AI can cut material waste and boost factory resource efficiency—case evidence shows up to 25% efficiency gains and 30% less scrap—but gains are uneven, with SMEs and firms in emerging markets falling behind due to data and capability constraints.
Artificial intelligence as a catalyst for the circular economy: a mixed-methods analysis of sustainable production ecosystems (2023–2024)
A systematic review and bibliometric analysis finds that AI applications in circular-economy production systems can deliver substantial resource-efficiency gains (reported up to 25%) and scrap reductions (up to 30%) in documented cases, but benefits are uneven and limited by data, capability, and adoption barriers—especially for SMEs and firms in emerging economies.
Abstract Artificial intelligence (AI) is emerging as a powerful driver of the circular economy (CE), enabling production systems to become more resource-efficient, less waste-intensive and strategically aligned with sustainability goals. This study offers the first comprehensive mixed-methods assessment of how AI transforms industrial production ecosystems in the post-ChatGPT era. Drawing on a bibliometric network analysis of 196 peer-reviewed articles (2023–2024) and a systematic review of 104 studies, we identify the main AI-enabled mechanisms advancing CE principles in smart manufacturing, waste valorisation, supply-chain transparency, and sustainable design. The evidence shows that AI can increase resource-efficiency metrics by up to 25% and reduce production scrap by as much as 30% in documented cases. We also highlight adoption barriers, particularly for small and medium-sized enterprises and firms in emerging economies, where capability and data constraints limit impact. This research deepens theoretical understanding by integrating CE principles, Industry 4.0 architectures, green innovation theory, and lifecycle assessment into a unified conceptual framework. It also delivers targeted recommendations for policymakers, managers, technology providers, and investors. The findings position AI not merely as an operational tool but as a strategic orchestrator of regenerative production systems, offering a clear roadmap for accelerating circular transitions in line with the Sustainable Development Goals.
Summary
Main Finding
AI is emerging as a strategic catalyst for circular-economy (CE) transitions in industrial production. Mixed-method evidence (bibliometric mapping + systematic review + cases) shows AI can materially improve resource-efficiency (documented increases up to ~25%) and reduce production scrap (documented reductions up to ~30%) when embedded in Industry 4.0 infrastructures. However, benefits are uneven: SME and emerging-market adoption is constrained by capability, data, capital and regulatory gaps. The paper integrates CE principles, Industry 4.0, green-innovation and lifecycle assessment into a unified framework and produces actionable recommendations for policymakers, firms, technology providers and investors.
Key Points
- Scope and contribution
- First comprehensive mixed-method assessment of AI–CE interactions in the post-ChatGPT period (papers published 2023–2024).
- Combines bibliometric network analysis (196 unique papers) with a systematic review (104 high-quality studies) and six illustrative case vignettes.
- Empirical headline effects
- Reported case evidence: resource-efficiency gains up to ~25%; production scrap reductions approaching ~30%.
- Theoretical synthesis
- Four integrated lenses: resource-based view & dynamic capabilities; socio-technical transition (MLP); ecological modernisation / industrial ecology; Industry 4.0 integration model.
- Two propositions:
- P1 (Dynamic-capability effect): Firms with stronger AI dynamic-capability bundles will outperform on CE metrics (material recirculation rate, waste-to-landfill reduction, energy/unit).
- P2 (Complementarity effect): AI impacts are magnified when embedded in mature Industry 4.0 infrastructures and driven by external pressures (regulation, resource scarcity).
- Mechanisms where AI advances CE
- Operational eco-efficiency: predictive maintenance, process optimisation, energy optimisation, waste minimisation.
- Design-for-circularity & transparency: generative design for material reduction, digital twins with real-time LCA, AI + blockchain provenance, AI-driven optical sorting/waste valorisation.
- Recent technological advances (2023–2024)
- Generative AI for eco-design, real-time LCA fed by IoT, digital twins for regenerative agriculture, AI–blockchain for traceability.
- Key barriers & research gaps
- Limited longitudinal and ecosystem-level impact studies; heavy reliance on single-case/short-run simulations.
