We’ve all seen the rise of AI agents across industries – and commodity trading is no exception. AI is showing up in pitch decks, software demos, and strategy slides everywhere. But how much of it is real capability, and how much is marketing dressed up as innovation?
Let’s look at a few common claims through a more critical lens:
“AI is making trades.”
It isn’t. Even in data-driven firms, AI isn’t executing trades – it’s generating signals and suggestions. Traders still make the final call, and for good reason: the risks, regulations, and context around a trade are too complex for automation to handle without oversight.
For example, Microsoft Copilot agents are now helping brokers by analyzing market data, summarizing trends, and recommending actions – but final decision-making remains in human hands (Microsoft, 2024).
“Forecasting tools are AI-powered”
That depends on where you look. Most tools use AI for data preprocessing – filling gaps, spotting anomalies, normalizing inputs. There is nothing wrong with that, but calling it “AI forecasting” is a stretch. The actual forecast often comes from traditional models, not machine learning.
A real world example: National Grid UK saw a 40% boost in solar forecast accuracy (Commodity Technology Advisory, 2025). But the breakthrough didn’t come from futuristic prediction models – it came from using AI to process live satellite imagery and feed higher-quality data into existing forecasting models. The result was better accuracy and tighter reserve planning, not a full AI overhaul.
“Natural language will transform CTRMs”
Eventually – but not yet. Real examples exist, but they are rare. Some vendors have launched LLM-based tools that interpret natural language and convert it into API calls or database queries. One such implementation runs completely on-premises, disconnected from the internet to manage security risks. It works, but the cost of replicating it at scale? Very high (Commodity Technology Advisory, 2025).
What’s real: Natural language is a great fit for reporting – especially where users can ask questions like “What’s our PnL exposure by counterparty this month?” and get usable results without writing SQL. That’s the low-hanging fruit, and it’s where the most progress is happening.
AI is rapidly becoming the future user interface (UI) for Microsoft Dynamics 365
AI does make Dynamics 365 easier to navigate. Copilot can summarise records, draft customer communications, and pre-fill complex ERP forms (Microsoft, 2024). This streamlines data entry and speeds up repetitive processes – particularly in finance and operations modules.
But domain expertise is still critical. AI can suggest entries, yet only trained professionals can ensure the inputs reflect contractual terms, compliance rules, and market realities. Sound decision-making still depends on contextual understanding, critical thinking, and maintaining data integrity.
So where is AI actually creating value today?
AI’s most tangible benefits in commodity trading and CTRM environments come from focused, well-defined applications that address clear business pain points.
- Productivity and workflow automation – AI agents such as Microsoft Copilot now draft emails, summarise meetings, extract action points, update ERP and CTRM records, and aggregate data from multiple systems (Microsoft, 2024). These capabilities reduce administrative effort and give teams more time for market analysis and decision-making.
- Software development acceleration – AI tools generate and review code, assist with debugging, create unit tests, and draft technical documentation. In many development teams, these capabilities deliver efficiency gains of 15–20% while still requiring human oversight for design and quality assurance (Commodity Technology Advisory, 2025).
- RPA-driven operational efficiency – Robotic Process Automation reduces manual workloads by as much as 80–90% in processes such as invoice reconciliation, nomination scheduling, and document matching. A midstream natural gas trading company cut costs by $245,000 annually after introducing RPA into its scheduling and balancing workflows (Commodity Technology Advisory, 2025).
- Short-term forecasting improvements AI-enhanced forecasting models in power and renewables combine live weather feeds and satellite data, giving forecasters a clearer view of volatility windows. This allows trading desks to adjust positions quickly when conditions change.
- Supply chain optimisation Machine learning models predict fuel tank refills at gas stations, optimise truck routing, and reduce delivery frequency. These capabilities lower transport costs and emissions while improving service reliability.
- High-speed information services GenAI and LLM-powered tools scan shipping movements, price indices, and satellite feeds to deliver targeted insights in seconds. Analysts can process far more information in less time, improving both speed and depth of market understanding.
