Tenth page in Operation Scarecrow. If you haven't read the argument page and the semantic-memory page yet, start there. Source citations at the foot of this piece.
The semantic-memory page earlier in this section argued a settled fact needs a named steward with real approval authority. That argument has to survive a documented graveyard, not a blank slate: Gartner reports that roughly 80% of data and analytics governance initiatives fail. This page checks the steward proposal against exactly why.
The steward fix is walking into a documented failure pattern
The specific, named pattern is worth borrowing directly: the "Unfunded Mandate," where an organization stands up a governance council that meets quarterly and approves high-level policy, but never actually appoints anyone empowered to execute it day to day. That's a precise, evidenced version of a line this section has already used elsewhere — everyone's responsibility means no one's — except the research shows a second failure mode sitting underneath the first one. Even where a steward role does get named, it often exists on paper without real authority: reporting to the wrong part of the org, excluded from the decisions it's supposed to govern, treated as an expendable documentation scribe rather than someone with actual veto power. Naming the role isn't the fix. It's the first half of a fix that fails constantly without the second half.
Why the knowledge base actually dies
The mechanism behind SharePoint- and wiki-era failure is more specific than "nobody used it." When no one is accountable for keeping content accurate, it goes stale. Once people notice it's stale, they stop trusting it. Once trust is gone, usage drops fast, and the system is functionally dead long before anyone officially retires it. That's a direct, causal validation of the risk this section already flagged for the semantic tier specifically: a governed fact with no active owner doesn't fail loudly. It just quietly stops being true while everything downstream keeps citing it as if it were current.
The same mechanism the section flags for the semantic tier specifically: a governed fact with no active owner doesn't fail loudly, it just quietly stops being true.
The rigorous answer already exists, and it isn't gamification
The industry's default prescription for this problem is thin: KPIs tied to small rewards, leaderboards, points, shout-outs for stewards who complete governance tasks on time. That's a surface-level fix for what's actually a governance-design problem, and a much more rigorous body of work already solved it, decades before AI memory was the thing being governed.
Elinor Ostrom won a Nobel Prize studying exactly this question — how communities successfully sustain a shared resource without it degrading into a tragedy of the commons — and her eight design principles for durable commons governance are specific enough to check this framework's steward proposal against directly: clearly defined boundaries around who's in the community and what's being governed, rules that fit local conditions rather than a one-size policy, participation rights for the people the rules actually affect, monitoring done by parties who are themselves accountable rather than a disconnected compliance function, graduated sanctions rather than all-or-nothing enforcement, cheap and accessible conflict resolution, respect for local governance authority from whoever sits above it, and nested layers of governance once the commons gets large enough that one flat rule stops fitting. A paper already applies this directly to AI agent memory architecture, finding that a well-designed system instantiates several of these principles at once: defined boundaries between private and shared memory, monitoring through provenance and version lineage, graduated requirements for ratification, and an explicit governance process as final arbiter rather than leaving conflict resolution to whoever argues hardest.
That's a sharper diagnosis of what's actually missing from the industry's leaderboard-and-shoutout approach: recognition programs address motivation, but the documented failure pattern is an authority and enforcement problem, not a motivation problem. A steward who's genuinely excluded from decisions doesn't need a badge. They need the reporting line, the veto, and a real consequence attached to ignoring their call — Ostrom's principles about monitoring, graduated sanctions, and respected local authority, not the principle of making it fun.
What this changes about the fix already proposed
The org-chart argument made earlier in this section — semantic memory needs a named steward with real authority to approve or reject a firmwide fact — turns out to be necessary but incomplete as stated. Naming the role is exactly the Unfunded Mandate pattern if it isn't paired with a second, previously unstated requirement: the steward's authority has to be structurally real, meaning an independent reporting line or a protected appeal path, actual veto power that's been tested and held at least once, and a graduated consequence for overriding it without going through the process. Without that, the fix this section has been recommending is indistinguishable from the governance council that meets quarterly and changes nothing — better branded, same failure mode.
There's also a direct answer here to a conflict-of-interest concern worth raising plainly: what happens when the steward and the person whose fact just got overruled report to the same manager. Ostrom's principle of cheap, accessible conflict resolution, and her insistence on monitors being accountable to the community rather than to whoever's easiest to please, both point the same direction: the appeal path has to exist independently of the local reporting structure, or the steward's authority only holds as long as it never actually costs the overruled person anything.
"The Unicorn the Org Chart Wants" arrives at the same authority-not-motivation diagnosis independently, from an org-design angle rather than a governance-research one: it argues a verification function without the authority to halt a work stream is "documentation of drift, not a remedy for it" — the Unfunded Mandate pattern in different words, reached by reasoning about agent-fleet management rather than by reading Gartner or Ostrom. It also names the shared operating context agents draw on as "the product backplane under another name... an org asset, not a per-PM hobby" — this whole research effort's subject matter surfacing independently in an essay that wasn't written about it.
The question professionalizing doesn't answer
One question this page can't settle: whether stewardship can be a genuine promotion track rather than a lateral, unglamorous assignment people rotate out of as soon as they can. Current market data shows the role professionalizing — dedicated titles, salary bands, a defined career path in some organizations — but doesn't show whether that professionalization actually holds people in the role, or just makes the role easier to backfill once someone leaves it. Those are different outcomes, and this page doesn't have evidence to say which one is actually happening.
Sources. Governance failure rate and the Unfunded Mandate pattern: "Data Governance Is Failing — Here's Why," CDO Magazine, cdomagazine.tech, and "Why Data Governance Fails in Many Organizations: The Accountability Crisis and Capability Gaps," Dataversity, dataversity.net. Knowledge-base and wiki failure mechanics: "8 Reasons Why Knowledge Management Fails," Bloomfire, bloomfire.com, and "Why Knowledge Management Fails in Most Companies (And How to Fix It)," agilityportal.io. Ostrom's commons-governance principles: "5. Elinor Ostrom & 8 Rules for Managing the Commons," Heinrich-Böll-Stiftung, tn.boell.org, and "A Practical Framework for Applying Ostrom's Principles to Data Commons Governance," Mozilla Foundation, mozillafoundation.org. Ostrom's principles applied directly to AI agent memory architecture: "Governed Collaborative Memory as Artificial Selection in LLM-Based Multi-Agent Systems," arxiv.org/pdf/2605.04264. Steward career-path and incentive-design data: "6 Strategic Steps to Enable Data Stewards," Atlan, atlan.com, and "What Is Data Stewardship? Framework, Roles & Tools 2026," OvalEdge, ovaledge.com.