The Integral Management Society · GenevaTEGRITY.AI / HOME / 2026-09 · v10.2

Operational Integrity

Tegrity.AI is an applied research, engineering and pre-standardization group of The Integral Management Society, a Swiss non-profit association based in Geneva. It advances AI Integrity Management through research, architecture, the review and development of standards, and technology for the common good.

§ 01

Ecosystem Awareness

In agentic systems, locally justified decisions can coexist with global indeterminacy. Ecosystem Awareness maintains a justified epistemic posture by continuously relating residual uncertainty to mission risk, determination capacity and the value of knowing more.

It preserves and communicates that posture across agent, human and system boundaries, whether the ecosystem is centrally orchestrated or not. For example, human oversight is itself a determination-capacity constraint: the relevant human may be known and authorised, yet unavailable within the useful decision window.

Ecosystem Awareness is a candidate pre-standardization architecture being developed publicly through ITU-T FG-TIDA contributor work to make system-level epistemic state legible and interoperable across independently governed human and agentic components.

Its differential is not uncertainty quantification, observability, human oversight or orchestration by themselves. It combines:

• Scoped epistemic state — what is determined, unresolved, potentially knowable and structurally residual.
• Non-fungible composition — confidence in one domain cannot erase uncertainty in another.
• Risk/capacity requalification — expand, narrow or stop determination according to mission risk, capacity and response horizon.
• Transport-neutral handoff — preserve the minimum scope, uncertainty, residual and dependency semantics needed across boundaries without a common internal reasoning model or central orchestrator.

Some distinctive features:

• Reduces false systemic confidence by preventing local closure, duplicated evidence or confidence in one domain from being laundered into global certainty.
• Improves the risk/resource frontier by directing observation, search, escalation and human attention only where additional determination has material decision value — and stopping when it does not.
• Enables heterogeneous agents, vendors and independently governed systems to interoperate epistemically without requiring a shared uncertainty model, central controller or disclosure of internal reasoning.
• Supports targeted requalification: when assumptions fail, the system can revisit the affected scope rather than rerunning or over-controlling the entire workflow.

Selected reading: EA foundations · general interfaces · Theme #13 (contributor @dakleyer) · EA contribution (contributor @dakleyer) · Theme #16 (contributor @dakleyer)

Ecosystem AwarenessArchitecture notebook · two coordinated views

EA-01Functional kernel and qualification loopsEA-02Integrated enterprise-architecture layers

§ 02

Minimum Sufficient Control Architecture

As an operational output of Ecosystem Awareness, Minimum Sufficient Control Architecture (MSCA) defines the least intervention-intensive configuration that can keep a system within its declared Objective Envelope under the current ecosystem, operating-context and regime conditions.

The Objective Envelope S(t) constrains a joint search over coordination scope (C), operational mechanisms (P) and enabling means (M). Sufficient configurations are then compared by justified burden.

It is not a single global optimum or a fixed minimal stack. Multiple control architectures may be sufficient under the same Objective Envelope, with different allocations of observation, early warning, containment, migration, ecosystem signalling, redundancy, human oversight and feedback. The relevant architecture is the one that remains within the feasible sufficiency region with the lowest justified burden for the current context.

Advantages:

• Resource-proportional control — allocates observation, control and human capacity where they materially support the Objective Envelope, reducing coordination overhead and improving both local and system-wide efficiency.
• Architectural interoperability and mapping — expresses different architectures, resources, objectives and apparently different use cases through a common sufficiency model, enabling comparison and extension without imposing a single stack.

Ecosystem Awareness provides the requalification signal: when ecosystem or regime conditions change, the current MSCA may remain sufficient, require targeted adjustment, or require a different control posture.

Current work: with industry, mapping agentic and pre-agentic systems against MSCA to reconstruct how objectives, context, controls and architectural choices evolve over time, and to test the same control logic across different domains and use cases.

ITU-T FG-AI4SSC (Focus Group on AI for Smart Sustainable Cities and Communities) covers system-level integration of AI-enabled urban technologies — requirements, interfaces, coordination, interoperability and assessment. This is where Minimum Sufficient Control was first submitted as a Focus Group input: read the submitted contribution, FGAI4SSC-I-097 (PDF).

Selected reading: MSCA working paper · minimum-control corpus · extensible mobility case · EA–MSCA interface

Minimum Sufficient ControlArchitecture notebook · three coordinated views

MSCA-01Objective-envelope qualification architectureMSCA-02Material-change requalification sequenceMSCA-03Case-to-validation standards traceability

§ 03

Regime Awareness / Sufficiently Good Early Warning

Regime Awareness is an important component of Operational Integrity. It links Ecosystem Awareness and MSCA by detecting when the operating conditions that justify observation, assessment and control are losing validity — even when the next regime is not known in advance. It is not forecasting: it asks whether present observable behaviour remains sufficiently compatible with the regime under which the current epistemic posture and MSCA are justified.

Because a real system cannot depend on a perfect detector, the pre-standardization object is a Sufficiently Good detector: a model- and domain-agnostic evaluation surface for whether detection is early, reliable and resource-proportionate enough to preserve useful action under the declared Objective Envelope and available response capability. Sufficiency is always evaluated relative to context; no preferred forecasting model, detector family or time-series method is imposed.

Concrete industrial detectors and operational solutions can therefore be compared against the same sufficiency benchmark across different systems, domains and series, rather than against an imaginary perfect detector.

Current work: research review with Universitat de València — QAVA (GIUV2026-019), covering concrete regime-awareness cases and the early-warning / Sufficiently Good detector architecture as part of the programme’s validation and pre-standardization work.

Selected reading: field cases · Good Enough EWS · H1–H3 validation proposal · EA–RA interface

See the full initiative index, with links to the underlying Field Notes series, in Research Initiatives.

Regime AwarenessEngineering notebook · five coordinated views

RA-01Semantic-window context architectureRA-02Warning-sufficiency validation surfaceRA-03Safety-governor posture lifecycleRA-04H1–H3 experimental architectureRA-05Causal selection and re-anchoring

Explanatory notebook

Selected executable examples behind the diagrams

Python 3.11+ · explanatory reference implementation v0.1.0

Four inspectable contract plates backed by dependency-free Python source, demos, nine tests and a SHA-256 manifest. Research prototype; not production safety logic.

Download Python reference (.zip)

§ 04

Pre-standardization roadmap

A compact contribution path from research and field evidence into evaluation, architecture and standards-facing work.

See the full list of research initiatives and their Field Notes in Research Initiatives, and the submitted contribution documents in Research Contributions.

§ 05

Industry Reference Group

The AI Integrity Management Industry Reference Group connects this roadmap with real architectures, products and operating cases. Participants can contribute requirements, implementation evidence, architecture mappings and reference cases while the candidate concepts and interfaces remain open.

§ 06

Last Publications

Browse all publications and Field Notes. The full corpus index is maintained in Structural Awareness Program.