AI-generated outputs can inform human decisions, but output alone does not establish that a user’s contextual Decision Need has been addressed, Decision Quality has improved, or downstream outcomes can be attributed to the system.

Building on the previously published Intent-to-Income™ Version 1.0, this paper develops its design rationale, operationalizes the Reference Decision Architecture through an illustrative end-to-end case, and documents specification revisions produced through an internal formative Structured Design Walkthrough.

The paper makes four connected contributions to the design and evaluation of AI-mediated Decision Support.

The paper uses the Model-to-Decision Gap to describe the design and evaluation problem of connecting AI-mediated outputs and decision-related signals to context-sensitive human Decision Support.

An AI Output may inform a decision, but its existence alone does not establish that the user’s Decision Need was addressed, the support was adequate, Decision Quality improved, or a downstream outcome was caused by the system.

The research object is Intent-to-Income™, HaNonn’s previously published proprietary Reference Decision Architecture for AI-first Decision Intelligence, initially focused on Customer Decisions and Commerce Decisions.

This paper examines the architecture’s theoretical rationale, operationalization, formative refinement, and future validation requirements. It does not present the architecture as newly discovered or empirically validated.

Prior architecture record:
Intent-to-Income™ Version 1.0

The paper draws on established research in Decision Support, human–AI decision-making, information behavior, measurement validity, and Design Science Research.

These foundations inform the requirements used to examine Intent-to-Income™ without treating adjacent concepts or prior research as inventions of the present study.

Intent-to-Income™ does not treat AI Output as a completed decision. It preserves the trace from observed signals and contextual interpretation to Decision Support, Decision Quality, and separately assessed outcomes.

Decision Context operates across the architecture by informing how signals are interpreted, needs are proposed, and support is designed.

Explore the full Intent-to-Income™ Reference Decision Architecture ➞

The paper proposes Evidence-Aware AI Decision Support as an orientation for keeping claims, source provenance, evidence sufficiency, uncertainty, and human-controlled next steps inspectable.

Its purpose is not merely to generate an answer, but to preserve the route from an interpreted Decision Need to the support actually presented.

The study combines conceptual synthesis with a Design Science Research orientation, an illustrative applied case, and an internal criteria-based formative evaluation.

Each source and artifact is classified according to the claims it can support, preventing conceptual, illustrative, and formative evidence from being treated as empirical effectiveness evidence.

The study operationalizes selected parts of Intent-to-Income™ through an illustrative Cattus case, followed by an internal Structured Design Walkthrough that records design findings and accepted specification revisions.

The paper operationalizes selected parts of Intent-to-Income™ through an illustrative cat-food decision scenario on Cattus.

The case combines a synthetic cat-owner profile, a constructed behavioral journey, and identified public product information to produce traceable Decision Support artifacts.

Explore the full End-to-End Case and Evaluation ➞

The case artifacts were examined through an internal, criteria-based Structured Design Walkthrough to assess architecture conformance, traceability, evidence handling, role separation, and preservation of human Decision Authority.

The evaluation was conducted by the architecture designer. Its recorded statuses concern design conformance—not application effectiveness, participant outcomes, or empirical validation.

Key Accepted Revisions

AI Output quality, interface activity, human Decision Quality, and downstream outcomes require different evidence. None should be used automatically as a proxy for another.

Proposed Decision Quality Dimensions

The architecture treats evidence governance and human oversight as design requirements. AI may organize evidence and support evaluation, but final Decision Authority remains with the person.

The current study establishes a theoretical rationale, traceable design artifacts, and a formative revision record. Whether the architecture improves human decisions or business outcomes remains an empirical question.

Proposed Empirical Programme

This paper develops the Model-to-Decision Gap as a bounded framing for distinguishing AI Output, context-sensitive Decision Support, human Decision Quality, and downstream Measurable Outcomes.

It extends the design rationale of the previously published Intent-to-Income™ architecture and shows how an illustrative case and internal formative walkthrough can produce traceable artifacts, design findings, and accepted specification revisions.

The result is a literature-informed rationale and an evaluation-ready design specification for future independent review, participant research, comparative evaluation, and cross-context testing.