Microsoft, ANZ, HSBC, and Lloyds Unveil AI Agent for Trade Finance — Automated MT700 Letter of Credit Processing at Sibos 2025

What did Microsoft announce at Sibos 2025?

On April 20, 2026, Microsoft published a blog post presenting a proof-of-concept AI agent for trade finance developed in collaboration with three major banks — ANZ (Australia and New Zealand Banking Group), HSBC, and Lloyds Bank. The solution was publicly demonstrated at Sibos 2025 in Frankfurt, the world’s largest gathering of professionals in financial message exchange and payments.

The goal of the collaboration is to solve a problem that has burdened international trade for decades — the manual processing of documentary letters of credit and matching shipping documentation against letter of credit conditions.

What specific problem does the AI agent solve?

The core of the solution is automating work with MT700 letters of credit — SWIFT-standardized messages carrying all conditions of a documentary letter of credit between the issuing bank and the beneficiary. The traditional process looks like this: an exporter sends the bank an invoice and shipping documentation, a bank employee manually reads the MT700, compares every item with the submitted documents, and searches for discrepancies. If the amount, date, goods description, or destination deviates — payment is blocked.

The AI agent does this automatically:

• Parses the MT700 letter of credit — extracts all conditions and restrictions into a structured format
• Compares with the invoice and shipping documentation — detects mismatches in amount, deadlines, goods specification, ports, and carrier
• Generates a discrepancy report for the treasury user to review and approve
• Offers a conversational interface — instead of reading tables, the user can ask “why is this shipment blocked” and receive a direct answer

What technology is behind the solution?

The foundation is the Microsoft Foundry platform — an enterprise layer for AI agent orchestration — combined with large language models that perform the actual document interpretation. Microsoft’s blog post does not reveal the exact model IDs used, but emphasizes that the architecture supports on-premises and hybrid deployments — crucial for banks due to regulatory requirements for data residency.

Each of the three banks tested the agent on their own real documents during the pilot, under internal compliance oversight. Results show the agent significantly reduces processing time per case.

Why is trade finance so difficult to digitize?

The estimates Microsoft cites are from industry reports: the global trade system produces approximately 4 billion documents daily, with only 1-2% in fully digital form. The reasons are systemic:

• Non-uniform standards between banks, jurisdictions, and logistics partners
• Legal requirements for original signed documents in many countries
• Historical investment in paper processes — old banks have decades of workflows built around PDFs and fax
• Fraud risk driving manual validation even when technology is available

The AI agent is the first layer that does not require the entire industry to digitize at once — it works with existing formats (scanned PDFs, MT messages, Word templates) and extracts structure from them.

Implications for the banking sector

For banks processing thousands of letters of credit per month, the potential savings are significant — not just in person-hours, but also in reducing the risk of errors. A person reviewing their 40th MT700 document of the day makes mistakes; an AI agent works consistently.

For Microsoft, the partnership is strategically important as it demonstrates Foundry as a serious enterprise platform — not a demo, but a solution that three global banks are publicly testing. The Sibos 2025 demonstration is a signal that trade finance, one of the most conservative segments of the financial industry, is beginning to accept AI agents as productive operators, not just assistance tools.

The next step toward production will require regulatory approvals, integration with existing core banking systems, and — most importantly — industry-wide trust that the agent is reliable enough for documents that drive global trade flows.