The material we can’t live without, and definitely can’t afford to ignore. Concrete is the backbone...
AI Is Everywhere—Including Concrete
AI adoption is spreading across traditionally conservative industries, and the concrete sector is no longer an exception. From mix design and performance prediction to cost and material optimization, AI-based tools are increasingly being part of everyday engineering workflows.
At the same time, sustainability has become a defining challenge for the sector. The cement industry alone accounts for about 8% of global greenhouse gas emissions, and under the current trajectory emissions from the sector could rise toward 3.8 billion tonnes per year.
In Sweden, this challenge is being translated into policy targets. National targets call for zero CO₂ emission from cement production by 2030, climate-neutral concrete to be available by 2030, and all concrete produced in Sweden to be climate-neutral by 2045. Similar ambitions are emerging across Europe, making sustainability a practical constraint rather than a long-term aspiration.
This convergence has led to a common expectation: the adoption of AI will naturally accelerate progress toward these goals by making sustainability decisions easier and faster.
However, in reality, sustainability in concrete is shaped by complex trade-offs, data limitations, and real-world constraints. AI can expand decision capacity and improve efficiency, but it does not automatically make concrete sustainable.
Why sustainable concrete design remains hard – despite AI
Applying AI to sustainable concrete design is challenging because sustainability is not a single-dimension optimisation problem. CO₂ emissions from concrete are influenced by many factors, including the chemical processes and energy demands of cement production, material availability and regional standards, many of which lie outside the control of mix designers. Reducing emissions in one area can easily introduce constraints or risks in another.
Meanwhile, concrete design is inherently multi-objective. Strength, durability, workability, cost and carbon footprint must all be balanced. Improvements in one dimension often degrade performance elsewhere.
These challenges are further compounded by data limitations: while relevant information exists across many sources, it is fragmented and stored in disparate formats, such as EPDs, laboratory reports and legacy databases, that are difficult to integrate at scale. Without a scalable data foundation, the effective and reliable application of AI remains constrained.
Why most AI models struggle in concrete mix design
Many AI approaches applied to concrete mix design are developed using limited or narrowly scoped datasets. To accommodate these constraints, problems are often simplified in ways that reduce their relevance under real design conditions. Consequently, model outputs may be difficult to validate, interpret, or apply with confidence in industrial settings.
This contributes to a persistent gap between reported research performance and real-world adoption. Addressing these limitations requires a reconsideration of how AI is applied, with greater emphasis on data representativeness, contextual constraints, and practical decision support.
ACORN (AI-powered COncrete Recipe geNerator) : An Innovative Approach
Against this backdrop, ACORN, the AI-powered Concrete Recipe Generator, was developed as a response to bridge the gap between the growing availability of concrete-related data and the practical limitations that have constrained efforts to reduce the climate impact of concrete in everyday engineering workflows.
The system applies an agentic AI to solve the need for simultaneous optimization for physical properties, cost, climate impact, and other constraints such as local standards and codes. Agentic AI refers to AI systems capable of iterative reasoning across multiple constraints rather than single-step prediction making ACORN a unique application in this domain.
It is not positioned as a promise of automated sustainability, but as a way to support more informed decisions at scale.
ACORN integrates large-scale data collection with machine learning to accelerate the development of high-performance, low-carbon concrete mixes. The system follows a data-first approach, combining state-of-the-art AI techniques, such as supervised learning and probabilistic models, with deep domain knowledge in concrete materials built on an agentic AI architecture. Experimental validation, sustainability metrics, and robust data management are embedded as core components of the architecture.
To address the fragmented data landscape typical of the concrete industry, ACORN uses natural language processing to structure information stored across disparate formats, including environmental product declarations (EPDs), mix specifications, laboratory reports, and legacy documentation. This allows AI models to work with datasets that better reflect real-world variability and constraints.
ACORN is currently being applied and further developed through an ongoing project between Ecometrix and Prism Johnson in India.
ACORN does not redefine sustainability; it operationalizes it within real engineering constraints. ACORN enables the identification and comparison of more sustainable concrete options within project-specific constraints, supporting incremental improvements that can be applied consistently and at scale.
Why ACORN Stands Out
- Scalable, data-first infrastructure
Aggregation of concrete-related data across multiple formats using natural language processing for structuring and standardization. By mid-2025, the system contained over 45,000 curated concrete mix recipes.
- Domain-informed AI models
Machine learning models developed in close alignment with concrete materials expertise, experimental validation practices, and sustainability assessment methods.
- Decision support for multi-objective design
Using Agentic AI for rapid evaluation of trade-offs between strength, durability, workability, cost, and carbon footprint without reducing the problems to a single optimization target.
- Practitioner-oriented interface
A user-friendly interface that allows practitioners to rapidly assess the expected performance and sustainability implications of proposed concrete mixes.
- Support for collaboration with data confidentiality
Modular and open system architecture that allows integration of proprietary data while maintaining control over sensitive information.
Conclusion: From algorithms to informed decisions.
The future role of AI in the concrete industry will be defined not by smarter algorithms alone, but by its ability to support better decision-making under real-world constraints. Its value lies in expanding decision capacity by reducing the friction caused by fragmented data, manual analysis, and limited visibility across performance, cost, and carbon considerations.
Systems such as ACORN illustrate how this shift can be achieved in practice. When grounded in data, domain knowledge, and practical constraints, AI can become a meaningful enabler of the incremental improvements required to advance lower-carbon concrete.
