How AI will reinvent Marketing

What happens in a world of infinite labor, infinite content, and mass personalization

You can predict gas stations, but not Los Angeles
There’s a saying that if you could have predicted the invention of the car, you could probably have predicted gas stations. After all, the closest metaphor would have been the horse — which humans used to traverse long distances, carry cargo, etc. (And in fact, I’ll remind you cars were originally called horseless carriages for exactly that reason). Just as horses need watering holes and grass, so too do cars need gas. Thus, gas stations. QED.

But it would have been very hard to predict the creation of Los Angeles, a city built on sprawling streets which could only be traversed by car. Or Walmart, enabled by the suburbs and exurbs that were enabled by cars providing long-distance daily commutes. Or the dynamics of the conflicts of the Middle East, driven partially by both blood and oil, and this indirectly from the car boom which enveloped humanity in the last hundred years. Of course these effects have all been huge — you’ll note that as of 2024, there are ~300 million cars in the US. That’s >1 per US adult. Wild.

Yes, this is a fancy way to say: First order effects of inventions are easy to predict but secondary, tertiary, and n+1 effects are very hard. That’s almost laughable to say about a sector like AI where new research / startups / demos / regulations / firings / resignations / fundings are happening at light speed. Like literally there’s something mind-blowing that happens every week if not every day. Thus it’s hard to predict what things will happen in the next year, much less what might happen over the next 10.

And yet we must speculate!

The big question
In this essay, I muse out loud: What will happen to Marketing in the next decade, as AI permeates every aspect of our technology products and experiences?

Think on this for a second.

The boring answers
What might be flashing through your head are the easy ideas:

• ad creatives (and landing pages, and really all assets) will have infinite variations, and it’ll be easy to generate/test many of them

• video, audio, and other forms of content will be much cheaper to create, thus more accessible to a wider swath of marketers

• LLMs will answer with ads embedded next to their answers, targeted at the content

• generative AI will be used throughout the creative process to streamline everything — from creative briefs to concepting to edits to the final assets

We could come up with a million of these, but I admit they are a bit boring.

They’re boring because they basically imagine a world that is unchanging, where we do all the same kind of work and consumers have the same types of experiences, but cute lil AI tools make it all a little bit better. It’d be like imagining a world of cars by simply copy+pasting cars into every mental image where a horse exists. You might get to gas stations from thinking that way, but not Los Angeles.

Infinite labor, content, interactivity, and more
We’d like to get to the strong-form ideas, the AI-native ideas that go beyond what we have today. So first, some building blocks — what does gen AI actually get us?

• Infinite labor. AI lets us use capital to buy compute which can then perform various types of labor. In the fullness of time, AI will let us transmute dollars into design work, project management work, or otherwise, as a team of agents execute your tasks

You can then point that labor towards everything marketing, and it will change our capabilities in revolutionary ways.

• Infinite content. If you have infinite labor, you can redirect that labor towards creating infinite content. Imagine content costs going to zero — you could leverage that in amazing ways, and not simply variations of ads or messages. You can also get mass personalization, where everyone gets their own personalized content — a video ad with a celebrity avatar talking to you and your particular needs. It can be real-time, where you create a billion variations of a new brand campaign, and if it decreases sentiment, you can update all the video/audio/image/text in near real-time without spinning up a multi-month project

• Instant internationalization. We think of launching products today country-by-country, usually to English-speaking markets, and then go from there. Why not simply launch to the entire world and have all the ads, message, and even product UX be in all the right languages? Why not have all the images reference the right cultural touchstones, with the best colors and imagery for each culture?

• White glove everything. Alongside that, onboarding into any new product, or dealing with problems, will feel like a concierge experience. Of course you get a detailed walkthrough of any product or experience, if you want it. Imagine the service you get at a 5 star hotel or at a high end boutique — that’s your new computing experience. Or if you want to just have the AI produce an outcome, you just ask for that. (“Don’t show me the latest jackets, just buy the one I want.”) And if there’s problems, customer support will just solve it automatically (or with extreme guidance and empathy), like a concierge.

• Explosion in the depth of content. Let’s say you create a complex product like a video game, and then you want to promote it. You would then use a lesser medium, like a video trailer, to demo the video game. But what if video games themselves became trivially low-cost to create? Would you then use a video game to promote a new toy you’re selling? And if so, why not 100 video games? Or a season of an entire TV show? And if you’re watching a TV show, and you hit pause, will you be able to then enter the world and endlessly interact with the characters? You could talk to them for hours. Or add new chapters or side stories for your favorite characters? Or better yet, would you have generative AI build an entire award-winning level TV show simply to promote your new watch brand? You could imagine the depth and complexity of different digital assets become interchangeable and infinite, such that what we consider ads right now might be better seen as experiences/products in themselves.

• Channels get upended. Here’s a fundamental thing: Marketing channels are created when customer behaviors enable marketing messages to be put next to organic content. If snail mail is invented, so too can junk mail. If you answer the phone to your friends, you can now dial a 1-800 number to learn about a new product or service. Newspaper ads exist when consumers want to open their paper in the morning. And it fails when people stop consuming. So in a world of endless content variation, it’s very hard to predict what happens to the surface area of organic consumption and conversation. Will LLMs generate so much content that SEO/SEM becomes worthless? That feels inevitable. Will talking to LLMs become commonplace and eventually the default? What happens if we go to voice and chat as the primary way to interact with computers, as we manage them more as we’d manage a team of employees? Will marketing messages become embedded in those conversations, either as sponsored recommendations or otherwise? Many new marketing channels will be invented, and many existing channels will have to be reinvented (or will ride off into the sunset). Hard to predict the dominant UI paradigm in the coming years.

