Toward an AI-Native Enterprise Operating System: An Evidence-Centered Reference Architecture for Modular Business Applications
Conference version of the architecture paper: the reference architecture and its evidence-centred invariants.
Research
Two books, two edited volumes and eighteen peer-reviewed papers on record — and the manuscripts now in progress, with the stage each one has reached.
Thought leadership
Two books (2009, 2012), two edited IGI Global volumes (2021, 2026) and eighteen peer-reviewed articles and papers (2006–2011) — 94 Google Scholar citations, presented in person in six countries.
IGI Global · published 20 May 2026
Edited volume (editor), published 20 May 2026: practical applications of LLMs in business intelligence and data analytics. ISBN 9798369366905 (e-book 9798369366929) · DOI 10.4018/979-8-3693-6690-5 · Library of Congress cataloguing names Fabio Perez Marzullo as editor.
2012
iOS development guide for the Brazilian developer community (Novatec, 2012). Assigned in curricula at UTFPR, IFTO, IFS and IFRS; five copies bought by UniRV for its Software Engineering library (2017).
2009
Service-oriented architecture for business innovation (Novatec, 2009). Required reading in 15 public hiring and promotion examinations by 10 public bodies (2011–2022) and assigned in curricula at UTFPR, UERN, IFMT, IFC and other institutions; 40 Google Scholar citations.
IGI Global, 2021
Edited volume (co-editor, with Felipe Antonio de Oliveira): eight chapters by authors in seven countries; indexed in Scopus and the Web of Science Book Citation Index; a representative text of Foothill College's course LINC 60C — Educational Game Design. DOI 10.4018/978-1-7998-5021-2.
Barcelona · Berlin · Portland · Honolulu · Las Vegas · Rome · Vancouver · Paris · Cologne
Nine academic events in four countries: ICEIS 2008 Barcelona · ECMDA-FA 2008 Berlin · IEEE/ACIS ICIS 2008 Portland · IEEE SERVICES / SOA Industry Summit 2008 Honolulu · WORLDCOMP 2006, 2008, 2009 and 2010 Las Vegas · IADIS WWW/Internet 2009 Rome. Industry events: XDS 2017 and 2019 Vancouver, Game Connection Europe 2017 Paris and Gamescom 2018 Cologne (Arcolabs delegate, presenting Amazing Wise Friends); Anima Forum 2016, Rio (consultant-evaluator, Anima Business: Coaching).
Eighteen articles, chapters and conference papers published by Springer, IGI Global, the Association for Information Systems, IEEE-, IADIS-, INSTICC- and CSREA-organized conferences and Brazil's CONSAD public-management congress, on model-driven architecture, domain-driven development, SOA and IT governance in government. 94 Google Scholar citations to eleven works; the ICIS 2010 paper has 244 downloads and 547 PlumX interactions on the AIS eLibrary, with readers in 37 countries.
Research in progress
Manuscripts written in 2026 and the stage each one has reached. Titles are working titles; a venue that reviews double-blind is named only after its decision.
Conference version of the architecture paper: the reference architecture and its evidence-centred invariants.
Companion paper on verification: which architectural invariants can be checked structurally before any behavioural test runs.
How Cortex keeps an enterprise agent's plans inert until policy, approval and idempotency gates allow execution, with an audit chain of what ran.
A fixed natural-language BI pipeline whose query gates fail closed, with deterministic verification and answers linked to their sources.
Extended journal version of the conference paper listed under review above, with the same working title: the full typed data model, the mechanism blocks with code excerpts, the design decisions, two workflow scenarios, the complete evaluation protocol and the ISO/IEC 25010 mapping.
A construction method for a self-describing business suite: a reflective ladder of self-awareness tests and a governed rule for when the system may modify itself.
A versioned semantic layer in which each tool publishes its own knowledge contract and cross-tool questions resolve with evidence attached.
What the literature means by an AI operating system for business software, a six-layer reference model, and how much evidence exists for each layer.
An LLM-assisted measurement architecture for tracking AI-generated software projects, continuing the PATT line of work.
Governance archetypes and competency models for public organisations adopting AI, extending a government IT-governance framework with AI decision rights.
Design and implementation of the protocol through which agents operate Ledgi tools, and what each message commits the system to.
System-description paper: design philosophy, shared services, the tool catalogue and the research agenda of the Ledgi platform.
Pre-registered protocol for a controlled, multi-domain study of AI-assisted enterprise workflows; the study itself has not been run.
Formal model of the platform as an AI operating system, with its algorithms and machine-learning methods.
Integrated logistic support as an AI-native asset operating model, with governed, evidence-linked maintenance decisions.
Turns the stereotype-based assessment of MDA4Eclipse into evaluation contracts bound to specifications in DevOS and Studio.
From XMI repositories shared over web services to knowledge contracts an LLM can consume in the Integration Suite.
Requirement-based costing for AI-generated software: token economics, verification effort and benefits realisation.
Performance contracts for generated queries, continuing the model-driven database profiling work in Insight and the Knowledge Node.
MAPE-K loops, CHOP properties and policy-bounded agents as the autonomic layer over Cortex and Flow.
Traceability graphs, LLM-proposed impact sets and governed re-generation for change-impact analysis in DevOS.
Requirements knowledge as verified organisational memory: classification, retrieval and reuse.
From model-interpreting cloud generators to a policy-governed autonomous software factory.
Student relationship management reframed as consent-governed relationship intelligence.