- SMEs under-represented; capability and data constraints limit adoption and impact.
- Contextual gaps on adoption in emerging economies (infrastructure, regulation, financing).
- Data-governance, upfront capital costs, interoperability and standards issues.
Data & Methods
- Data sources and period
- Web of Science Core Collection, Scopus, Dimensions (searched 15 Jan 2025); publication window 1 Jan 2023 – 31 Dec 2024; English-language articles & reviews.
- Search string: (“artificial intelligence” OR “machine learning” OR “deep learning”) AND (“circular econom” OR “closed-loop” OR “resource efficien” OR “industrial symbiosis”) AND (“manufactur*” OR “production” OR “supply chain” OR “smart factory”).
- Sample sizes
- Initial harvest: 278 (WoS), 312 (Scopus), 241 (Dimensions); de-duplicated to 196 unique records for bibliometrics (158 empirical, 38 reviews).
- Systematic review corpus: 104 high-quality studies (selected from the same databases + Google Scholar; PRISMA flow used).
- Methods & tools
- Bibliometric mapping: Bibliometrix (R) 4.3.2, VOSviewer 1.6.20; thresholds e.g., minimum 5 keyword occurrences; sensitivity checks (lowering thresholds) produced ~87% overlap in high-centrality nodes.
- Reference management: EndNote; qualitative coding & synthesis: NVivo; inter-rater screening agreement κ = 0.92.
- Case analysis: six purposive cases mapped to theoretical quadrants (SME to multinational examples).
- Outcomes extracted and coded
- Publication venue, method, industrial sector, AI technique, CE principle targeted, reported performance metrics, contextual moderators (firm size, region, regulation).
- Robustness & limitations of the evidence base
- Cross-platform bibliometric checks, PRISMA standards, high inter-rater reliability.
- Empirical evidence skewed toward short-run or single-factory cases; ecosystem- and longitudinal metrics sparsely reported.
Implications for AI Economics
- For firm-level productivity and returns
- AI investments can generate measurable resource-efficiency and waste-reduction gains, creating direct cost savings and potential revenue from recovered materials. These benefits strengthen AI’s role as a source of (heterogeneous) firm-level competitive advantage consistent with RBV/dynamic-capabilities theory.
- Complementarity matters: returns to AI are higher when firms also invest in sensors, data infrastructure, and interoperable systems (Industry 4.0). Economists should model AI returns as non-linear and conditional on complementary capital.
- For market structure and inequality
- Potential to widen performance gaps: large firms and digitally mature producers capture outsized CE benefits; SMEs and firms in emerging economies risk being left behind without targeted support. This raises distributional concerns and justifies public policy to correct market failures (finance, skills, infrastructure).
- For measurement and empirical research in AI economics
- Suggested CE metrics for empirical studies and cost–benefit analysis: material recirculation rate, waste-to-landfill reduction, scrap rate, secondary-material yield, energy-per-unit output, dynamic LCA indicators. Integration of these with standard productivity metrics will allow fuller valuation of AI investments.
- Need for longitudinal, ecosystem-level datasets and quasi-experimental studies to estimate causal effects, spillovers, and diffusion dynamics.
- For policy and investment strategy
- Market failures (financing constraints, coordination problems, data public-goods) imply roles for targeted subsidies, R&D support, data-governance frameworks, and interoperability standards to accelerate inclusive CE adoption.
- Carbon/resource pricing and procurement rules that value circular outcomes can raise private returns to AI-enabled CE investments (magnifying P2).
- For modeling environmental-economic trade-offs
- AI can facilitate partial decoupling of output from environmental impact, but evidence is context-dependent. Economists should incorporate rebound risks, lifecycle boundaries, and supply-chain spillovers into welfare and climate-policy models.
- Research agenda for AI economics suggested by the paper
- Estimate heterogeneous treatment effects of AI adoption across firm sizes, sectors and countries.
- Evaluate public interventions (grants, concessional finance, procurement) that lower adoption barriers for SMEs and emerging-economy firms.
- Build macro/sectoral models linking firm-level AI-driven circular improvements to national-level material flows and SDG indicators.
- Cost–benefit and distributional analyses that internalise lifecycle environmental benefits and potential labor-market adjustments.
Short summary conclusion: AI is not just an operational optimiser but a strategic orchestrator for circular production when combined with Industry 4.0 complements and supportive policy. The economic promise is substantial but uneven; rigorous, longitudinal and ecosystem-scale economic research and targeted public interventions are needed to turn potential gains into inclusive, scalable outcomes.
Assessment
Paper Typereview_meta
Evidence Strengthmedium — Presents collated empirical results from a systematic review and bibliometric analysis, including reported improvements in resource efficiency and scrap reduction, but relies largely on heterogeneous case studies and short-run evaluations without consistent causal identification or meta-analytic synthesis and is susceptible to publication and selection biases.
Methods Rigormedium — Uses systematic review and bibliometric network analysis on a defined corpus (196 articles, 104 studies) and integrates theoretical frameworks, but the scope is narrow (2023–2024), the quality and design of included studies are heterogeneous and not uniformly assessed in the abstract, and no clear causal-identification or aggregated effect-size estimation is reported.
SampleBibliometric network analysis of 196 peer-reviewed articles (2023–2024) and a systematic review of 104 studies examining AI applications across smart manufacturing, waste valorisation, supply-chain transparency, and sustainable design; evidence drawn from case studies, pilot deployments and empirical evaluations reporting resource-efficiency gains (up to 25%) and scrap reductions (up to 30%).
Themesproductivity adoption innovation
GeneralizabilityShort publication window (post-ChatGPT 2023–2024) limits temporal generalizability, Likely publication bias toward positive case studies and pilot projects, Geographic and firm-size bias: SMEs and emerging-economy contexts underrepresented in demonstrated impacts, Sectoral focus on manufacturing, waste valorisation and supply chains limits transferability to services or other sectors, Heterogeneous definitions of AI and outcome metrics across included studies constrain comparability, Evidence dominated by case studies and pilot evaluations rather than randomized or quasi-experimental designs
Claims (8)
| Claim | Direction | Confidence | Outcome | Details |
|---|---|---|---|---|
| AI can increase resource-efficiency metrics by up to 25% in documented cases. Organizational Efficiency | positive | high | resource-efficiency metrics |
n=104
up to 25% increase
0.24
|
| AI can reduce production scrap by as much as 30% in documented cases. Error Rate | positive | high | production scrap (waste generated during production) |
n=104
as much as 30% reduction
0.24
|
| Artificial intelligence is emerging as a powerful driver of the circular economy (CE), enabling production systems to become more resource-efficient, less waste-intensive and strategically aligned with sustainability goals. Organizational Efficiency | positive | high | resource efficiency and waste intensity of production systems |
n=300
0.24
|
| This study offers the first comprehensive mixed-methods assessment of how AI transforms industrial production ecosystems in the post-ChatGPT era. Other | null_result | high | novelty / comprehensiveness of the study |
n=300
0.04
|
| The study identifies the main AI-enabled mechanisms advancing CE principles in smart manufacturing, waste valorisation, supply-chain transparency, and sustainable design. Innovation Output | mixed | high | AI-enabled mechanisms advancing circular economy principles (e.g., in smart manufacturing, waste valorisation, supply-chain transparency, sustainable design) |
n=196
0.24
|
| Adoption barriers exist, particularly for small and medium-sized enterprises and firms in emerging economies, where capability and data constraints limit impact. Adoption Rate | negative | high | adoption barriers / limitations to AI impact (capability and data constraints) |
n=104
0.24
|
| This research deepens theoretical understanding by integrating CE principles, Industry 4.0 architectures, green innovation theory, and lifecycle assessment into a unified conceptual framework. Other | null_result | high | conceptual/theoretical integration (framework development) |
n=300
0.12
|
| The findings position AI not merely as an operational tool but as a strategic orchestrator of regenerative production systems, offering a clear roadmap for accelerating circular transitions in line with the Sustainable Development Goals. Innovation Output | positive | high | role of AI in enabling/regenerating production systems and accelerating circular transitions |
n=300
0.24
|