or manage careers responsibly, and struggle to figure these mechanics out (I struggle with some of this as well). Why not assist these managers with AI that crunches the data and suggests what to ask and do? Humans will do more leadership and less management in the age of AI.
7. We will realize our sensitivity to algorithms and our vulnerability to being polarized.
Algorithms that optimize for grabbing attention ultimately drive polarization. When you’re engaging with content that reinforces anything you’re passionate (or angry) about, modern algorithms not only feed you more similar content but also widen the content queue with variations of content from adjacent people. The algorithms are designed to keep and intensify your attention. As a result, the surface area of topics or reasons that stoke and feed your anger increases - and this often includes manipulated media and fake news that was tailor-made to get you amped up, clicking, and sharing even more. The AI driving these algorithms quickly learns that a rational or “both sides” view is less likely to sustain your attention. But the rage-inducing stuff keeps us swiping. The outcome of all this is expanding the surface area of what triggers you (along with the opportunity to serve you more ads). This is amazing when it’s a hobby like football, but polarizing and brainwashing when it is something far more triggering.
Old-school algorithms targeted our ego, but the newest algorithms target our id. The first wave of algorithms driving our feeds were driven by our social graphs and interests we’d tell anyone about ourselves – “I love soccer, I’m a hiker, I play piano” – and would serve up content about those interests (as well as updates about our friends’ dogs, etc.). In contrast, this new generation of algorithms seem to burrow deep within our ancestral lizard brain to the raw emotions – hate, fear, anger, lust, envy – that don’t make us proud but are hard to resist.
How will we navigate this new world? It starts with modern education for digital literacy. Just as third graders now learn how to use a keyboard, we must help children understand how algorithms manipulate us and what questions we must ask ourselves. For us product designers and leaders, what is our Hippocratic oath for the modern era? Should we be designing constraints into the system that purposely diversify feeds and viewpoints at the expense of engagement? Should we be suggesting breaks to social media users after a certain period of rabbit-holing at the expense of optimal engagement? I (cautiously) think this is one of those areas where regulation could help, and it becomes increasingly clear why TikTok as we know it doesn’t exist in China. In China, a different app called “Douyin” owned by the same parent company of TikTok, is a more child-friendly app with educational videos and a time limit. It has been described as “spinach” compared to our “opium.” While I hesitate to resort to moderation and regulation, I am deeply concerned about the long-term implications of letting these algorithms continuously get more effective at polarizing us as they run loose.
8. Many traditional business models are disrupted and reimagined in the age of AI.
Time-based business models are liable for disruption via a value-based overhaul of compensation. Today, as most designers, lawyers, and other trades charge by the hour, the AI-powered step-function improvements in workflows are liable to shake things up. Let’s first tackle this by considering the ultimate SOURCE of the differentiating value delivered to a client: It is less about time and more experience. Of course, there are some exceptions like yard work or lawn-mowing where trade experience may be less of a differentiator in the value delivered, but let’s focus on trades where the spectrum of outputs is long and varied. In such fields, the differentiators that matter are one’s honed skills from formal education and practice, taste and intuition, creativity, network of relationships, and even data and algorithms honed through volume of past experiences. In such a world, the classic time-based model of billing for lawyers, designers, consultants, freelancers etc is officially antiquated. So, how might the value be captured in a future where we no longer bill by the hour? Perhaps there is a new source-of-truth for the “value” of tasks across professional trades via a third-party billing service that determines price? Much like the billing codes and market pricing for medical procedures, these prices could be negotiated with trades and could vary based on one’s years of experience. Or perhaps we enter an era of results-based compensation that is far more objective and measured? In some industries where we start paying less for something, perhaps we end up making up for it in volume? Not sure where we land, but new pricing models are overdue to replace time-based and finger-in-the-wind pricing in the age of AI.
Increasing perversion of certain business models that are liable to be gamed or constrained by AI: We’re shifting from a world where data analysis required long cycles (analysts need lots of time to run queries, analyze, and then present findings in a way that people understand) to a new world of real-time optimization and insights automated by AI. When businesses start optimizing themselves, all sorts of crazy things might start happening. For dating apps, where the perfect match of two people increases churn, will Tinder or Bumble constrain the efficiency of AI so the product doesn’t become too “unsustainably effective?” Or in the world of music streaming: Since Spotify pays artists per song, will Spotify automatically optimize its algorithms to favor longer songs ultimately lowering its costs? As AI gets really good at optimization, we need to be wary of the unintended implications.
