research

Research grounded
in real operations.

My work sits at the intersection of enterprise AI, industrial knowledge systems, and operations management. It addresses the conditions real deployments face — fragmented data, legacy infrastructure, governance requirements, and decisions with operational consequences — rather than the conditions benchmarks assume.

3research areas
2papers (published & forthcoming)
5research essays
1available to download
research areas

Three interconnected
domains.

Industry 5.0 reframes the goals of industrial transformation — from pure efficiency and automation to a model where technology serves human capabilities, builds organisational resilience, and supports sustainable operations at enterprise scale.

Where Industry 4.0 asked "how do we automate this?", Industry 5.0 asks "how do we make this human-AI collaboration work reliably, at scale, under real operational constraints?" That requires rethinking operating models, knowledge structures, and AI governance from the ground up.

research threads

  • Collaboration architectures for human-AI decision-making in industrial settings
  • Operating model design for AI-augmented, people-centred enterprises
  • Governance frameworks for responsible AI deployment in operations
  • Transition pathways from automation-first to augmentation-first industrial systems

Most enterprise AI deployments underperform not because the models are wrong, but because the knowledge infrastructure is insufficient. Rules, relationships, constraints, and institutional logic remain locked in documents, systems, and people — inaccessible to AI at the moment a decision is needed.

My work here develops practical knowledge graph architectures that hold up in real enterprise conditions: imperfect data, legacy systems, limited semantic expertise, and governance requirements that cannot be waved away.

research threads

  • File-based knowledge graph architectures for practical enterprise adoption
  • Ontology governance and versioning in multi-team environments
  • RAG (Retrieval-Augmented Generation) grounded in structured enterprise KGs
  • Knowledge graph design patterns for manufacturing, construction, and retail
  • Provenance, traceability, and explainability in KG-backed AI systems

Operations systems are being redesigned from first principles. The question is no longer how to automate more tasks or layer AI onto existing workflows. It is how to architect systems where physical processes, digital intelligence, and human judgment are structurally integrated — with defined boundaries, reliable handoffs, and governance built into the design itself.

This domain addresses the engineering and organisational design of next-generation Human-Cyber-Physical Systems (HCPS): environments where the physical and the digital are tightly coupled, where human agency is preserved by design, and where operational decisions carry consequences that demand auditability and resilience.

Prefabricated & Panelised Construction

Construction 5.0 demands reconfigurable production systems where design intent, manufacturing logic, logistics coordination, and assembly execution must remain coherent across long, multi-stakeholder project lifecycles.

Smart Retail Operations

Retail intelligence requires synthesising demand signals, inventory positions, pricing rules, and supplier constraints in real time. The focus: operational context layers that make AI pricing, allocation, and procurement decisions reliable and auditable.

Advanced Manufacturing

Deeply integrated production environments where quality, scheduling, knowledge management, and human oversight operate as a unified system rather than separate applications.

papers

Published &
forthcoming work.

Papers are shared as PDFs by email. Leave an address and it arrives immediately — no list, no follow-up.

Book chapterpublishedAPMS 2025 · Springer, Cham

Robotic Process Automation: A Qualitative Journey Through RPA's Impacts on Company Employees

abstract

Robotic process automation is usually justified in headcount terms, which is the least interesting thing about it. This study takes a qualitative route instead: thirteen semi-structured interviews with RPA practitioners across four multinational companies, asking what automation actually did to the people around it. The finding is that RPA does not straightforwardly displace employees, but it does reshape their work and the governance around it — and how much it reshapes depends on which automation strategy the organisation chose. Task-oriented programmes lean on citizen developers and move quickly at the cost of a coherent process view; process-oriented programmes rely on professional developers and demand far more structured governance. The chapter draws these threads into an integrated framework linking automation strategy, governance model, upskilling, and employee adaptation.

authors

Edgar Simões, Ana Correia Simões, José Coelho Rodrigues, Pedro Miguel Lourenço

Robotic Process AutomationDigital TransformationGovernanceWorkforce AdaptationQualitative Research

key contributions

01
Strategy shapes impact
Task-oriented and process-oriented automation produce different effects on work
02
Governance requirements
What each automation strategy demands of the operating model around it
03
Citizen vs professional developers
Where each model accelerates delivery and where it accrues debt
04
Integrated framework
Automation strategy, governance, upskilling, and employee adaptation as one system
Read the published version →
Conference paperin progress

Reconfigurable Smart Production System for Prefabricated Panelised Construction

abstract

Prefabricated panelised construction offers significant efficiency gains but requires production systems that can adapt to multiple product variants without costly process redesign. Traditional fixed-sequence production lines struggle with design variability, leading to bottlenecks and rework. This paper presents a reconfigurable production architecture that maintains production coherence across variant design spaces while coordinating fabrication sequences, logistics constraints, and multi-site assembly workflows. The approach is grounded in a real industrial deployment and demonstrates how structured ontology-driven production planning can enable factory-level flexibility without sacrificing traceability or quality control.

authors

Pedro Miguel Lourenço

Prefabricated ConstructionReconfigurable ProductionIndustry 5.0Smart Manufacturing

key contributions

01
Reconfigurable architecture
Design framework for production systems that adapt to multiple product variants without process redesign
02
Production-logistics integration
Methods to maintain coherence between design intent, manufacturing sequences, and logistics constraints
03
Multi-site coordination
Governance model for coordinating fabrication, assembly, and site delivery across distributed teams
04
Knowledge traceability
Systems approach to capturing and reusing production knowledge across the project lifecycle
The paper is not available yet
essays

Shorter
arguments.

Essays work through a single problem in the space between a research note and a full paper. Same delivery: leave an email, get the PDF.

Research essayforthcoming

File-Based Knowledge Graphs and Retrieval-Augmented AI for Complex Project Delivery

abstract

Complex engineering and transformation projects generate heterogeneous knowledge that is difficult to integrate, trace, and reuse across multi-year lifecycles. Although knowledge graphs (KGs) and retrieval-augmented generation (RAG) have independently matured, many teams still lack a practical path from fragmented documents to explainable AI-assisted decision support. This paper develops an end-to-end design pattern for file-based project knowledge graphs: KGs whose canonical representation resides in structured files (Markdown + YAML + explicit links), rather than in dedicated graph databases. It is based on a prototype built for a real-case, multi-year, multi-site industrial MES implementation project, and presents a detailed design that covers ontology governance, graph encoding patterns, agentic retrieval loops, provenance rules, human-in-the-loop write controls, and production-oriented cost optimisation.

authors

Pedro Miguel Lourenço

Knowledge GraphsRAGComplex Project DeliveryMESOntologyAgentic AI

key contributions

01
File-based KG architecture
Markdown + YAML + explicit links as canonical graph representation
02
Agentic retrieval loops
RAG pipelines grounded in structured project knowledge graphs
03
Ontology governance
Versioned ontology management for multi-year project contexts
04
Provenance & explainability
Human-in-the-loop write controls and traceable AI decisions
05
Cost optimisation
Production-oriented strategies for sustainable enterprise RAG
Research essayforthcoming

Anatomic Taxonomy-Based Medical Element Recovery from Speech-to-Text AI Transcripts in Radiology Reporting

abstract

Radiology reporting remains a critical bottleneck in diagnostic imaging workflows. While speech-to-text technology has dramatically accelerated dictation, converting unstructured narrative transcripts into queryable clinical data remains manual and error-prone. Transcription errors, anatomic terminology variation, and spatial relationship ambiguity further complicate automated extraction. This paper develops a taxonomy-driven framework for recovering structured medical elements from speech-to-text radiology transcripts by grounding language understanding in formal anatomic ontologies and spatial relationship models. The approach integrates medical NLP with standardised clinical taxonomies (SNOMED, RadLex) and demonstrates how controlled vocabulary recovery can enable reliable downstream tasks — from quality assurance to evidence extraction to epidemiological analysis — without requiring manual correction of transcript errors.

authors

Pedro Miguel Lourenço

RadiologySpeech-to-TextClinical OntologyAnatomic TaxonomyMedical NLPHealthcare AI

key contributions

01
Anatomic taxonomy framework
Structured representation of anatomical concepts and spatial relationships for reliable entity extraction
02
Speech-to-text reliability
Methods to handle transcription errors and clinical terminology variation in radiology narratives
03
Clinical ontology integration
Mapping mechanisms between speech outputs and standardized medical taxonomies (SNOMED, RadLex)
04
Structured data recovery
Conversion from natural language radiology reports to queryable, auditable clinical data structures
The essay is not available yet
Research essayin progress

Ontology Alignment Patterns for Manufacturing ERP Integration

abstract

Every enterprise ontology project eventually collides with the ERP. The data is there, the semantics are not, and a full schema migration is never on the table. This essay works through the practical design patterns for mapping ERP data structures onto a domain ontology without rewriting either — where to put the mapping layer, how to version it as the ERP configuration drifts, and which classes of semantic mismatch are worth modelling versus worth ignoring.

authors

Pedro Miguel Lourenço

OntologyERPSemantic IntegrationManufacturing

key contributions

01
Mapping layer placement
Where the ontology-to-ERP mapping should live, and what it must not own
02
Drift management
Versioning the mapping as ERP configuration changes underneath it
03
Mismatch triage
Which classes of semantic mismatch justify modelling effort and which do not
The essay is not available yet
Research essayin progress

Human-AI Decision Handoff Models in Industrial Operations

abstract

When should an agent escalate to a person? Most deployments answer this with a confidence threshold, which is the wrong instrument — confidence measures the model's certainty, not the cost of being wrong. This essay formalises the conditions, triggers, and interface patterns for reliable human-in-the-loop operation in settings where a bad automated decision has physical consequences, and argues for handoff rules grounded in consequence and reversibility rather than model self-report.

authors

Pedro Miguel Lourenço

Human-AI CollaborationIndustry 5.0AI GovernanceOperations

key contributions

01
Handoff triggers
Escalation conditions grounded in consequence and reversibility, not confidence
02
Interface patterns
How an escalation should be presented for a person to act on it quickly
03
Accountability boundaries
Who owns the decision at each stage of an assisted workflow
The essay is not available yet
Research essayin progress

Cost Optimisation Strategies for Enterprise RAG Pipelines

abstract

A retrieval pipeline that is correct but uneconomical does not reach production. This essay collects the techniques that reduce inference cost in production RAG systems without giving up retrieval quality or explainability: routing by task rather than by default, caching at the semantic layer instead of the prompt layer, structuring retrieval so that the expensive model sees less but better context, and measuring the whole thing in cost-per-answered-question rather than cost-per-token.

authors

Pedro Miguel Lourenço

RAGEnterprise AICost OptimisationLLM Architecture

key contributions

01
Task-based routing
Matching model capability to task difficulty instead of defaulting to the largest model
02
Semantic-layer caching
Caching where meaning is stable rather than where prompt strings match
03
Cost-per-answer metrics
Measuring pipelines by answered question, not by token
The essay is not available yet
approach

What
“research-led”
means here.

01

Primary research

Original design patterns, prototypes, and frameworks developed from real project contexts — not retrospective literature reviews.

02

Applied prototyping

Ideas are validated through working prototypes built on real operational environments, not toy examples.

03

Industry validation

Findings are tested against the constraints of actual deployments — messy data, legacy systems, real governance requirements.

04

Open publication

I publish what I find and make it available to the wider community of practitioners and researchers.

research notes

Thinking in
progress.

Threads I'm exploring — not yet ready for publication but worth documenting. These graduate into essays and papers as they mature.

○ exploring

Provenance models for multi-agent enterprise systems

When several agents contribute to one conclusion, what does a traceable record even look like? Working through the representation before the tooling.

note pending
● active

Ontology versioning under organisational change

Ontologies drift when the business reorganises. Looking at how to version a governed schema without invalidating the history that references it.

note pending
● active

Evaluation harnesses for grounded industrial agents

Benchmarks for agents that operate on plant data barely exist. Building an evaluation approach around traceability and correct escalation, not answer quality alone.

note pending

Want to know when new notes and papers go up? Get in touch and I'll let you know.

Interested in the research?

I'm open to collaborations with industry partners, academic institutions, and practitioners working in the same domains.

Get in touch →