When Facebook arrived there was a lag. Any company that got onto it early and used it well grew its brand exposure enormously, far beyond what it paid for, and in the moment that felt like a permanent shift in how marketing worked. Over the long term it was an arbitrage: a gap between what attention cost and what it was worth, open only until the rest of the market noticed. As everyone joined, it averaged back out.
AI is the newest of those windows, and most businesses are reading it the way they read Facebook, as the shift itself, when what they are looking at is the gap.
How a window closes, in the platform's own words
In June 2014 Brian Boland of Facebook published Organic Reach on Facebook: Your Questions Answered, for the businesses asking why their Pages were reaching fewer people for free. His answer was competition. Sharing had become easy, people were liking more Pages, and the average person had about 1,500 possible stories waiting each time they opened News Feed, of which it showed roughly 300, ranked by relevance. More content was chasing the same attention, so it was harder for any one story to be seen.
Nobody switched the window off. The rest of the market climbed through it. That is what the end of an arbitrage looks like from the inside: the tactic still works, the platform still works, and the multiplier that made an ordinary operator look brilliant has quietly gone back to one.
Alex
Alex is a composite, not a real person, and I tell the story because there are a lot of people living some version of it.
Alex is genuinely talented at Facebook ads and gets a job with a traditional shoe manufacturer, hired to set up a new direct ecommerce channel on Shopify and run Facebook ads into it. Because Alex is an early mover the impact is multiplied many times over. No other shoe manufacturer is selling direct, so those ads are close to the only brand awareness in the market, and sales take off. Alex gets promoted from the tactic, Facebook, to the channel, ecommerce, and then further up the organisation on the strength of what ecommerce did, into titles like marketing manager and chief revenue officer. Alex got there through work, grit, study and curiosity.
Then everyone else arrives, the multiplier averages out, and the platform lessons and tactics don't really hold up to the lens of time. Alex is left holding a senior title without the formal marketing education to carry it, and a lot of people in that seat are struggling. The platform was the edge, and the edge was rented.
The window minted job titles on its way through as well. Advertising, media buying, ad strategy and the rest are new words for very similar work, and each one sounded like a new discipline while its platform was young.
Play the platforms, don't live in them
Every platform cycle runs faster than the one before, so the stance I use is to play a platform while it is cheap and never move in. Use the window to build trust with people while the platform still allows it, before it narrows to paid awareness, and carry what you build somewhere you control: a website, owned media, a list of customers you can reach directly.
The test for anything you make inside a window is whether it is an asset or a liability. An asset keeps working after the window shuts. A liability needs the window open to be worth anything. Build the first kind, the things that work again and again.
Spread the paid work across more than one channel for the same reason. If an ad is performing badly, optimising harder inside the one platform is unlikely to turn it into a winner. Cherry-pick the best customers from several channels instead of chasing a harder audience deeper into one.
AI is the newest window
AI content is programmatic SEO continued. Search marketing turned into AI by way of word spinners, local SEO, content velocity and programmatic page generation, and I have been using language models since 2021, starting with Jasper and then ChatGPT. Every one of those was cheap output for whoever had it first, and none of them stayed an edge.
The models make the same move, faster. The better they get, the closer every business using them comes to the same answer, and an answer everyone holds is worth nothing. The tool that was supposed to give you an edge gave everyone the same edge, and the collapse of differentiation is what that looks like once the market has caught up. The other way renting goes wrong is quieter: a rented engine does exactly what it is fed, at volume, so one bad input becomes an error at scale.
ChatGPT, Claude, Gemini and whatever ships next are the edge, and you rent them. So are the CRM, the scheduler and the marketing tools. Rent them as interchangeable execution engines, the kind you could replace without losing anything, and expect the window on any one of them to close. A business whose whole method is "we use ChatGPT" has to rebuild its method the day ChatGPT stops being the best option.
What doesn't close
None of what outlasts a tool advantage is built on a particular technology. The System Seed is what the business knows, how it sounds and what is verified. Input quality is the whole game: the real moat is concentrated expertise, not the tools you use but the inputs you feed them, and third-party software on its own is an illusion of advantage. Every competitor can rent the same model this afternoon. Nobody can rent your inputs.
The core is where the truth about the business lives: its customers, products, jobs and the history of every change to them. Models are the edge too, and an owned core is what they read from and write to. Swap the engine and the history stays.
The loop inside the four quadrants is the discipline that runs on both. Human leads, AI executes, human validates, the system deploys, and feedback refines the next round. That sequence does not care which model sits in Q2, the AI execution step, which is exactly why it survives a change of model. Who leads what gives each step an owner, and cascading updates is what the sequence does when something in the business changes and every channel has to follow.
Teams lose whole meetings comparing one tool with another, when three years out, at board level, the choice barely registers. A good team performs with almost any technology. A team without the seed, the core and the discipline performs with the newest one, until the window on it closes too.
The same idea holds in search. On Opportunity + Authority authority is what running this discipline with integrity produces over time.
Where to start
One cycle of the framework on real work is enough to start: decide what you believe and write it down, let the machine draft from it, check every output against what is true, then ship it. Keep what the check taught you in the seed. If you are not sure where your own loop breaks, the diagnostic asks.
Play the window while it is open. Build the things that are still yours when it shuts.