The business of traditional entertainment creation will evolve, but not as we expect. There has been a lot of focus lately on potential job displacement. However, I have come to view the future of entertainment as a “core and periphery” model, where the core (Hollywood - and all the players involved with original story creation) only gets stronger and more efficient, and the periphery (user-generated content, unsanctioned sequels, and long-tail spin-offs) grows by 100x. As every brand floods the zone of our consciousness with AI-generated content, we will crave story, meaning, and originality more than ever before. The efficiencies of AI reduce the cost of content creation so we can TAKE MORE CREATIVE RISK. Instead of green-lighting five ideas, perhaps we can green-light fifteen? Perhaps Hollywood will spend less time replaying safe playbooks (sequels and familiar storylines) and more time developing NEW franchises and imaginative storylines? With AI, the core can get 10x better, and the periphery will grow 100x larger. The business model disruption here is where the money is spent in traditional studios. Why not offload the derivative content (the sequels, the animated short offshoots, etc) to a select group of long-tail creators and fans that leverage AI…and then reallocate the saved capital to the core?
Mechanisms that help match the best talent with the right opportunities will drive more creative meritocracy — and challenge “old boy” networks. Imagine a world where IP becomes a little more “open” to a longer tail of talent to play with, but with guardrails (this relates to the periphery type of content we discussed above)? Perhaps a brand like Marvel could invite 100 passionate creators beyond the walls of the studio to engage with their characters using AI models to explore new plot ideas? Perhaps AI will help user-generated content not only improve in terms of quality, but also get exposure from a higher-signal network of curators? So far, social platforms have surfaced content based on what the “critical mass” thinks (number of likes) rather than what the “credible mass” thinks (WHO actually liked the content, and how credible they are as tastemakers).
Gratitude & Let’s Get Ready
I started this monthly exercise just one year ago. Chronicling what I observe/learn, and then having a process to consider the implications has proven a tremendously helpful practice as a builder of products, leader of strategy, and investor. But the real value has been your input in the form of suggestions, challenging the ideas shared in this forum, and helping connect the dots. So, thank you. I appreciate you following along, and I also want to extend appreciation to those who subscribed to the monthly extended edition of Implications - you’ve helped support a number of non-profit organizations I support including the Cooper-Hewitt National Design Museum (and if you’re a colleague or founder I work with and wish to sign up, just ping me!).
Now, as we look to the year ahead, here’s what is top of mind:
Another whirlwind year of breakthroughs: The clock speed of the technology industry will not slow down, and I suspect we’ll spend each week catching up on the latest breakthroughs and reconsidering plans as a result. It is exciting, but tiring as well. Let’s keep analyzing and debating.
Landmark legal declarations that set the bit for AI: We enter 2024 with many open questions related to the use of copyrighted training data (see the recent NY Times lawsuit against OpenAI among others), the use case and possible requirements of content credentials to determine what we can trust, and the regulatory approach to AI especially open source models that are becoming increasingly powerful. Ultimately my bet is some form of safe harbors - much like those bestowed on the early days of the internet - alongside some guidelines on what can and cannot be used for training data.
Human cravings shift toward story and scarcity: In the age of abundance - and the rapid commoditization and accessibility of content and entertainment - there will be a growing human desire for authenticity and story. As every brand “floods the zone” just because they can, I am interested in the unique human desires that empower more niche and non-scalable experiences (and who benefits from providing such services).
AI protects us. There is so much focus on the risks of AI falling into the wrong hands, rampant scamming, and AGI overlords. But I am more focused on sensational new solutions to criminal activity, junk calls, impersonations, and risks of all kinds that are unlocked by AI. From blocklists and filters to remarkable new capabilities that identify risks for all of us in real-time, my bet is that AI will protect us far more than it will harm us.
AI-driven spontaneity and a better first mile for human connection: We need technology that helps us connect with each other, discovering our commonalities more efficiently. Whether dating, professional connections, or serendipitous run-ins that were always possible but rarely realized, AI can crunch large amounts of data and surface these discoveries for us. Perhaps such technology will promote more mutual understanding and vastly increase the surface area of human connection? Let’s hope that we become closer and more human through the technology we create.
Happy new year, folks. 2024, let’s do this.
Receive Implications on a regular basis: