COSO's 2013 Internal Control Framework (five components, seventeen principles) and 2017 ERM update provide the globally dominant standard for governance design. COBIT 2019 defines forty governance and management objectives across five domains (EDM, APO, BAI, DSS, MEA). The IIA Three Lines Model (2020) reframes role differentiation from defensive posture to value creation. TOGAF provides architecture governance including architecture boards, compliance reviews, and architecture contracts. SOX (2002) creates legal accountability requirements (Section 302 CEO/CFO certification, Section 404 internal control assessment, Section 802 criminal penalties for document destruction). Despite this maturity, all frameworks share a structural limitation: they prescribe WHAT organizations should do without providing the infrastructure for HOW to capture evidence that it actually occurs in decisions. COSO distinguishes "present and functioning" from "operating effectively" — the most critical distinction in governance — but provides no mechanism for the evidence that bridges them.

Jensen & Meckling (1976) formalized the agency problem in the Journal of Financial Economics (nearly 100,000 citations): when principals delegate to agents, information asymmetry creates space for divergent interests. Agency costs decompose into monitoring costs (principal observes agent), bonding costs (agent signals alignment), and residual loss (irreducible divergence). Eisenhardt (1989) extended the framework in Academy of Management Review: information is a purchasable commodity, and organizations choose between behavior-based and outcome-based governance depending on what's observable. Gailmard (2014) synthesized: no single accountability mechanism eliminates agency loss; effective governance requires combining ex-ante mechanisms (selection, contracting, structural design) with ex-post mechanisms (monitoring, evaluation, enforcement).

The critical gap: traditional agency responses — monitoring, incentive alignment, bonding — reduce information asymmetry at the margins but do not fundamentally restructure the information architecture. Decision lineage represents a structural reduction in information asymmetry: principals can see what agents intended, what authority they invoked, what evidence they considered, and what outcomes they expected.

Davis, Schoorman & Donaldson (1997) proposed in Academy of Management Review that under certain conditions (professional environments, mission-driven organizations, strong-culture contexts), agents function as stewards whose interests naturally align with organizational objectives. Stewardship governance emphasizes empowerment and trust rather than monitoring and incentives. The critical insight for governance infrastructure: most organizations contain both agency and stewardship dynamics simultaneously. Some roles require monitoring (financial controllers); others benefit from empowerment (researchers, clinicians). Governance infrastructure must serve both — providing accountability evidence (agency function) and knowledge preservation (stewardship function) through the same mechanisms.

Bovens (2007) defined accountability in European Law Journal as "a relationship between an actor and a forum, in which the actor has an obligation to explain and to justify his or her conduct, the forum can pose questions and pass judgment, and the actor may face consequences." Three phases: (1) Information — the actor provides factual information about conduct, (2) Explanation/Debate — the forum examines and engages in dialogue about reasoning, (3) Consequences — the forum evaluates and determines outcomes. Bovens (2010) distinguished accountability as virtue (personal quality) from accountability as mechanism (institutional structure). Governance infrastructure is an accountability mechanism — it enables accountability regardless of individual attitude.

Decision lineage directly serves all three phases: decision records provide the factual information (Phase 1), documented intent, authority, and evidence provide the basis for explanation (Phase 2), and documented expected outcomes provide the standard against which actual outcomes can be evaluated (Phase 3).

Schedler (1999) identified accountability as inherently two-dimensional: answerability (obligation to inform and justify) and enforcement (capacity to impose consequences). Answerability without enforcement is merely transparency — organizations report but face no consequences. Enforcement without answerability is authoritarianism. Genuine accountability requires both. Schedler further identifies overlapping accountability-adjacent concepts: surveillance, monitoring, oversight, control, and accountability proper. Decision lineage infrastructure supports the first three (surveillance, monitoring, oversight); integration with governance consequence systems enables the remaining two (control, accountability).

