Industrial facilities account for a significant share of electricity demand and can provide opportunities for both energy efficiency and demand-side flexibility. Industry accounted for nearly 40% of global electricity demand growth in 2024, with industrial electricity consumption increasing by nearly 4%. Improving visibility into asset-level consumption and demand patterns can help facilities identify inefficiencies, manage peak demand and respond to changing system conditions. 

Edge AI energy intelligence solutions analyze data from facility control, metering and energy management systems close to the point of use. Applications can include asset- and line-level monitoring, load forecasting, anomaly detection, peak-demand management and identification of operational efficiency opportunities. Local data processing can also reduce reliance on external connectivity and enable sensitive operational data to remain within the facility. Where operational flexibility exists, optimized industrial loads could also participate in demand-side flexibility services, extending facility-level optimization to the wider power system.

Main Outcomes of the Solution

• Affordability: Identification of avoidable consumption, demand peaks and equipment performance issues can reduce energy and maintenance costs and improve energy efficiency per unit of output.
• Reliability and Flexibility: Asset- and line-level forecasting can support peak management and more informed load planning. Where operational flexibility exists, optimized loads may also participate in demand-side flexibility programs.
• Sustainability: Reducing avoidable electricity consumption can lower associated emissions, with impacts measured against site-level consumption baselines.

Key Deployment and Scaling Factors

1. Integrate existing facility systems as a modular layer across control, metering and energy management systems, reducing the need for equipment replacement and limiting operational disruption.
2. Maintain human oversight by allowing operators to review recommendations before implementation and preserving an auditable decision trail.
3. Establish site-level baselines and measurement protocols to distinguish realized savings from projected benefits and verify impacts using metered consumption.
4. Design for replication across industrial sectors while allowing configuration for site-specific factors such as tariffs, operating schedules, equipment and consumption patterns.

Case study

Nommas.ai (2025), Saudi Arabia

Nommas.ai is a Saudi Arabia-based industrial AI company building the intelligence layer for industrial operations, with energy management as a core application. In October 2025, the company deployed its WattX edge AI energy intelligence platform at a manufacturing facility in Saudi Arabia. The platform integrates with the facility’s existing systems and monitors energy consumption across production assets and utilities. 

The deployment uses an advisory-first model in which identified issues are translated into quantified recommendations for operator review before action. A baseline based on historical consumption is used to assess impact, with savings measured against metered consumption following corrective actions.

The platform provides facility-level visibility into energy consumption and is used to identify abnormal consumption, support forecasting-led peak management and detect indicators of equipment degradation. A generative AI layer translates each finding, including leaks and idle consumption, into a specific, quantified recommendation, and the system’s accuracy improves with every verified outcome. Nommas.ai targets a 10–20% reduction in energy cost per facility within the first year. At scale, Nommas.ai aims to aggregate optimized industrial loads to participate in utility-facing demand-side flexibility services. 

References:

1. Electricity – Global Energy Review 2025 – Analysis - IEA
2. Information provided to World Economic Forum by Nommas.ai (June 2026)
3. How AI agents are changing manufacturing in the Global South | World Economic Forum