Human leads. AI executes. Human validates. The system deploys. Feedback refines. That is the loop inside the 4-Quadrant Framework in one line, and every part of it is a rule about who leads rather than who is busy. Each of its four steps, Q1 human strategy, Q2 AI execution, Q3 human validation and Q4 deployment, has a non-negotiable answer to that question, and an operation using AI across a whole marketing department needs the answers decided before the machine starts producing, not after.
Job titles are a poor guide to the answers. Marketing titles keep changing while the work underneath them barely moves: advertising became media buying became ad strategy, the customer avatar became the persona and then the ideal customer profile, and the job of understanding who buys and why stayed the same job the whole way through. So the quadrants are assigned here by the kind of work, and whoever holds that work leads it, whatever their business card says.
Q1: the person who knows how the business actually works
Human strategy is led by the people who make decisions inside the business every day, usually the founder or the owner, because the first thing Q1 produces is the System Seed: the expert knowledge, the position, the voice, the verified product facts and the customer logic, written down where a machine and a validator can both read it. Marketing without intimate knowledge of how the business works, or without the power to change the product, has nothing true to report, and it fills the gap with hype that is louder than it is true.
An agency working on promotion alone can describe the business's promise, but it cannot change any of it, which is why in-house marketing that can change the product beats an agency that can only change the ad.
Skip Q1 and everything downstream breaks. The machine in Q2 produces generic output, and the validators in Q3 have nothing to check against.
Q2: the machine, fed by someone who owns the feed
AI executes, and the person who leads Q2 is not writing sentences. Their job is what the machine gets fed and how it gets fed, so that the seed reaches every brief instead of living in a document nobody attaches. The rule worth holding is that output gets rewritten from the source, never edited: when a page is wrong, stiff, or does not sound like the business, the fix goes into the source and the page is produced again.
That rule marks where the Q2 lead's job ends. Someone who quietly corrects sentences downstream is running a second, unrecorded Q1, and the next batch repeats the mistake because the source never heard about it. A correction that changes the seed is a lesson the whole operation keeps. A correction that changes one paragraph is a lesson the operation has to learn again tomorrow.
Q3: whoever holds the seed checks against it
Human validation is led by someone who knows the seed well enough to say no, because the seed is the benchmark: does this number match it, is this a promise the business has made, does this sound like us. In a small operation that is often the Q1 person again, and in a bigger one it has to be someone the Q1 person trusts with the word no. Traffic + Offer calls the habit the Bullshit Police, and Q3 is that habit applied to everything the machine produces. On The Trust Algorithm the same split reads as trust: Q1 captures the expertise, and Q3 guarantees the trustworthiness.
A check at the seed is worth the most, because a fix there lands everywhere, and it sits alongside the person reading what ships rather than replacing them. An operation without a seed has nothing upstream to check at all, and what happens when Q3 is skipped is a single mistake travelling into every asset built from it.
Q4: the system deploys, and specialists run the channels
Deployment is led by the system, the publishing, scheduling and channel machinery, with people running the channels inside it. When those people are hired they sit in the rented column of the four quadrants, each one a specialist who takes full responsibility for one platform or tactic, however the work happens to be chopped up. Hiring specialists sets out the checks, starting with whether what they learn comes back to the business.
The decision about where to grow still belongs to whoever leads Q1. An agency running one channel will recommend more of that channel, and the business might do better adding a channel or lifting conversion, so the channel's owner advises and the seed's owner decides. Channels and tools are the rented edge. The people running them are judged by whether the business keeps what they learn.
Feedback refines: the handoff that breaks first
The first check before hiring anyone into Q4 is whether the learning flows back into the company. If a specialist finds a winning offer or headline on one platform, there has to be a channel for it to be learned, stored and reapplied on the others, so the company stays coherent rather than becoming a pile of ad hoc messages, one per platform. That is the Q4 to Q1 handoff, and it is where the cycle closes: the seed takes the new fact, and a cascading update carries it to every asset that states the old one.
Every handoff in the cycle passes the seed along. Q1 hands it to the machine as the brief, Q2 hands its output to Q3 to be checked against it, Q3 hands only what passed to Q4, and Q4 hands back what the market taught so the seed can change. A handoff with no seed in it is a guess passed from one person to the next, and the machine makes guesses at volume.
One person in every step
At 19 I ran a lawn-mowing round, and I was every step of the loop at once. I decided what I offered, lawn mowing on the front of the flyer and section clearing and one-off jobs on the back, I paid to have 10,000 flyers delivered at a time, and I tested colour flyers against black and white ones. Colour lost to black and white. When a round was big enough I sold it on as a business at 12 times a week's takings. Nobody else was needed to lead anything, but every step was still there, the decision, the production, the check and the delivery, each one done by the same person in a different mode.
A small team can do without most of the Q4 machinery and publish by hand, and it can build the Q2 tooling slowly. It cannot do without Q1 or Q3, because the seed and the check against it are the two things the machine cannot supply for itself. The diagnostic asks, one quadrant at a time, whether anyone is actually leading.