Human–AI Collaboration for Real-World Decision-Making
centers on one of the defining questions of our moment: what does it actually mean for humans and AI to make decisions together?
Across 88 pages, practitioners, CEOs, and researchers from energy, fashion, logistics, healthcare, and supply chain share one conviction technology must follow values, not lead them.
From LLMs that finally explain what the optimizer decided and why, to agentic systems navigating hospital floors and factory lines, every contribution asks the same underlying question: does this tool serve the problem, or are we retrofitting a problem to the tool?
WHEN THE OPTIMISER SPEAKS by Matteo Pozzi, CEO at Optit
HOW GEN AI IS TRANSFORMING THE CREATIVE PROCESS IN FASHION by Carmine Cerrone, Founder at Four Clicks
DECISIONS WE CAN TRUST: BUILDING AI TO EMPOWER RATHER THAN REPLACE by Annalisa Andaloro, Head of EU Project Portfolio at 28DIGITAL
FROM DATA SCARCITY TO DECISION INTELLIGENCE HOW BAOBAB APPROACHES COMPLEX DECISION-MAKING by David Sánchez, Chief AI Sales Officer at Baobab
THE DECISION LAYER – WHY OPTIMIZATION IS ESSENTIAL TO ENTERPRISE AI by Oliver Baster, Chief Technology Officer at Gurobi
JAMES WATT AND ARTIFICIAL INTELLIGENCE by Armando Miguel Nieto Ranero, President at Divina Seguros FROM AUTOMATION TO SUPERVISED ORGANIZATIONAL AUTONOMY by Massimo Mistretta, CEO at Exentriq
VISIBILITY WON’T SAVE YOUR SUPPLY CHAIN Interview to Fabio Luraschi, Inventory Strategy Manager at GUESS Europe
FROM DATA TO ACTION ON THE FACTORY FLOOR: HUMAN–AI COLLABORATION AT THE WRIST by Mauro Mazzolini, Co-founder at J&W srl FROM DATA TO DECISIONS: WHY AI IN TRANSPORT STILL NEEDS HUMAN INTELLIGENCE by Ennio Andrea Adinolfi, CEO & Co-founder at Minervas
BEYOND THE TRADITIONAL WMS: AGENTIC PLATFORMS AND INTELLIGENT ORCHESTRATION FOR MODERN LOGISTICS by Marco Crasnich, CEO at Overlog srl
BEYOND THE LLM HYPE: ENGINEERING INTELLIGENCE FOR FASHION Interview to Alice Calamita, AI & Optimization Specialist at Syanctica HUMAN–AI COLLABORATION IN REAL-WORLD DECISION-MAKING: BUILDING RESILIENT INTELLIGENCE FOR COMPLEX ENVIRONMENTS by Immacolata De Santi, Co-Founder, Co-CEO, CMO at TheFace5.0 Italia S.r.l.
PROOF BEFORE PRESS by Andrea Cinelli, CEO FoolFarm Spa & Member of the Advisory Board of Deloitte DCM
WHEN THE ALGORITHM ASKS FOR A SECOND OPINION by Francesca Antonella Aiello, R&D Coordinator at Math Biology
NEAR-ZERO INVASIVE MEDICINE: A DECISION SCIENCE FRAMEWORK FOR HUMAN-CENTRED, MODEL-DRIVEN CLINICAL CARE by Raffaele Maccioni, CEO & Chief Research Officer at Math Biology & President at Decision Science Alliance
WHEN HUMANS AND MACHINES FINALLY AGREE ON A DECISION Maddalena Mazzali, Community Director at Decision Science Alliance
Welcome to the New Edition of IDA, guided by its key voices
LETTER FROM THE PUBLISHER
VALUES FIRST, TECH FOLLOWS A DECISION SCIENCE PERSPECTIVE ON HUMAN–AI COLLABORATION
The pace of AI innovation in 2026 is not just fast, it is directionally ambiguous. Large Language Models, agentic systems, and autonomous decision tools are diffusing across healthcare, supply chains, manufacturing and public policy with a velocity that compresses the time available for critical thinking. As decision scientists, we face a specific and concrete risk: mistaking acceleration for progress.
Our position as a community is clear. Technology follows values, not the reverse. The chain is non-negotiable: values → objectives → model of the system → decision. Break that chain anywhere and you don’t have a decision support system. You have an expensive shortcut.
Five imperatives follow from this. Match tool to problem. LLMs are powerful and genuinely transformative – for the problems they fit. They are not universally optimal. Simulation, mathematical optimization, and causal modeling each address problem structures that LLMs handle poorly. Defaulting to dominant paradigms wastes resources and produces brittle solutions. Build hybrid infrastructures. The most robust decision support systems in 2026 combine techniques, LLMs for language and pattern, simulation for dynamic complexity, optimization for constrained resource allocation, human judgment for values and accountability. Integration is the discipline, not selection. Count the cost of computation. Energy consumption is a dimension of solution quality. Computational responsibility is not a compliance checkbox, it is part of what it means to build well. Infrastructure before autonomy. Before deploying agents, build an agentic culture: clear decision rights, auditable workflows, explainable outputs. Engineering readiness precedes automation. Shortcuts here create systemic risk. Distinguish innovation from application. Using a well-known technique creatively is innovation. Developing new methodology is research. These require different teams, different risk tolerance, different investment horizons. Confusing them is organizationally expensive.
The hospital is not a random example. It is one of the most compelling convergence points in decision science: a complex system where clinical urgency, resource scarcity, regulatory accountability, and human vulnerability intersect simultaneously. People in need are at the center, and every decision propagates consequences across the entire system. Consider a regional hospital network optimizing patient flow and resource allocation. The rigorous approach begins top- down: define patient outcome values first, then model the system architecture before selecting any technology. A discrete-event simulation captures dynamic patient pathways and bottleneck behavior. Mathematical optimization governs scheduling and resource constraints. At the data layer, a Large Language Model handles synthesis of heterogeneous clinical documentation, but a Small Language Model, fine-tuned on local clinical vocabulary and running on-premise, manages real-time triage classification and alert generation. The two models cooperate: the SLM acts fast and efficiently at the edge; the LLM reasons across broader context when depth is required. Together they reduce computational cost without sacrificing capability. Crucially, this architecture is designed to be updatable and harmonized, a living model of the system, not a static deployment. As patient populations shift or new protocols emerge, the simulation parameters and model layers adapt while the value layer remains stable. And the workforce is not only human. Software robots handle administrative workflow automation. Hardware robots, autonomous guided vehicles, medication dispensing systems, and physical logistics. Human clinicians, coordinators and specialists retain full decision authority on patient care. The result is a human-robot cooperative system where each layer does what it does best, governed by a coherent architecture built from values down. This is what responsible Human–AI collaboration looks like in practice: not one model doing everything, but a designed system where intelligence, artificial and human, is allocated deliberately.
Raffaele Maccioni IDA Publisher Director and DSA President & Co-Founder
Raffaele Maccioni is an entrepreneur-manager focused on building initiatives from research to create tangible market value. Electronic – Control Engineer from the Polytechnic of Milan. Over 25 years of experience and passion for analytical models and algorithms applied to complex systems. Raffaele is now leading and supporting as R&D Executive a few companies and entrepreneurship initiatives.
