DEV Community

Cover image for BNY And Microsoft Let AI Bots Haggle Over A Treasury Bill Trade
Fathom News
Fathom News

Posted on Originally published at fathom.news

BNY And Microsoft Let AI Bots Haggle Over A Treasury Bill Trade

BNY and Microsoft completed what BNY called the first AI agent negotiation of a U.S. Treasury bill trade, a test of whether governed agents can handle market workflows without removing human control.

Key Takeaways

  • BNY and Microsoft completed what BNY called an industry-first pilot negotiation of a U.S. Treasury bill using AI agents
  • BNY's Agentic Negotiation Protocol let the firms' agents exchange offers and counteroffers within predefined rules and approved pricing inputs
  • Human traders reviewed and approved the proposed terms, and existing internal systems handled execution and booking
  • BNY plans to keep testing the protocol beyond the initial bilateral use case and explore repo and other contractual negotiations

BNY announced that its Agentic Negotiation Protocol, or ANP, handled the back and forth between the firms' agents and checked a central ledger before proposed terms went to traders.

Microsoft Treasury gave its agent specific trading instructions, while BNY controls checked the counterparty, security, size, settlement terms, and available pricing before prices could be negotiated within authorized limits.

How ANP Worked

BNY designed ANP as a control layer for authorized agents from different organizations to negotiate securely and transparently. It verifies participants, lets them connect securely, and lets agents exchange offers and counteroffers until they reach a shared understanding of the proposed terms.

Check Who Checks It
Counterparty Validated by BNY controls
Security Validated by BNY controls
Size Validated by BNY controls
Settlement Terms Validated by BNY controls
Available Pricing Validated by BNY controls

Participant-side controls use predefined rules and AI safety checks tied to approved pricing inputs. Before trader approval is requested, ANP checks that both parties' agents submitted matching terms.

Also Read: Coinbase Lets Anyone Trade Stocks And Crypto Right Inside ChatGPT

Human Authority Stayed In Place

Human traders reviewed and approved the proposed terms, and the trade was executed and booked through the firms' existing internal systems and controls. Agents could not execute the trade for either firm.

The process keeps a record of who took part, the messages exchanged, the control checks, and the human decisions so the negotiation can be reviewed and reconstructed. BNY said the approach is meant to streamline the traditional message-based process while preserving existing risk and transaction-processing controls.

BNY said it serves over 90% of Fortune 100 companies and nearly all the top 100 banks globally. As of Sept. 30, it oversaw $63.4 trillion in assets under custody and/or administration and $2.2 trillion in assets under management.

What Comes Next

BNY plans to keep testing ANP, expand beyond the initial bilateral use case, and evaluate additional negotiated workflows.

Over time, future applications could include repo and other contractual negotiations. BNY did not give dates for those next tests.

Read Next: Microsoft Teams With Lumen To Keep Bank Data Off Web


Originally published on Fathom News.

Top comments (0)