Digital Twins and AI

This white paper synthesises the key themes presented during a lecture on “Digital Twins and AI” delivered to a research delegation from Nomura Research Institute (NRI) on September 18, 2025. It introduces the S.M.I.L.E. methodology (Sustainable Methodology for Impact Lifecycle Enablement) as a comprehensive framework for implementing digital twin strategies across construction, healthcare, manufacturing, logistics, and human-centric applications.

Drawing on current market research, Japan’s Society 5.0 vision, and real-world case studies from Nordic and global contexts, this paper argues that the fundamental barrier to digital twin adoption is not technological but methodological — and that reality, not data, must serve as the persistent layer upon which all digital transformation is built.

The global digital twin market was valued at USD 13.6 billion in 2024 and is projected to reach USD 428.1 billion by 2034, growing at a compound annual growth rate (CAGR) of 41.4% (Grand View Research, 2025). This explosive growth reflects a fundamental shift in how organisations conceptualise, design, operate, and optimise physical systems through their digital counterparts.

However, the vast majority of digital twin implementations remain siloed, data-centric, and disconnected from the broader organisational and societal context in which they operate. The technology exists. What is missing is the methodology.

This white paper presents a methodology-first approach to digital twin implementation, grounded in twenty years of cross-industry experience spanning construction, healthcare, manufacturing, smart cities, telecommunications, space, and robotics. The central thesis is simple yet profound:

This principle — “Impact First, Data Last” — forms the foundation of the S.M.I.L.E. framework and was the core message delivered to the NRI research team during the September 18, 2025 lecture in Tokyo.

Japan’s digital twin market is experiencing significant growth across multiple sectors. The Japan Healthcare Digital Twins Market alone is estimated at USD 48.85 million in 2025 and anticipated to reach USD 360.29 million by 2033, growing at a CAGR of 28.40% (Transpire Insight, 2025). The Japan Building Digital Twin Market is accelerating at a 26.95% CAGR as smart buildings, AI-driven energy optimisation, and net-zero mandates reshape the built environment (OpenPR, 2025).

Japan’s Project PLATEAU, launched in 2020 by the Ministry of Land, Infrastructure, Transport and Tourism (MLIT), represents one of the world’s most ambitious 3D urban digital twin initiatives, aiming to extend models to 500 cities by 2027 (MLIT, 2025). This initiative aligns directly with the SMILE methodology’s first phase of Reality Emulation — creating a shared reality canvas using geographic information systems (GIS) and Building Information Modelling (BIM).

Japan’s Society 5.0 framework envisions a human-centric society that balances economic advancement with the resolution of social problems through the deep integration of cyberspace and physical space. The concept introduces Human Digital Twins as a central element of Cyber-Physical Systems designed to interact with humans in a personalised fashion (Springer, 2024).

The EU-Japan e-VITA project, coordinated by Tohoku University’s Smart Aging Research Centre (SARC), exemplifies this convergence — developing “Digital Twins for Smart Aging” that combine European and Japanese research competencies (Granrath, 2024).

Nomura Research Institute, as Japan’s premier consulting and IT services firm, has been instrumental in driving digital transformation across Japanese enterprise. NRI’s involvement in the Earthbrain consortium (jointly established in 2021 with Komatsu, NTT Docomo, and Sony Semiconductor Solutions) demonstrates the firm’s commitment to applying digital twin concepts to construction and heavy industry (NRI, 2021). NRI’s multicloud strategy and Digital Trust Service further position the firm as a critical enabler of the secure, governed digital infrastructure required for enterprise-grade digital twin deployments (NRI, 2025).

S.M.I.L.E. — Sustainable Methodology for Impact Lifecycle Enablement — is a proprietary, benefits-driven digital twin implementation methodology developed by Nicolas Waern. Its core principle inverts the conventional approach to digital transformation:

Rather than the typical “collect data, build models, derive insights” approach, SMILE begins with the desired impact and works backwards to determine what data, systems, and people are required.

The methodology operates through six progressive phases, each building upon the previous:

Every SMILE implementation is evaluated through three complementary lenses:

SMILE defines a progressive AI maturity path:

This journey maps directly to the lecture’s core argument: digital twins must progress from passive data repositories to active, explainable decision-support systems that ground their outputs in shared visual reality.

SMILE draws on several established academic frameworks:

Do not start with data. The structuring of data, database schemas, data governance policies — these are activities for later phases. The problems observed across healthcare, manufacturing, construction, robotics, and cybersecurity are fundamentally the same: everyone looks at the world through different applications, which means no one can truly communicate with each other.

The primary barriers are not technological:

The lecture presented a detailed walkthrough of building a hospital using SMILE methodology:

This approach has direct relevance to Japan’s Project PLATEAU initiative and NRI’s Earthbrain consortium work.

Japan’s healthcare digital twins market is projected to grow from USD 48.85 million (2025) to USD 360.29 million (2033). Key drivers include:

In a real-world example discussed during the lecture (an unnamed major pharmaceutical company), the methodology was applied to a lab retrofit project using Autodesk Construction Cloud. The critical insight: by requiring vendors to demonstrate data flows, AI capabilities, and interoperability standards within the 3D model before physical installation, the organisation avoids the typical post-deployment discovery that “10 different systems cannot communicate, data cannot be combined, and AI can only operate in silos.”

The convergence of edge computing, federated learning, and digital twins is transforming manufacturing. Recent research (Nature Scientific Reports, 2025) demonstrates a method combining Digital Twin, Edge AI, and Federated Learning for real-time co-simulation in smart factories, enabling edge devices near the factory floor to process data and share learnings while keeping sensitive production data within factories.

