AI Agent Operational Lift for Befesa Zinc Metal in Mooresboro, North Carolina

Why AI matters at this scale

Befesa Zinc Metal, operating as American Zinc Products in Mooresboro, NC, is a mid-sized secondary zinc smelter that recycles hazardous electric arc furnace (EAF) dust from steel mills into high-purity zinc. With 201–500 employees and an estimated $150M in revenue, the company sits in a capital-intensive, energy-heavy industry where even marginal efficiency gains translate into significant cost savings. AI adoption at this scale is not about moonshot projects but about practical, high-ROI applications that can be deployed with existing data infrastructure.

What the company does

The Mooresboro facility processes EAF dust through a Waelz kiln to produce zinc oxide, which is then refined into Special High Grade (SHG) zinc metal. This circular-economy model prevents hazardous waste from landfills while supplying zinc for galvanizing, die-casting, and other industries. The process is continuous, energy-intensive, and sensitive to feedstock variability—ideal conditions for AI-driven optimization.

Three concrete AI opportunities with ROI framing

1. Predictive maintenance for rotary kilns and furnaces
Unplanned downtime in a smelter can cost $50,000–$100,000 per day. By instrumenting critical assets with vibration, temperature, and acoustic sensors, and applying machine learning to historical failure data, Befesa can predict bearing failures, refractory wear, and burner issues days in advance. A 20% reduction in downtime could save $1–2 million annually, with a payback period under 12 months.

2. Real-time process optimization
The Waelz kiln and refining stages involve complex interactions between feed rate, temperature, oxygen levels, and flux additions. Reinforcement learning models can continuously adjust setpoints to maximize zinc recovery while minimizing natural gas and electricity consumption. A 3% improvement in recovery and 5% reduction in energy use could yield $2–3 million in annual savings.

3. Quality prediction from feedstock variability
EAF dust composition varies by steel mill and batch. Using X-ray fluorescence data and historical production logs, a supervised learning model can predict final zinc purity and suggest optimal blending ratios. This reduces off-spec production and rework, potentially saving $500,000 per year.

Deployment risks specific to this size band

Mid-sized manufacturers like Befesa face unique hurdles: limited in-house data science talent, legacy control systems with poor data accessibility, and a workforce accustomed to operator-driven decisions. To mitigate, start with a cloud-based IoT platform (e.g., Azure IoT) to centralize sensor data, partner with a specialized AI vendor for initial models, and run a pilot on one furnace line. Change management is critical—operators must see AI as a decision-support tool, not a replacement. With a phased approach, Befesa can achieve quick wins and build internal capabilities for broader AI adoption.