Agents Prompt People: Using Agents for Long-Horizon Financial Work
· Source: Latent Space
Arca, a wealth‑management firm that brands itself as “AI‑native,” is deploying artificial‑intelligence agents to automate much of the back‑office work in financial planning, while keeping strict controls that require human advisors to make the final decision. According to Aman Singh, senior AI engineer, the agents can handle tasks that span up to ten days, with multiple decision points and intermediate deliverables. To maintain continuity, the company has built an internal memory system that records each client’s status, short‑ and long‑term pending actions, and the responsibility of each operator, ensuring that information persists throughout the process.
Arca uses models from vendors such as Anthropic and OpenAI, with Claude as its primary model, and has developed its own “harness” that automatically selects the most suitable model based on internal metrics. This proprietary architecture allows the firm to switch providers without interruption. Regarding security, the agents only perform analysis and propose solutions; any trading activity or movement of funds requires a human advisor’s authorization. Security constraints are deterministic and are complemented by sandboxed code execution and prompt‑injection detection mechanisms.
This story is noteworthy because it demonstrates how AI can accelerate complex processes in the financial sector while preserving safeguards that protect clients, offering a model that other firms could emulate to combine technological efficiency with human responsibility.
Read the original article on Latent Space
This summary is an informational synthesis produced by dataqbs.com. All rights to the original content belong to its author and the cited media outlet. We act solely as curators of technology news and claim no authorship.