The most consequential strategy document at my company is four lines of code:
export interface MessagingPillar {
pillar: string
description: string
}That's it. That's a messaging pillar (the thing agencies bill five figures to put in a deck) reduced to a name and a description, sitting alongside the rest of our brand definitions where every product we build can reach it.
I know what it looks like. It looks like the opposite of strategy: strategy flattened into a database field, the soul run through a form validator. Every operator instinct says strategy lives in the deck: the narrative arc, the positioning triangle, the slide where the pillars get their own icons.
Here's what changed my mind. In July I ran a search for that field across everything we've built. Ninety-seven files reference it. Fifty-three of them are real users of the field, not automated checks: scoring systems, content engines, report generators, product surfaces, the feed we hand to outside AI agents. If you deleted that four-line definition, half a dozen product surfaces would lose their strategy layer.
| The four-line field | The strategy deck from your last agency engagement | |
|---|---|---|
| Delete it. What breaks? | Half a dozen product surfaces would lose their strategy layer. | Nothing breaks. |
That's the tell. Strategy that nothing depends on isn't strategy — it's decoration.
A field is a promise with teeth
Every brand I've operated had strategy documents. Five brands' worth of decks, brand books, positioning one-pagers. I can tell you from experience what happens to them: they get presented once, admired, and then every downstream decision — every brief, every packaging tweak — gets made from memory and vibes, drifting a little further from the deck each week. The deck can't object. It's a PDF. It has no way of knowing it's being ignored.
Strategy that nothing depends on isn't strategy — it's decoration.
A structured field is a different kind of document. It's a promise the rest of the system holds you to. When a pillar is a field, vagueness stops being an aesthetic problem and becomes a mechanical one: the tools downstream either receive something usable or they receive a gap, and the system treats those differently. There's no third state where everyone nods at the slide and quietly reverts to instinct.
The sharpest version of this in our stack is in a voice-analysis template, where content pillars come with a receipts requirement built into the field definition itself. The pillar section is described as "What they return to. Content pillars with their specific angle," and each slot demands a pillar name, a signature angle, and evidence pieces: "Up to 3 piece titles/URLs from the corpus where this pillar shows up."
Read that last field again. The template will not accept a pillar as pure assertion. A pillar is a claim about a pattern, and the field forces the claim to arrive holding examples of the pattern. No agency deck I ever paid for held itself to that standard. There's nothing stopping you from imposing it on your own strategy today: for every pillar in your current deck, ask for three real pieces of your existing content where that pillar actually shows up. A pillar that can't produce its evidence is a wish.
What actually depends on it
The census is the argument, so let me walk a few of the things that depend on the field.
A brand with no pillars gets no polite note; its score visibly bleeds, everywhere the score is shown.
When a report section is weak, you can trace it to the exact strategy field that was thin.
A reviewer reading that line doesn't have to argue about whether the draft is on-strategy.
The customer-facing artifact and the machine-facing field are the same object.
Our brand preview takes a homepage and reads it back as a first brand model: archetype, audience tribes, positioning wedge, messaging pillars.
The strategy field is literally part of a contract someone pays for.
The scorer prices strategy gaps. Our brand-completeness scorer keeps its judgment in one place a person can edit by hand; the note at the top says the mechanics live elsewhere and "this file holds the JUDGMENT." In that file, messaging pillars sit in Tier 1 alongside mission and positioning, and the tiering rationale is blunt: "A missing Tier-1 field misranks the whole funnel, so it costs most." A brand with no pillars gets no polite note; its score visibly bleeds, everywhere the score is shown.
Content engines declare their diet. Each analysis lens in our report engine opens with a declared list of the strategy fields it consumes (messaging pillars, tribes, pain points, the positioning wedge), a plain statement of which strategic inputs feed which output. When a report section is weak, you can trace it to the exact strategy field that was thin. Try doing that with a deck.
You can watch a pillar get used on one of our product pages today. Our writing product's demo of its review step passes every draft through three judges before a person sees it, and the brand judge's finding is a single line that names its evidence: no banned words, then the pillars the draft actually hit, by name. A reviewer reading that line doesn't have to argue about whether the draft is on-strategy. The strategy's own nouns are sitting in the verdict.
A pillar that can't produce its evidence is a wish.
Product surfaces render it. The strategy summary a customer sees builds its "Messaging Architecture" section directly off the list of pillars — pillar and description, straight from the field. The customer-facing artifact and the machine-facing field are the same object. There is no second, prettier version that drifts.
The same field is there the moment a brand first meets us. Our brand preview takes a homepage and reads it back as a first brand model: archetype, audience tribes, positioning wedge, messaging pillars. The brand-read report has a section headed "Messaging pillars", and that section is drawn only when the read found pillars to put in it. No pillars, no heading. Not a placeholder, not three generic bullets to fill the space. The absence of strategy renders as absence, which is the first time a lot of operators have seen their gap drawn to scale.
