Joseph Byrum is an accomplished executive leader, innovator and cross-domain strategist with a track record across multiple industries.
getty
Every year, companies invest in content teams, SEO agencies and PR firms. Yet AI still returns a competitor's name in search results. This isn't because any one function failed, but because no one was managing the system they collectively created.
Russell Ackoff, W. Edwards Deming and Toyota's production engineers diagnosed this problem before digital marketing existed. Each solved a different component of the same failure. Digital marketing inherited all three.
The Local Optimum Trap
Ackoff spent decades studying why some well-run organizations produced mediocre results. In a 1999 essay, he put this first among the obvious propositions he had found to be false: improving the parts of a system separately does not necessarily improve the whole. It can make the whole worse. The performance of a system is not the sum of its parts. It is the product of their interactions.
Digital marketing often runs this failure at scale. Content strategy optimizes for engagement, SEO optimizes for ranking, PR optimizes for mentions and schema markup is deployed where convenient. Each discipline is accountable for its own metrics and reports success on its own terms. The result can be a company that AI fails to recognize, recognizes incorrectly or resolves in favor of a competitor.
The practitioner isn't necessarily failing for lack of skill. The problem is that no one has modeled how the parts interact.
Managing What You Can't See
Knowing the architecture is broken doesn't fix it. Deming's contribution was the next layer: once you accept that the system is what matters, you need a formal model of it; not to eliminate variation, but to make it visible before it compounds into losses.
In the engagements I've led, the pattern holds. Companies tracking content performance, SEO rankings and earned mentions as separate dashboards are reacting to outputs. None has a model of the system producing those outputs or can calculate whether their combined investment clears the threshold required to hold an AI authority position. A Fortune 500 CMO survey found that just 16% of brands systematically track AI search performance at all. They are measuring the parts while the system producing those results remains invisible.
In one marketing-data acquisition I advised, the combined company surfaced in roughly half of the AI queries that mattered to its market. That number did not move for nine months after close. Every dashboard was green. Content was shipping, rankings held and mentions accrued. Nothing in the reporting registered that the merged entity had no resolved identity at all.
Without a formal model, management is reactive. You find out something went wrong after it already cost you a board seat, a partnership or a deal.
The Partial Implementation Trap
The Toyota Production System (TPS) worked when organizations treated it as an integrated operating system rather than a collection of tools. Where companies selectively adopted its techniques without the underlying management practices and culture, the results were usually limited or unsustainable.
Organizations often adopt the parts that are easy to understand and skip the parts that require systemic discipline: kanban without jidoka. This shows up in visual signals without the machine-level check that catches a defect before it moves downstream and continuous improvement language without the stop-the-line authority that makes it operational. They get local improvements and mistake them for the system.
Digital marketing's adoption of entity signals follows the same arc. Companies deploy structured data solutions on their homepage. They establish a graph record, publish in authoritative outlets and skip coordinated signal seeding. They skip monitoring how AI systems are resolving their entity across platforms. They skip the formal model.
Partial implementation isn't a version of the system. It is a collection of practices without the system that makes them work together.
The Governing Condition
The reason a fully implemented system is achievable (rather than aspirational as TPS was for most of its adopters) is that the governing condition can be stated precisely. It is a rate condition: the rate at which an organization constructs machine-readable authority signals must exceed the rate at which those signals decay and are displaced by competitors. When that condition isn't met, the AI resolves a competitor. The gap that follows shows up in revenue, not in impressions.
In the acquisition I described, coordinated implementation moved that number from roughly half to 72% in four months. In my experience, the comparable timeline runs closer to 18. The difference was not effort. It was that the whole was modeled before any part was touched.
This is exactly what the discipline Deming built around statistical control did for manufacturing. It didn't eliminate variation; it made it visible, continuously and against a formal standard. A scored diagnostic against a baseline produces three things management can act on: a current position, a rate of change and a measurable gap to the threshold. That turns management from reaction into anticipation so the system can be steered.
Digital marketing inherited all three failure modes simultaneously. The content era optimized parts, the agency model federated accountability and the acronym proliferation (SEO, GEO, AEO, LLMO, etc.) renamed the surface without examining the substrate.
The organizations that have already made the transition aren't running a better version of what came before. They're running a discipline I've coined as "entity engineering," a formally modeled, fully implemented system in which every signal is coordinated, every threshold is measured and the governing condition is a number on a dashboard, not a judgment call in a conference room.
The entry point is a diagnostic, not a strategy session or a content audit. It's a scored baseline that produces a position, a rate and a gap—the same output Deming argued manufacturing couldn't improve without. That number tells an executive exactly where they stand relative to the threshold that produces competitive AI authority. What they do with it is their decision.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?