Engineering the Conditions of Agency: A Comparative Study of Field-Native and Conventional AI Agent Development

Enterprise AI agents are increasingly embedded in operational systems that transform records, coordinate decisions, invoke services, and influence institutional claims about completed work. The engineering problem extends beyond individual model capabilities: an agent configuration acts within changing conditions of authorization, evidence, dependencies, timing, and organizational responsibility. Yet conventional software engineering already provides substantial methods for describing and governing those conditions. A scientifically credible Field-Native proposal cannot claim novelty merely by naming relationships, introducing contracts, or integrating existing controls. This conceptual and methodological paper develops a decision-centered comparison of strong conventional engineering (B), integration-complete conventional engineering (B+), and Field-Native Engineering (C). It defines Conditions of Agency as a bounded, mission-relative architectural perspective and specifies architectural requalification episodes as the proposed unit of comparative analysis. The design distinguishes parity of information and resources from differences in semantic representation, separates comparative performance of methodological packages from causal attribution to individual mechanisms, and connects architectural decisions to downstream outcomes and lifecycle economics. It describes a prospective evaluation program with independently constructed cases, negative controls, explicit equivalence margins, and differentiated outcomes: incremental benefit, integration-only improvement, conventional sufficiency, negative net value, and inconclusive evidence. The paper does not report agent experiments or empirical Field-specific advantages. Its contribution is a falsifiable methodology for investigating where Field-Native practices could add explanatory, engineering, or economic value and where their claims should be reduced to competent conventional architecture.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-16
DOI
https://doi.org/10.5281/zenodo.24293408
Primary Topic
Artificial Intelligence Applications
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Engineering the Conditions of Agency: A Comparative Study of Field-Native and Conventional AI Agent Development

Rogério Figurelli
3 links
Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence Applications
preprint

Engineering the Conditions of Agency: A Comparative Study of Field-Native and Conventional AI Agent Development

Rogério Figurelli
preprint en

Abstract

Enterprise AI agents are increasingly embedded in operational systems that transform records, coordinate decisions, invoke services, and influence institutional claims about completed work. The engineering problem extends beyond individual model capabilities: an agent configuration acts within changing conditions of authorization, evidence, dependencies, timing, and organizational responsibility. Yet conventional software engineering already provides substantial methods for describing and governing those conditions. A scientifically credible Field-Native proposal cannot claim novelty merely by naming relationships, introducing contracts, or integrating existing controls. This conceptual and methodological paper develops a decision-centered comparison of strong conventional engineering (B), integration-complete conventional engineering (B+), and Field-Native Engineering (C). It defines Conditions of Agency as a bounded, mission-relative architectural perspective and specifies architectural requalification episodes as the proposed unit of comparative analysis. The design distinguishes parity of information and resources from differences in semantic representation, separates comparative performance of methodological packages from causal attribution to individual mechanisms, and connects architectural decisions to downstream outcomes and lifecycle economics. It describes a prospective evaluation program with independently constructed cases, negative controls, explicit equivalence margins, and differentiated outcomes: incremental benefit, integration-only improvement, conventional sufficiency, negative net value, and inconclusive evidence. The paper does not report agent experiments or empirical Field-specific advantages. Its contribution is a falsifiable methodology for investigating where Field-Native practices could add explanatory, engineering, or economic value and where their claims should be reduced to competent conventional architecture.

Zenodo (CERN European Organization for Nuclear Research)
Artificial Intelligence Applications
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon