AI is ready.The human layer isn't.
Every AI governance framework depends on human oversight. WHEEL measures whether that oversight can actually hold. It diagnoses the adoption, trust, workflow, and accountability constraints that policies assume but organizations rarely test.
A diagnostic for the human layer of AI.
Most AI initiatives stall not because the technology fails, but because the human layer around it was never measured. WHEEL is a constraint-based diagnostic that identifies the single factor most likely to break AI adoption, weaken oversight, or turn governance into theater, before it compounds.
The binding constraint
Not ten recommendations. One constraint. WHEEL applies Goldratt's Theory of Constraints to organizational AI readiness. It finds the single ring governing your system's output. Fix it, and the system accelerates. Fix everything else first, and nothing changes.
Eight dimensions
WHEEL scores organizational AI readiness across eight dimensions, from identity and psychological safety at the organizational layer, through workflow and environment at the operational layer, to leadership, learning, and measurement at the root cause layer. Three tiers. One binding constraint.
A constraint map
Not a maturity model. Not a survey. A scored diagnostic that shows which ring is governing your entire system's output, paired with a prioritized remediation sequence that moves in weeks, not quarters.
Human governance for autonomous agentic systems.
Once AI systems are running autonomously, making decisions, executing workflows, and operating at machine speed, the question is no longer whether governance exists on paper. The question is whether the humans assigned to execute that governance can actually do so under real operating conditions.
WHEEL OS is an eight-ring diagnostic framework that surfaces the gap between governance architecture that exists in policy and governance capacity that functions under the conditions the agentic environment actually creates.
It is the only published governance framework that diagnoses whether the humans governing deployed AI agents have the authority, capacity, and accountability structures to govern at machine speed.
Applied to EU AI Act Articles 14 and 26 compliance in published research. Available now for general application.
WHEEL synthesizes four established scientific traditions into a single constraint-based diagnostic architecture. The foundation traditions are public. The diagnostic methodology is proprietary.
Twelve working papers. One body of work.
Each paper is a sector-specific or methodological application of the same underlying architecture. They are not separate projects.
The Accountability Layer: A Constraint-Based Diagnostic Methodology for Enterprise AI
Names and formalizes the constraint-based diagnostic methodology operating across the full portfolio. The methodological spine of the portfolio.
The Capacity Ledger: A Measurable Specification for Capacity-Degradation Risk in Human-AI Collaboration
Formalizes capacity-degradation risk, the measurable risk that repeated AI collaboration erodes the human judgment the collaboration still depends on. Its central result is that observed output quality can decouple from retained human capacity by design, which makes the most obvious performance measure an actively misleading one. Specifies a capacity ledger, a capacity-dynamics model, and a two-leg instrument pairing an offload-ratio estimator with unaided probes. Offered as a reference specification, not as field validation.
Measurable Specifications for the Supervisory Layer in AI-Governed Critical Infrastructure
Human oversight is a control only if it can be performed under the conditions in which the system actually operates. Specifies a cross-sector measurement architecture for determining whether assigned supervision is performable, built on four failure conditions: alert fatigue, automation bias, authorization by decision class, and override and intervention design under load. Each is defined in terms that hold across sectors, because the conditions are properties of human cognition rather than of any one technology. Worked through four critical-infrastructure sectors and mapped onto the NIST AI Risk Management Framework.
Who Governs the Governors? Diagnosing the Human Capacity Gap in EU AI Act Compliance
Articles 14 and 26 of the EU AI Act require human oversight of high-risk AI systems. This paper argues that governance frameworks are mature at documenting oversight structures and immature at measuring whether the humans assigned to oversight can execute it under real operating conditions. Introduces WHEEL OS as a diagnostic scaffold for that gap.
The Token Governance Trap: A Constraint-Based Diagnostic Architecture for Organizational Readiness and Value Attribution Under Goodhart's Law
Applies the WHEEL constraint-based architecture to enterprise AI token economics. Names the Token Governance Readiness Model (TGRM) and a derivative diagnostic index. Addresses agentic AI economics, parallel-agent token consumption, and the Goodhart's Law failure mode in AI value measurement.
The Heart Preservation Problem: A Design Standard for Preserving Human Capacity in AI Collaboration
AI systems can succeed at the task while weakening the human responsible for it, a failure mode current evaluation does not capture. Argues that AI evaluation must measure what repeated use leaves the human able to do over time, and defines Heart as an operational design construct of seven capacities: agency, discernment, authorship, moral attention, humility, courage, and truthful self-relation. Reframes human in the loop as insufficient unless the capacity required to participate in the loop is itself preserved. Proposes seven design principles, a baseline and withdrawal evaluation, and a Capacity-Degradation Risk Report. Conceptual and hypothesis-generating.
Human Governance as a Distinct Layer of Autonomous Agentic Systems: A Constraint-Based Diagnostic Architecture
Governance architecture and governance capacity are not the same thing. Frameworks specify what oversight should include; they do not establish whether the humans assigned to it have the authority, cognitive bandwidth, accountability clarity, intervention pathways, and feedback control to execute it under agentic operating conditions. Proposes WHEEL OS across eight rings, with governance capacity bounded by the weakest ring. Specifies what validation would require and the conditions under which the architecture would be revised or rejected.
The Observation Layer: Why the Governance, Risk, and Compliance Platform Category Cannot Produce the Decision Infrastructure That Agentic Systems Require
Documentary analysis of how the five largest enterprise GRC platform vendors repositioned their product lines around AI governance between November 2025 and April 2026. Finds the repositioning substantive in marketing terms and minimal in architectural terms: every vendor's AI governance product is an extension of the observation and workflow automation infrastructure it already sold.
Multiplicative Models of Human AI Collaboration: Two Equations for Assessing Collaboration Quality and Human Legacy
Proposes that the binding constraint on human-AI collaboration may be the human operating the system rather than the technology. Presents two multiplicative equations, the Collaboration Equation and the Legacy Equation, whose shared structural property is that any variable approaching zero collapses the entire output. Higher AI capability cannot, by this architecture, compensate for absent Heart.
The Foundational Three
The first three papers establish the diagnostic architecture: why AI adoption fails at the human layer (SSRN 6446445), the three non-substitutable conditions for sustained organizational behavior change (SSRN 6479924), and WHEEL OS as a human governance diagnostic framework for autonomous agentic systems (SSRN 6545198).
Full portfolio: twelve working papers available at SSRN Author Page →
I have spent 40 years watching organizations buy technology they could not absorb. The pattern is not new. The stakes are.
WHEEL is a diagnostic instrument. It finds the human constraint most likely to break your AI initiative before it compounds. I built it, I published the methodology, and I deploy it with organizations that are serious about the problem.
The methodology is grounded in four established scientific traditions: Goldratt's Theory of Constraints, Edmondson's psychological safety research, Tajfel and Turner's Social Identity Theory, and Lee and Parasuraman's work on trust in automation. The diagnostic architecture is proprietary and is not vibe-coded theater. It is based on scientific foundations that are established, published, and independently verifiable.
If that is the problem you are solving, let's talk.
Want to know whether your human layer will hold?
WHEEL identifies the constraint most likely to break AI adoption, weaken oversight, or turn governance into theater. TokenGap applies the same diagnostic logic to enterprise AI token economics.
Substantive inquiries welcome.
If the work is relevant to what you are building, researching, or governing, I am interested in the conversation. No forms. No calendaring links. Direct contact only.
Appropriate inquiries include: methodology discussion, research collaboration, EU AI Act compliance context, NIST engagement, and partner or institutional conversations. Sales solicitations and unsolicited pitches are not appropriate uses of this contact.