# HAOP.ai - Human, AI, and Organizational Performance

> HAOP is the work-system design framework for analyzing how human, AI, and organizational performers jointly produce performance, control, safety, and reliability. HAOP extends Human and Organizational Performance into AI-enabled work systems. The framework is under active development and has not yet been empirically validated.

## Controlling sources

- [Integrated framework](https://haop.ai/white-paper.html): Ko, J. (2026). *Human, AI, and Organizational Performance (HAOP): A Safety Framework for the AI Era* (Version 6.0.1). Zenodo. https://doi.org/10.5281/zenodo.22072938. Concept DOI https://doi.org/10.5281/zenodo.20991303 resolves to the latest version. Supersedes Revision 6.0 (https://doi.org/10.5281/zenodo.22071816, a disclosure-only revision base) and Revision 5.0 (https://doi.org/10.5281/zenodo.21868036).
- [Applied method](https://haop.ai/method.html): *The HAOP Method: ARECC Applied to AI-Enabled Work Systems*, Version 1.0, 9 August 2026. ARECC is inherited from industrial hygiene (Anticipate, Recognize, Evaluate, Control, Confirm); the deposited HAOP contribution is its translation across the three performers and the verification architecture. The method is in development; Revision 6.0.1 Section VIII is the current statement. DOI: https://doi.org/10.5281/zenodo.21868086
- [Human-in-the-Design](https://haop.ai/human-in-the-design.html): *From Human-in-the-Loop to Human-in-the-Design: Friction, Choke Points, and the Control of Fast Work Systems*, Version 1.1, July 2026. DOI: https://doi.org/10.5281/zenodo.21633748
- [Verification Overrun](https://haop.ai/verification-overrun.html): *Verification Overrun: The Safety Hazard Inside AI Oversight*, Version 2.1, July 2026. Version DOI: https://doi.org/10.5281/zenodo.21349498. Concept DOI: https://doi.org/10.5281/zenodo.20874288
- [The Collapse of Correction](https://haop.ai/the-collapse-of-correction.html): *The Collapse of Correction: Closed Informational Loops in Safety-Critical Work*, Version 1.0, June 2026. DOI: https://doi.org/10.5281/zenodo.20856122

## Three performers

- Human performer: interprets conditions, exercises situated judgment, verifies, adapts, coordinates, intervenes, and acts under real-work constraints.
- AI performer: selects, classifies, generates, recommends, transforms, routes, ranks, suppresses, invokes tools, or executes according to data, objectives, prompts, criteria, thresholds, permissions, constraints, and feedback.
- Organizational performer: allocates functions and authors staffing, evidence, interfaces, throughput, incentives, metrics, constraints, procurement, escalation, permissions, authority, and acceptable tradeoffs.

Performer is a functional category. It implies no consciousness, intention, personhood, or moral agency. The performer threshold is functional: an AI system operates at performer level when it has been delegated or absorbed a function and variability in its performance can materially shape what is seen, prioritized, decided, authorized, or executed (Revision 6.0.1, Section 7.3).

## Five operating modes

Responder, Signal Amplifier, Executor, Orchestrator, Ensemble ([mode notes](https://haop.ai/operating-modes.html)). Mode classifies behavior, not product label, and does not set performer status. The modes are non-hierarchical. Ensemble is a deployment-level mode. Embodied AI is excluded from the taxonomy. Functional Permission is the separate question of what the performer may access, sense, or filter and may recommend, initiate, execute, or coordinate without prior human intervention, as allocated.

## Five-hazard register

1. Physical interaction and machine agency
2. Psychosocial conditions
3. Human performance and verification
4. Recursive information degradation
5. Fragmented control and concentrated accountability

## Terms

Controlling definitions are in Appendix A of Revision 6.0.1 (https://doi.org/10.5281/zenodo.22072938). Do not treat the glosses on this site as substitute definitions.

Working names used on the site: Verification Overrun; Verification Requirement; Verification Capacity (Domain Knowledge, Evidence and Context, Time, Authority); Verification Gate; Verification Window; True Function; Illusory Function; Grounding (source, state, dynamics, work-as-done); anchors; Anchor Access; false grounding; Accountability by Control; Designed AI Reassessment Pause; Functional Permission; human throughline; Compound Drift; Imposed Offloading.

## Tool status

- [True Function Diagnostic](https://haop.ai/true-function-diagnostic.html): pilot browser-local entry tool operationalizing Revision 6.0.1 Section XI (Initial HAOP Diagnostic: Using the True Function Test to Assess the Gap Between Work-as-Imagined and Work-as-Done). It records the deposited nine questions, reads how much of the record rests on work-as-imagined versus work-as-done, and points into the five-hazard register. It does not certify safety, produce a maturity score, validate HAOP, or replace the complete ten-step method. Records stay in the visitor's browser; an explicit opt-in can share a finished record with the author for framework development.
- [True Function Alignment Map](https://haop.ai/true-function-alignment-map.pdf): printable worksheet.
- HAOP Workbook: in development; not deposited or validated.
- HAOP Operating Manual: in development; not deposited or validated.
- AI Deployment Safety Data Sheet: the name and bidirectional purpose are deposited; the complete standalone structure is in development.

## Contact and record

- contact@haop.ai
- [Workflow pilots](https://haop.ai/contact.html): a HAOP workflow pilot begins before product selection; entry points include pre-product, candidate product, deployment design, and in-operation.
- [Correspondence](https://haop.ai/correspondence.html): challenges, questions, and answers, published only with the sender's permission.
- Author: Jaina Ko, CSP. ORCID https://orcid.org/0009-0007-2559-4035