Towards intelligent supply networks: AI‑enabled simulation in the evolution of digital twins for autologous cell therapy manufacturing

Autologous cell therapy (AuCT) supply chains are complex in ways that differ fundamentally from those of modern industrial products such as automobiles, aerospace, or cell phones. Industrial supply chains manage thousands of parts across multi‑tiered networks, whereas AuCT supply chains are complex because of their tightly coupled production and logistics. Each AuCT batch is linked to a single patient, requiring chain‑of‑identity, chain‑of‑custody, and chain of control. Moreover, each AuCT batch must align clinical appointments with manufacturing slots and logistic schedules to ensure reliable operation. In practice, courier delays, quality release holds, contamination events, and manufacturing capacity conflicts can propagate across stages, extending vein‑to‑vein time and hence reducing patient benefit. In recent years, simulation models have emerged as a tool for AuCT supply chains to evaluate patient service levels, throughput, and cost. However, researchers still need to define ‘what‑if’ scenarios, configure supply chain parameters, and interpret results to run experiments on simulation models, rendering this approach difficult to scale and reproduce. An AI‑enabled supply chain simulation framework is proposed to formalize simulation‑based experimentation through an agentic orchestration layer. The framework simultaneously monitors performance degradation, generates causal hypotheses, conducts targeted comparative experiments, and ranks resilience priorities. Rather than treating the digital model merely as a stress‑testing tool, the framework structures and automates the reasoning process surrounding experimentation. This opens new academic opportunities for studying AuCT supply chains. The proposed framework operates as an AI‑enabled supply chain simulation model and is positioned as a foundational step toward the ultimate realization of a supply chain digital twin. While it does not incorporate real‑time data integration or closed‑loop control, it establishes a structured, hypothesis‑driven approach for scalable and reproducible experimentation in AuCT supply chains.

Autologous cell therapy (AuCT) supply chains are uniquely complex – every batch is patient-specific, requiring chain-of-identity, chain-of-custody, and chain-of-control. An AI-enabled agentic simulation framework is proposed to formalise, automate, and scale hypothesis-driven experimentation across AuCT supply networks, as a foundational step toward full digital twin realisation.

What you will learn

01

Why AuCT supply chains are uniquely complex – and why conventional simulation approaches cannot scale to meet this challenge

02

How an agentic orchestration layer structures simulation-based experimentation as a coordinated, hypothesis-driven reasoning process

03

How this framework supports resilience prioritisation, reproducibility, and transparency – and what it means for decentralised AuCT manufacturing

The four cooperative agents

1

Monitor Agent – detects significant KPI degradation & converts outputs to symptom signals

2

Root Cause Agent – generates causal hypotheses & designs controlled comparative experiments

3

Decision Agent – ranks critical nodes & prioritises resilience investments based on validated evidence

4

Helper Agent – maintains experiment state memory & ensures reproducibility across teams

Key findings

AuCT supply chains require formalised, hypothesis-driven experimentation – current manual, expert-driven approaches are too difficult to scale or reproduce across decentralised networks

The framework operates as an external orchestration layer over existing simulation models – it automates the reasoning loop without altering simulation internals or performing end-to-end reinforcement learning

Structured causal logging by the Helper Agent addresses a key reproducibility gap in simulation-based research, enabling reconstruction of decision paths and cross-institutional benchmarking

This framework is positioned as a foundational step toward full digital twin realisation – real-time data integration and closed-loop control are explicitly scoped as future-phase developments

Key interests

Autologous Cell Therapy

Digital Twins

Supply Chain Simulation

Agentic AI

Resilience Optimisation

CAR-T Manufacturing

Disruption Modelling