HC Group and FT Longitude surveyed 131 senior executives across trading houses, asset-backed organisations and financial institutions on the state of AI in commodity trading. Two findings stand out. Executives expect AI to deliver an average additional 3.34% in trading profit and loss (P&L) by 2027.

The same executives rank data quality, access and organisation as the biggest barrier to scaling AI, ahead of systems integration, governance, talent and return on investment.

An industry that prices risk for a living has priced in the upside. The barrier it names sits closer to home.

Two findings, one asset

The expectation and the barrier describe the same asset from different sides. AI earns its P&L in trading by working with the information that runs the business: contracts, prices, positions, counterparties, exposures and the internal logic that connects them. HC Group's analysis is direct on this point. Value emerged first where data quality was highest and integration simplest. Where data sits fragmented across systems, poorly governed or hard to join up, AI stalls at isolated use cases.

The 3.34% depends on solving the data problem. The survey's own respondents say so.

Why trading data stays fragmented

Fragmentation is easy to read as a legacy accident. Some of it is. Some of it is deliberate caution. This data is the commercial edge, so it is guarded: split across the Commodity Trade and Risk Management (CTRM) system, spreadsheets, shared drives and inboxes, with access controlled team by team.

That caution is rational under one condition: that organising the data means preparing it to leave. A firm that consolidates its contracts, positions and counterparty history into a clean pipeline feeding a third-party AI running outside its jurisdiction has organised its way into exposure. Hesitation is the correct response, and hesitation is part of what the survey is measuring.

The architecture question

Where the AI runs decides whether the data problem is worth solving. If the AI is external, every improvement in data quality raises the stakes of the transfer: cleaner data, more to lose. If the AI runs inside the firm's environment, consolidation carries no exposure penalty. The better the data gets, the more the firm keeps.

The CTRM is the natural place for this work. It already holds contracts, positions, exposures, invoices and P&L in structured, governed form. Private AI inside the CTRM works on the data where it lives.

What this looks like in practice

Euclid AI deploys on client-owned infrastructure or inside Euclid's private Swiss data centre, connected directly to the platform. Today it does three things, all within that private environment.

It processes trading documents. An operator drops a contract, invoice or confirmation into the integrated chat and the AI populates the Deal Capture interface for human review and commit. In Euclid's own measurement, data entry on a physical contract fell from eight minutes to under ten seconds, about 98% less time on manual document integration. Verification stays a human step.

It ingests market data, parsing high, low and mean prices from daily PDF reports for automatic database updates.

It answers platform questions in context, giving users step-by-step guidance drawn from Euclid's technical documentation.

Each capability works because the model sits with the data. The same architecture is the foundation for extending private AI across the five domains the platform covers: Trading, Risk, Operations, Finance and Management.

The advantage on the other side of the barrier

Asked where AI will create competitive advantage over the next three years, the surveyed executives pointed to superior proprietary data, faster decision cycles and more accurate risk pricing. All three are internal advantages. All three are internal advantages that compound quietly and resist copying. Each depends on proprietary data staying organised and staying private.”

Your trading data is your edge. The firms that put it to work inside their own perimeter are the ones this survey says will win.

See how private AI works inside your CTRM.