• Companions as a dominant new channel. I do think it’s pretty clear that talking to companions deserves its own mention. We already have so-called parasocial relationships with streamers and celebrities where we feel like they are friends, but actually we’re never going to meet them. AI companions can of course be that but on steroids. Today a startup might never hire a mega streamer to promote their new app because it’s just expensive and costly to buy a sponsorship. But in the future, as AI allows the $-to-labor conversion, every startup could create a streamer that just demos their product 24x7. Or maybe an army of streamers. That’s almost too overtly commercial and the better way to think of it might be that we primarily interact with companions the way we interact with apps today — it’s almost like they are a new UX for a computer, and we will find many ways to insert ads and marketing messages into this new format. Sometimes it’ll be overt, as in a companion that is fully commercial. They sell you hard. But other times it might be more like the futuristic implementation of product placement or a sponsorship, where a very pleasant and engaging companion casually mentions certain products you’re meant to buy.

• Instantaneously traversing the OODA loop. A lot of what marketers do today is to follow a version of John Boyd’s OODA loop — observe, orient, decide, act — which I’m stealing from military strategy but I think applies here. Marketers spend time observing their customer (via survey, analytics, qualitatively, or otherwise). They orient upon that data, creating segmentations and the marketing channels to reach them. They decide what to do, proposing campaigns to reach those segments. And they execute the campaigns. Today this loop happens constantly, with some strategies requiring months (or years!) of planning. Imagine shortening this loop with AI. Imagine shortening it to hours or minutes or instantaneously. In a world of infinite labor, of course you could measure analytics and brand sentiment in real time. Of course you could constantly create new strategies, test them, and execute on them. And some of those strategies might be very large content creation exercises, building new games and TV series to promote something. You’d be able to act in real-time, adjusting as the world changes.

• Authenticity and dilution of design/beauty/more. In a world where everything is perfect and beautiful, the right way to counter-signal might be to be ugly, authentic, and real. Generative AI will every ad creative and video and image perfect. The people in the marketing will be beautiful (because they won’t be real) and the environments their in will be generated. They’ll say all the right things, because they’ve iterated on billions of variations. In a world that’s saturated with that, what will people respond to? I think we get a hint from today’s social media ecosystem where “corpospeak” social media accounts are boring and receive low engagement — yes, you know the ones I’m talking about, where every post is properly capitalized, has well-crafted messages, etc. They’re robotic. In today’s social media landscape, people respond to other people… people who spell incrrectly and have typos occasionallly so you know they are real. And I think in a world of infinite beauty and polish, we will see “Proof of Human” and a desire for authenticity become a major component of market. (And of course AI will learn to spoof that, and then what…)

The convergence of Marketing and Sales
I have a big point about how Marketing changes that I want to bring up, that I’ve briefly touched on above but it deserves its own point. The idea is that Marketing and Sales are soon to converge — and here’s why.

Today, Marketing has huge deficiencies. It broadcasts to people over large public media formats, starting with radio and television, and now the Internet. For the most part the message that it broadcasts is static, and targets segments of potentially millions of people, rather than individuals. We invented marketing, because it's cost-effective. Somewhere, along time ago, someone hung up assign above their shop, and marketing was invented. But that sign says the same thing to everybody that passes by, and thousands of years later, or techniques are very much the same.

With smarter AI-powered conversations, marketing will look more like sales over time. Rather than 1:many broadcast, we will have many 1:1 agents selling people over chat/phone/video and providing a truly personalized pitch. We only have marketing because 1:1 sales for everything is too expensive. But with AI allowing people to convert $ to labor, we will see unique combinations of mass 1:1 sales with brand efforts to give your virtual salesforce air cover. And along with sales, mass personalized landing pages, product experiences, and so on. Everything will be white glove and concierge, rather than mass-produced.

When a marketer kicks off in new campaign, it might be more like spinning up an instance of millions of virtual AI sales people -- or better yet, "sales companions" -- that go out and engage consumers in the exact way they want to be engaged. That might be chat, or they might buy online ads (but each one tailored, 1:1), or send emails. Or call.

These agents might speak every language in the world. They might know every idiom and every way to be persuasive no matter who you are. They might not only relay an initial message but know how to follow up exactly the right way. Maybe it won't resemble selling at all, but instead they'll be your friend, and part of being your friend is I'll make recommendations on where you should go when you travel.

Thus, the next generation of AI-driven marketing might be like scaling up a massive team and having millions or even billions of one-on-one conversations. Perhaps you still want some form of broadcast marketing to help build brand and provide air cover for your sales team, so that there's trust. Or maybe it doesn't matter.

If this sounds silly to you, I ask you to consider how marketing has changed in the past hundred years. We started out in a world with three television channels and marketer simply bought ads against a small set of programs. We've seen an incredible explosion of marketing channels, whether you're talking about the hundreds of television channels, the millions of websites and apps, the world of influencers, esports teams, micro brands, and everything else. The only thing that keeps marketers from being able to cover the entire surface area of marketing channels, and deliver breathtakingly new creative against all these channels, is the cost of planning creating an executing all the campaigns. But imagine this goes to zero -- maybe we'll be able to cover all the surface area, no matter how many, and how complex.

What do you think?
I’ve had great fun writing this essay and many of the ideas were inspired by folks who replied in the original X post here. Appreciate you all for the inspiration.

And if y’all have more thoughts to add, please comment below! Thank you!