The reports grade it. Our reports product scores brand coherence on four axes, and one of them is message pull-through, which asks a single question: do the message pillars trace back to the positioning? The grade follows mechanically from which brand fields are present, so the same record always yields the same scorecard, and a signal that is wholly absent renders unscored, never as a guessed number. The tier above the grade is the operating documents the grade says you're missing, the positioning, the messaging, the voice, written from the extracted record rather than a template.
Outside machines get it too. The version of the brand we expose to external AI agents includes the pillars under a label that just says "strategic layer," and the paid version carries a warning that "Changing this changes what customers bought." The strategy field is literally part of a contract someone pays for. Decoration doesn't carry contract language.
The version those agents read is a fixed list of nineteen fields, identity, voice, positioning, and the pillars are on the list. Everything else is left out simply by not being named. If a field gets renamed upstream, that version refuses to build instead of leaking quietly. Rename a pillar in a deck and nothing notices. Rename the field and the build refuses until every reader agrees on the new name.
The compile step
The part I find most satisfying is what happens at the moment strategy gets decided, because it's the closest thing I've seen in brand work to a compile step: a human decision turned, mechanically and without interpretation, into every field the products run on.
In our launch-planning tool, strategy is generated as a three-direction pack — three genuinely different strategic directions, each with its own positioning, pillars, personality, values. A human operator reads them and confirms one, by name. Then a part of the system built for exactly this takes over. Its own description calls a confirmed strategy entry "the brand-DNA heart", and says it "MECHANICALLY collapses" the chosen direction into the flat brand fields every downstream product reads. The collapse is a verbatim re-map: in the system's own words, "it re-homes the direction's own words, and unmapped fields stay honest gaps."
Three genuinely different strategic directions, each with its own positioning, pillars, personality, values.
A human operator reads them and confirms one, by name.
The machine compiles the human's decision into every field the rest of the system executes against, inventing nothing along the way.
An absent description is a gap, not a fabrication.
Notice the division of labor, because it's the whole philosophy in one handoff. The machine drafts directions. The human makes the strategic call. That's the one step with no automated path around it. And then the machine compiles the human's decision into every field the rest of the system executes against, inventing nothing along the way. Even the gaps are principled: the pack's pillars arrive as bare phrases, so their description fields are left empty, with a note explaining that "an absent description is a gap, not a fabrication." Strategy in, executable brand out. AI generates, humans decide, and the deciding compiles.
The obvious objection is that this makes strategy whatever the machine read off your homepage. It doesn't, and the reason is the same handoff. Nothing collapses until a person confirms a direction, and the products that seed from the record keep that posture downstream. Our image studio seeds a brand's visual style from the record across seven dials (medium, grade, finish, lens, time, lighting, mood), and every dial stays yours: confirm the seed, override it, or clear it. An empty dial means clean photoreal, never a hidden default. The machine proposes from the field, the human owns the field, and a blank stays visibly blank.
What it costs
Honest ledger, because this approach is not free.
Compiling doesn't make strategy good. A field checks shape, not judgment. Mediocre pillars flow through the same ninety-seven files as brilliant ones, and the system will faithfully execute mediocrity at scale, which is arguably worse than a forgotten deck, because the deck at least had the decency to be ignored. The structure removes drift and vagueness as failure modes; it does nothing about taste. That stays a human job, which is exactly where we want it, but nobody should mistake the plumbing for the wine.
The structure is opinionated, and reality isn't. A structured strategy can only hold what the structure anticipated. The interesting brands are interesting partly in ways no field captures, and evolving the structure is real work: a new field means checking every place that reads the old ones, because ninety-seven files is leverage in both directions.
Gaps become visible, and visible gaps are uncomfortable. When strategy was a deck, a brand with no real positioning could hide behind good typography. When strategy is a scored, structured input, the absence shows up as a priced gap in the top tier: on dashboards, in reports, in the funnel math. We think that discomfort is the product working. Not every customer enjoys the X-ray.
The decoration test
The operator takeaway is one sentence: if your strategy can't be consumed by your tools, it's decoration.
When strategy was a deck, a brand with no real positioning could hide behind good typography.
If yes: your strategy is load-bearing, and every dollar you spent on it is compounding.
If no: what you have is a PDF with a strategy's picture on it, and every brief and campaign since has been quietly compiled against whatever was in someone's head instead.
We chose to make ours four lines of code with ninety-seven dependents. It's the least glamorous strategy document I've ever owned, and the only one that ever fought back when we tried to ignore it.
If your strategy can't be consumed by your tools, it's decoration.
See how a brand gets gradedDoesn't this just make strategy whatever a machine read off your homepage?
Nothing collapses until a person confirms a direction, and the products that seed from the record keep that posture downstream. The machine proposes from the field, the human owns the field, and a blank stays visibly blank.
Does putting strategy in a field make the strategy any good?
A field checks shape, not judgment. The structure removes drift and vagueness as failure modes; it does nothing about taste. That stays a human job, which is exactly where we want it.
What happens to a brand that has no pillars at all?
No pillars, no heading. Not a placeholder, not three generic bullets to fill the space. The absence of strategy renders as absence, which is the first time a lot of operators have seen their gap drawn to scale.