Companies like Ramp are developing financial AI agents using a tiered autonomy model akin to self-driving cars (L1-L5). By implementing robust guardrails and payment controls first, they can gradually increase an agent's decision-making power. This allows a progression from simple, supervised tasks to fully unsupervised financial operations, mirroring the evolution from highway assist to full self-driving.
Related Insights
Fully autonomous agents are not yet reliable for complex production use cases because accuracy collapses when chaining multiple probabilistic steps. Zapier's CEO recommends a hybrid "agentic workflow" approach: embed a single, decisive agent within an otherwise deterministic, structured workflow to ensure reliability while still leveraging LLM intelligence.
Frame AI independence like self-driving car levels: 'Human-in-the-loop' (AI as advisor), 'Human-on-the-loop' (AI acts with supervision), and 'Human-out-of-the-loop' (full autonomy). This tiered model allows organizations to match the level of AI independence to the specific risk of the task.
As AI evolves from single-task tools to autonomous agents, the human role transforms. Instead of simply using AI, professionals will need to manage and oversee multiple AI agents, ensuring their actions are safe, ethical, and aligned with business goals, acting as a critical control layer.
Frame AI agent development like training an intern. Initially, they need clear instructions, access to tools, and your specific systems. They won't be perfect at first, but with iterative feedback and training ('progress over perfection'), they can evolve to handle complex tasks autonomously.
Avoid deploying AI directly into a fully autonomous role for critical applications. Instead, begin with a human-in-the-loop, advisory function. Only after the system has proven its reliability in a real-world environment should its autonomy be gradually increased, moving from supervised to unsupervised operation.
To enable agentic e-commerce while mitigating risk, major card networks are exploring how to issue credit cards directly to AI agents. These cards would have built-in limitations, such as spending caps (e.g., $200), allowing agents to execute purchases autonomously within safe financial guardrails.
The evolution of Tesla's Full Self-Driving offers a clear parallel for enterprise AI adoption. Initially, human oversight and frequent "disengagements" (interventions) will be necessary. As AI agents learn, the rate of disengagement will drop, signaling a shift from a co-pilot tool to a fully autonomous worker in specific professional domains.
A useful mental model for AGI is child development. Just as a child can be left unsupervised for progressively longer periods, AI agents are seeing their autonomous runtimes increase. AGI arrives when it becomes economically profitable to let an AI work continuously without supervision, much like an independent adult.
The future of AI in finance is not just about suggesting trades, but creating interacting systems of specialized agents. For instance, multiple AI "analyst" agents could research a stock, while separate "risk-taking" agents would interact with them to formulate and execute a cohesive trading strategy.
The financial system is unprepared for the coming wave of AI agents. These agents will perform tasks and require payment, creating trillions of micropayments. Current infrastructure from Stripe, Visa, or Mastercard cannot handle this volume, creating a massive opportunity for new protocols to facilitate the 'agent economy'.
