AI Agent Operational Lift for Tellus Products, Llc in Belle Glade, Florida

Why AI matters at this scale

Tellus Products operates a single, mid-sized molded fiber facility in Belle Glade, Florida, employing 201–500 people. At this scale, the company sits in a challenging middle ground: too large to manage production with spreadsheets alone, yet too small to have a dedicated data science team or modern cloud data infrastructure. The packaging and containers sector runs on thin margins, where material costs (recycled paper, bagasse) and energy for drying represent 60–70% of cost of goods sold. AI is not a luxury here—it is a tool to protect those margins by squeezing out waste and variability that human operators cannot see in real time. For Tellus, the most practical AI entry points are edge-based computer vision and lightweight predictive models that can run on existing industrial controllers without requiring a full cloud migration.

Three concrete AI opportunities with ROI framing

1. Computer vision for inline quality control

Molded fiber trays often suffer from cosmetic defects—thin spots, clumps, or cracks—that lead to customer rejections and wasted material. By mounting industrial cameras over the forming line and training a defect classifier, Tellus can catch bad units before they enter the dryer. At $45M in revenue and typical fiber waste rates of 8–12%, a 15% reduction in scrap could save $500K–$800K annually. The hardware cost for a single line is under $50K, with payback in less than six months.

2. Predictive maintenance on thermoforming presses

Unscheduled downtime on a high-speed press can cost $5,000–$10,000 per hour in lost output. Vibration, temperature, and cycle-time data from PLCs can feed a simple anomaly detection model to forecast bearing or seal failures 2–4 weeks in advance. Even preventing two major breakdowns per year covers the cost of sensors and a part-time data engineer, while also extending asset life.

3. Dynamic scheduling to reduce changeover waste

Every time Tellus switches between product molds or colors, the line produces transitional scrap and loses productive time. A constraint-based optimization engine—ingesting order backlog, mold availability, and material constraints—can sequence jobs to minimize these losses. A 10% reduction in changeover time could free up 200+ hours of capacity annually, equivalent to adding a week of production without capital expenditure.

Deployment risks specific to this size band

Mid-sized manufacturers face unique AI risks. First, data poverty: many machines lack sensors, and quality records may live on paper. Without a digitization step, models will fail. Second, workforce readiness: operators may distrust or override AI recommendations if not involved early. A "human-in-the-loop" design where AI flags issues but humans make the final call is critical. Third, IT/OT convergence: connecting factory floor systems to any cloud service opens cybersecurity risks that a small IT team may struggle to manage. Starting with fully on-premise, edge-deployed models mitigates this. Finally, vendor lock-in: Tellus should avoid proprietary AI platforms that demand multi-year contracts. Open-source tools (e.g., TensorFlow, Node-RED) and modular sensors keep the company in control and allow incremental scaling.