The “Invite to Innovate” model: when supply chain disruptions (tariffs, embargoes, geopolitical shifts) require rapid supplier evaluation: scan candidate manufacturing plants using 360-degree cameras and point cloud technology; enable remote “teleportation” into these facilities from anywhere in the world; deploy domain experts who can assess operational maturity, equipment quality, and organisational capability through visual inspection — leveraging 25-30 years of tacit knowledge; combine AI-generated shortlists with human expert evaluation; and make supplier selection decisions within days rather than months.

The lecture addressed the emerging paradigm of Physical AI — robots trained in digital twin environments before physical deployment. Spatial World Models: working with Nokia Bell Labs to create models where 360-degree cameras record reality, generating NeRFs (Neural Radiance Fields) and Gaussian Splats that allow users to “teleport” into any recorded space and view it from arbitrary angles at any point in time. Context-Specific Robot Generation: rather than deploying generic humanoid robots, generate robot designs tailored to specific physical spaces and required tasks. A robot for an agricultural orchard will look fundamentally different from one for a manufacturing plant. Knowledge Transfer: retiring workers transfer knowledge into digital twins through observation and direct interaction. This knowledge trains the next generation of robotic systems that embody “part Sarah, part me, part someone else.”

This principle aligns with critical developments in 2025: the EU NIS2 Directive requires data localisation and stringent controls on cross-border data transfers; the EU AI Act requires transparency and accountability for AI systems, which is more achievable when intelligence resides at the edge; Data Sovereignty ensures users and organisations own and control their data; and Edge computing provides operational continuity regardless of connectivity status.

The digital twin market analysis confirms this trend: the convergence of digital twins with edge computing ensures faster data processing closer to the source, enabling seamless integration across enterprises while supporting remote monitoring and real-time collaboration (Belden, 2025; Litmus, 2025).

Life Atlas, founded by Nicolas Waern, represents the application of SMILE methodology to the most complex system of all: the human being.

The lecture outlined a future where every person has a Large Quantitative Model (LQM) — a physics-based model of themselves, distinct from statistical LLMs. This “soul” (in the Ghost in the Shell sense) transfers between physical embodiments. The companion manifests through augmented reality, physical robots, or purely digital interfaces. The individual maintains complete sovereignty over their digital twin and AI companion.

Japan faces a projected shortage of 11 million workers by 2040, with nearly 30% of the population over 65 by 2042 (IMF, 2025). The care sector alone faces a 570,000 worker shortfall by 2040 (HumansAreObsolete, 2026). 2025 marked the 14th consecutive year of population decline, maintaining an average decrease of half a million people per year.

Knowledge Preservation: record the reality of existing workplaces through digital twins; capture retiring workers’ tacit knowledge through observation and interaction recording; embed this knowledge in AI systems that train the next generation of workers and robots.

Contextual Robot Deployment: do not replace workers with generic robots — instead, generate robot designs optimised for specific physical spaces and required outcomes; train robots virtually first in the digital twin, informed by decades of human operational knowledge; use multi-agent systems where robots, AI agents, and human supervisors collaborate through the digital twin.

Solving Loneliness: AI companions that “mask intelligence and insight and personalisation” to provide genuine relational support; digital twins that visit environments on behalf of their human counterparts, assessing suitability; augmented reality interfaces that overlay digital companions onto physical space.

Labour Mobility and Upskilling: digital twins of manufacturing plants enable global experts to assess, advise, and train without physical travel; education delivered through shared reality, not siloed classroom instruction; upskilling of both humans and robots through the same digital twin platform.

The lecture concluded with a critical warning: the technical capability to automate and robotise exists today. The greater challenge is the societal fabric. Without addressing the social, legal, and ethical dimensions — universal income, meaningful occupation, human dignity — the technological transformation will create more problems than it solves.

The Extreme Collaboration (XC) methodology originated at NASA’s Jet Propulsion Laboratory (JPL) in 1994. Facing planning cycles that had extended from days to 3–9 months for space missions, JPL developed a radically compressed approach: assemble 20 cross-functional specialists (rather than hundreds); place all information in one shared space; enable real-time visibility so that when one specialist modifies parameters, all others immediately see the change and can assess impact on their domains.

The SMILE methodology extends NASA’s XC approach into the digital twin paradigm. The “room” is now the digital twin itself — accessible from anywhere in the world. All information is embedded in the 3D model, not distributed across documents and databases. Specialists teleport into the shared reality rather than traveling physically. AI agents supplement human expertise with real-time analysis and simulation.

Japanese corporate culture — with its emphasis on consensus (nemawashi), attention to detail (kaizen), and long-term thinking — is particularly well-suited to the SMILE methodology’s concurrent engineering approach. The methodology does not replace consensus-building; it accelerates it by providing a shared visual reality in which consensus can form more rapidly and with greater fidelity.

The lecture delivered to Nomura Research Institute on September 18, 2025, presented not merely a technology briefing but a paradigm shift in how digital transformation should be conceived and executed. The SMILE methodology offers a structured, repeatable, and scalable approach to digital twin implementation that addresses the root causes of project failure: fragmentation, sequential working, data-first thinking, and the absence of shared reality.

Japan stands at a unique inflection point. The convergence of Society 5.0 ambitions, Project PLATEAU infrastructure, acute demographic pressures, and world-leading robotics capabilities creates an opportunity to lead global digital twin adoption — not through technology alone, but through methodology.

By anchoring digital transformation in shared visual reality, Japan can bridge the boundaries between disciplines, generations, and the physical and virtual worlds.

The future is not data-driven. It is impact-driven, virtually-first, and grounded in the reality we all share.