AI Agent Operational Lift for Montana Renewables, Llc in Great Falls, Montana

Why AI matters at this scale

Montana Renewables operates a 200+ employee biorefinery in Great Falls, converting waste fats and oils into renewable diesel and sustainable aviation fuel. As a mid-sized, capital-intensive manufacturer, the company faces tight margins, complex chemical processes, and volatile feedstock markets. AI offers a path to operational excellence by extracting insights from the vast sensor data generated across the plant. At this scale, even a 1% yield improvement or a few days of avoided downtime can translate into millions of dollars in annual savings. Moreover, with the recent $350 million expansion, the facility now has the scale to justify investments in advanced analytics and machine learning infrastructure.

What Montana Renewables does

Montana Renewables is a leading producer of low-carbon fuels, utilizing hydroprocessing technology to turn agricultural waste and used cooking oil into drop-in replacements for petroleum diesel and jet fuel. The plant’s output helps meet California’s Low Carbon Fuel Standard and federal Renewable Fuel Standard obligations, positioning the company as a key player in the energy transition. With a workforce of 201–500, it blends the agility of a growth-stage firm with the process rigor of a refinery.

Why AI matters in renewable fuels

The renewable fuels sector is data-rich but insight-poor. Thousands of temperature, pressure, and flow sensors generate terabytes of time-series data daily. AI/ML models can correlate these variables to predict catalyst deactivation, optimize hydrogen consumption, and reduce energy intensity—areas where traditional control systems fall short. Additionally, feedstock quality varies significantly, and AI can dynamically adjust process parameters to maintain yield and product quality. As regulatory carbon accounting becomes more stringent, AI-powered lifecycle analysis tools can automate compliance reporting and identify the lowest-carbon feedstock blends.

Three concrete AI opportunities with ROI

1. Predictive maintenance for critical rotating equipment. Hydroprocessing reactors rely on high-pressure pumps and compressors. Unplanned failures can halt production, costing over $500,000 per day. By training models on vibration, temperature, and oil analysis data, the plant can forecast failures weeks in advance, schedule maintenance during planned turnarounds, and reduce downtime by 20–30%. Payback is typically under 12 months.

2. Real-time process optimization with reinforcement learning. The hydrotreating and isomerization units have dozens of setpoints that affect yield and energy use. An AI agent can continuously explore optimal combinations, learning from historical and real-time data to maximize diesel yield while minimizing natural gas consumption. A 2% yield gain on a 15,000 barrel-per-day plant could add $10–15 million in annual revenue.

3. Feedstock blending intelligence. The plant processes a mix of soybean oil, tallow, and used cooking oil, each with different free fatty acid profiles and contaminants. AI models can predict how blend ratios impact catalyst life and product quality, enabling procurement to buy the cheapest mix that still meets specs. This could reduce feedstock costs by 1–3%, a significant margin lever.

Deployment risks specific to this size band

Mid-sized manufacturers like Montana Renewables face unique AI adoption hurdles. First, legacy OT/IT integration: many sensors and control systems were not designed for cloud connectivity, requiring middleware and edge computing investments. Second, talent scarcity in Great Falls makes hiring data engineers and ML specialists difficult; partnering with a system integrator or using managed AI services is often more practical. Third, change management: operators and engineers may distrust black-box recommendations, so explainable AI and a phased rollout with human-in-the-loop validation are critical. Finally, cybersecurity risks increase when connecting operational technology to the cloud, demanding robust network segmentation and monitoring.