Corporate finance teams are exploring new architectures to manage AI agent spending without creating thousands of individual bank accounts. The proposed model separates actual liquidity from the authorization to spend, allowing treasury departments to maintain centralized control while providing agents with programmable spending authority.
Virtual cards are emerging as a primary tool for agentic payments, as they allow for programmatic generation, merchant restrictions, and transaction limits. This approach enables companies to utilize existing payment networks like Visa and Mastercard without needing to hold separate capital for each autonomous agent.
Stablecoins offer potential for high-frequency, small-scale transactions, such as purchasing compute or API access. Experts suggest a model where treasury maintains a central balance, while agents operate via virtual wallets or sub-ledgers, ensuring capital efficiency through centralized liquidity.
Recent data highlights the adoption of these technologies. A September 2025 report indicated that 7% of US enterprise CFOs have deployed agentic AI in finance workflows, with another 5% running pilots. Meanwhile, a readiness gap persists, with 75% of tech firms reporting familiarity with agentic AI, compared to 33% of goods firms and 38% of services firms.
