What does it mean for an AI system to be agentic? Agentic Intelligence: AI Methods for Bond Markets answers this through modeling one of modern finance’s hardest problems: Government bond market trading. As structural pressures and regulatory concerns mount, traditional approaches fail in these complex, partially observable and bilateral trading environments. Following unprecedented pandemic-era government debt issuance, authorities including the Reserve Bank of Australia, Bank of Canada, Bank of England, and Federal Reserve Bank of New York have prioritised enhancing the market’s liquidity and functioning. This book responds directly by building Agentic AI models – showing that Agent-Based Models (ABMs) provide the natural computational foundation for transparent market modelling, and that when LLMs are embedded with discipline, Agentic AI can uncover emergent features and behavioural realism that traditional models cannot. Dynamically adaptive and able to operate over a virtually infinite space of possibilities, Agentic AI offers a new paradigm. The author makes three practical contributions: (1) a liquidity-centred framework that treats the ability to transact as the central systemic object; (2) Agentic AI methods that capture heterogeneous participant interaction, non-linear cost structures, and emergent liquidity; and (3) the AI-simulator testing paradigm, in which the simulator is the laboratory and the agent is the experiment – providing market participants and regulators a basis on which Agentic AI can be deployed, governed, and trusted. Readers will gain implementable methodologies for moving Agentic AI from concept to market application, supporting trustworthy AI in the trading, supervision and design of financial markets. I have spent five decades writing about fixed income markets, and in that time the analytical toolkit has evolved considerably, but the fundamental challenge of modeling liquidity in bilateral, over-the-counter markets has remained
...
Read Full Text