Meyer & Rowan (1977) introduced "ceremonial conformity" in American Journal of Sociology: organizations adopt formal structures as "rational myths" that confer legitimacy regardless of operational effectiveness. Organizations "decouple" formal structures from actual operations — policies exist but aren't followed, committees meet but don't decide. DiMaggio & Powell (1983) explained how this spreads through three isomorphic mechanisms in American Sociological Review: coercive (regulatory pressure), mimetic (uncertainty-driven imitation), and normative (professional standards). Scott (2001) synthesized three institutional pillars: regulative (rules), normative (values), and cultural-cognitive (shared assumptions). For governance infrastructure to achieve institutional stability, it must be supported by all three pillars — legally required, professionally expected, and culturally assumed.

The ceremonial conformity problem is the deepest challenge for governance: organizations can adopt COSO frameworks for legitimacy without operationally implementing them. The resolution is architectural — embedding governance documentation in the governance activity itself so that decoupling becomes structurally difficult. When the documentation is the decision, organizations cannot easily maintain governance on paper without practicing it.

Weill & Ross (2004) studied 250 enterprises at MIT CISR and found that governance is "not about making specific decisions — management does that — but rather determines who systematically makes and contributes to those decisions." Their key empirical finding: high-performing organizations tend toward clear, integrated governance rather than diffuse or anarchic arrangements. The critical variable is not who decides but clarity about who decides. Even imperfect governance that is clearly understood outperforms theoretically superior governance that is ambiguously implemented. Three governance mechanisms: decision structures (who has authority), alignment processes (decisions align with strategy), and communication approaches (decisions are announced and explained).

This finding validates the focus on decision infrastructure: the primary governance problem is not better decision rules but clearer decision documentation — who decided, under what authority, why, and with what expected outcome.

Berle & Means (1932) identified the foundational problem: in large corporations with dispersed shareholders, owners cannot directly control managerial decisions. The Cadbury Report (1992) established corporate governance principles including the "comply or explain" approach — organizations either follow standards or explain deviations. Useem & Zelleke (2006) found that boards increasingly treat delegation of authority "as a careful and self-conscious decision" with annual governance calendars and written protocols. This evolution validates the need for formal authority allocation infrastructure.

The IIA's 2020 Three Lines Model reframes governance from defense to value creation, defining first-line (operations/risk management), second-line (oversight/expertise), and third-line (independent assurance) roles. The governing body delegates authority and defines risk appetite. Six principles govern: governance structures, governing body roles, management roles, third-line roles, third-line independence, and value creation alignment. The critical quote: "Because of internal audit's independence from management, the assurance it provides carries the highest degree of objectivity." This independence principle creates a governance requirement: decision lineage must be accessible to third-line assurance without first/second-line ability to alter records. Immutability is a governance requirement, not merely a technical feature. But role definition without decision documentation means whether roles actually fulfill their functions is unknowable without retrospective investigation.

Busuioc (2021) identified three AI-specific challenges in Public Administration Review: opacity (algorithmic decisions may be unexplainable), diffusion of responsibility (multiple actors share accountability), and regulatory lag (governance frameworks designed for human decision-making). Wieringa (2020) systematically reviewed 242 articles on algorithmic accountability at FAccT, finding lifecycle-based accountability across development, deployment, operation, and evaluation phases. These challenges extend principal-agent information asymmetry: with AI agents, even the principal (deploying organization) may not fully understand the agent's (AI system's) decision process.

Mulgan (2003) identified three dimensions: external constraint (deterrence), retrospective judgment (evaluation), and social interaction (dialogue). He also identified the network accountability challenge — mechanisms designed for hierarchies work poorly in distributed governance. Grant & Keohane (2005) identified seven mechanisms: hierarchical, supervisory, fiscal, legal, market, peer, and public reputational. No single mechanism suffices. The convergence: effective accountability requires infrastructure that supports multiple mechanisms simultaneously — providing the common decision record that each mechanism can access for its specific accountability function.