Artificial intelligence (AI) is beginning to reshape some of the world’s most energy-intensive industries. In cement and steel – two sectors responsible for roughly 14 per cent of global CO₂ emissions – the technology is helping operators extract more performance from aging plants, optimize fuel use and reduce waste while building more resilient production systems. In doing so, AI acts a a digital accelerator across the five building blocks of industrial decarbonization.

In cement, Daniel Summerbell, Co-founder and Chief Solutions Officer at Carbon Re, estimates that each deployment of their AI operating system reduces emissions in the order of 10,000 tons a year of carbon per plant. In steel, Tarun Mathur, Global Digital Lead for Metals at ABB, points to projects avoiding around three kilotons of CO₂ per year and others delivering 10 to 20 per cent improvements in energy and process stability. 

Both stress that AI is not a replacement for breakthrough technologies like hydrogen-based steel making or carbon capture. But it’s a practical tool for reducing emissions today by helping plants run more consistently and efficiently. For emerging and developing economies (EMDEs), which host much of the world’s cement and steel capacity and often rely on older equipment, these gains matter even more.

Finding efficiency in everyday cement operations

Daniel’s early consulting work showed that manufacturing operations often had room for improvement, without needing new capital equipment. His PhD research confirmed this was also true in the cement industry: “if you compared a good day and a typical day at a cement plant, there was quite a significant difference,” revealing that kilns could burn less fuel if their operation was stabilized.

Scaling those insights required automation. When Daniel met the AI experts who became his co-founders, he saw how machine learning could take on “the constant monitoring, the deep data analysis, pulling together disparate operating data sets” that operators had no time for. And that’s how Carbon Re was born.

Today, the platform starts by mapping years of plant data and understanding how each sensor behaves depending on location, maintenance and calibration. Only then can machine-learning models reliably simulate kiln performance two to three hours ahead. Operators receive recommended control settings that balance efficiency, quality, emissions and fuel use. 

Cement producers are always making trade-offs, Daniel explains. Increasing alternative fuels can cut carbon and costs, but may raise energy use or risk product quality. The simulator makes those consequences explicit, predicts how each choice affects competing indicators and offers settings that strike the balance the plant is aiming for. “This structured way to navigate trade-offs is one of our biggest added values”, he says. The system also retrains itself automatically using live data from the cloud, keeping its models accurate as conditions shift. 

With this mix of services, Carbon Re typically achieves around a three per cent increase in alternative fuels use, which can cut coal consumption by up to 15 per cent, a meaningful shift when scaled across global production.

Stabilizing high-temperature steelmaking processes 

Steel production faces different constraints to cement. Equipment varies widely across plants, processes are highly interdependent and temperatures in melt shops reach 1,600°C. Yet the opportunity is similar: vast, energy-intensive systems where even small gains matter.

Early in his career, working on energy-optimization projects in metals, Tarun noticed that the sector was “kind of a laggard compared to other industries in terms of adoption of AI.” With rising demand for green steel, he says, “that work feels even more relevant.”

Where Carbon Re uses AI to stabilize the behaviour of a cement kiln, ABB applies a similar logic to the molten-metal environment of steelmaking. Tarun describes their AI-enabled process digital twin as “an autopilot for production…you can run it in an energy efficient mode or a productivity mode, and it automatically optimizes the parameters.” The system sits on top of existing automation layers, making real time adjustments based on plants operational priorities.

In upstream steelmaking, where coordination is complex with cranes moving molten metal, AI helps to manage material movement and workflow timing. Using radar, laser and camera systems, ABB maps these movements and feeds them into AI models that streamline the workflow. One project, Tarun notes, helped avoid thermal losses equivalent to “around three kilotons of CO₂ per year” while increasing production by 24,000 tons.

While green steel routes will fundamentally change how this vital product is produced, AI’s role is more immediate. “It tackles inefficiency and waste in the existing operations,” he says. “It makes operations smarter, but more importantly, it is making sustainability and decarbonization more profitable by linking carbon reduction with operations excellence.”

The challenge of data environments

Both leaders say the biggest challenge is not the AI model itself but the data environment needed to make it work. 

In steel, many facilities rely on old machinery built for reliability rather than connectivity. ABB retrofits legacy machines with edge devices using open communication protocols to give them “a digital voice”. Standardizing the data is equally important, ensuring that information from multiple vendors can “speak the same language”. 

In cement plants Daniel says, “data and data infrastructure is probably the number one challenge,” because identical sensors can behave differently depending on placement, maintenance or cleaning cycles. His team often spends months rebuilding this context so the models can separate meaningful signals from routine noise.

Some companies are also cautious about cloud-based solutions, preferring AI tools to sit inside their own IT environments, even if that requires expensive bespoke builds. Smaller firms in EMDEs can sometimes be more open to cloud-based AI.

Once reliable, standardized data flows are in place, both ABB and Carbon Re can scale their systems across plants with only light adjustments. But for many EMDE plants basic connectivity remains the first hurdle.  

The human factor 

AI adoption ultimately depends on people. Across both steel and cement plants, digitalization raises skills gaps and concerns about how new tools will affect daily work. Tarun says workers need support to use digital systems confidently, and companies must invest in training, reskilling and clear career pathways. Some clients are even partnering with universities to build data science expertise and close the talent gap.

Operators must believe the AI understands their production realities and complements their expertise rather than replacing it. For Tarun, the human side of adoption can be just as critical as the technology itself. Organizational readiness, clear roles and workforce confidence ultimately determine weather AI delivers on its potential. 

AI’s footprint

With growing attention on AI’s energy use, both Tarun and Daniel are clear about scale. Daniel says Carbon Re’s compute footprint is roughly 50 tons of CO₂ a year, tiny compared with the 10,000-ton reduction delivered at each plant using the software.

Tarun stresses that industrial AI is fundamentally different from large consumer models trained on massive cloud datasets. ABB’s systems run close to the process on edge devices and are “very lightweight and optimized for industrial environments.” Their energy use is negligible compared with the savings enabled by more efficient furnace operations and smarter load planning.

Next frontiers

Daniel and Tarun expect industrial AI to move from optimizing single processes to coordinating whole production chains. Tarun envisions AI orchestrating material flows, energy use and output across departments and warns that retiring specialists risk taking decades of “tribal knowledge” with them. AI powered assistants, he says, could help capture and transfer that expertise.

Daniel believes leaders should prepare for “an AI industrial revolution happening under the radar,” which will require strong digital foundations, standardized data and readiness to scale AI across multiple sites. He also sees AI playing a broader role in energy systems shaped by renewables. As plants become more flexible and able to adjust demand dynamically, he argues, they could help stabilize grids and make better use of intermittent supply. 

For both, AI in heavy industry has moved well beyond pilots. It is already cutting emissions and improving performance, and the pace of adoption is set to accelerate.