How AI Agents Are Reshaping Long-Term Wealth Management

Wealth management has always involved long chains of research, follow-up, review, and decision-making. Arca is building an “AI-native” approach around that reality, using agents to handle much of the background work while human advisors remain responsible for final decisions.

Aman Singh, lead AI Engineer at Arca, described the company’s approach during a 26-minute Latent Space podcast episode listed on Oct 31. He also presented at AI Engineer New York this month. The central idea is simple: AI should guide the process, but it should not control the movement of money.

Agents That Track Work Over Ten Days

Arca’s agents can work on tasks that take up to ten days, a long window compared with the short exchanges people often associate with chatbots. Singh said the firm’s workflow is built around the idea that “agents prompt people more than people prompt agents.”

That changes the role of an AI system. Instead of waiting for an advisor to ask every question, an agent can track open work, identify what needs attention, and propose the next step. The advisor still makes the decision, but the agent keeps the process moving in the background.

“It’s very stop-start, there’s many actors in between, many decision points, many handoffs,” Singh said. Arca’s response is an internal memory system that keeps a persistent record for every client.

“So we have this persistence for every client,” Singh said. The system tracks what has happened, what needs attention over the next few days, what belongs to the long-term plan, and which person is responsible for each action.

Singh described the memory system in practical terms: “What has happened so far? What are we tracking for next? What needs to happen immediately in the next couple of days [and] what are we tracking for the long term? Which individual is responsible for which action?” The agents contribute to that record, giving the firm a shared view of ongoing work.

Arca recently raised $48.5 Million in a round led by General Catalyst, with participation from Index Ventures and Venrock. The company provides wealth management services and is building its infrastructure around these longer-running agent workflows.

Models, Guardrails, and Human Control

Arca uses models from frontier labs such as Anthropic and OpenAI, with Claude as its primary model provider. The company also built its own harness, which chooses the best model for a task using Arca’s internal benchmarks.

Owning that harness gives Arca control over how models are tested and assigned. It also allows the company to change model providers if needed, instead of depending on a managed agent harness.

The agent’s role remains limited to analysis and proposing solutions. Trading and any other money movement are handled by a human advisor. Agents do most of the background work, but guardrails ensure that advisors make the final decisions.

“All of our safety restrictions are fully deterministic,” Singh said. “And they take the decision making off the agent’s hands.” That distinction matters in financial work, where an incorrect suggestion and an unauthorized transaction carry very different risks.

Arca also runs verification checks for known AI security issues, including prompt injections. The agent itself is not sandboxed, but Arca uses sandboxes for computation and has built systems that let code inside the sandbox access company data safely.

Singh explained that Arca translates code written by an agent into a strict format before sending anything to the back end. The original code is discarded, and guardrails are applied to the approved format. Agents can still use code to query, aggregate, and work with data to produce items such as a financial plan.

A “meta-agent” reviews the data from these systems and suggests changes to the harness. That creates a feedback loop around model selection, testing, and the way agents handle tasks, while the firm keeps safety restrictions outside the agent’s own decision-making.

The Advisor’s Work Is Changing Too

Arca’s approach fits into a wider discussion about autonomous AI and financial services. AMCAP Global Capital Management wrapped up its Internal Intelligent Finance Summit at its headquarters in New York. The symposium focused on the operational convergence of autonomous AI and global capital markets, while AMCAP Global’s strategic roadmap is to pioneer agentic financial infrastructure amid shifting macroeconomic dynamics.

The technology is also reaching individual wealth advisors. Rick Nott, a senior managing director at Angeles Investments and a private wealth advisor at Santa Monica, California-based Angeles Wealth Management, spoke about using AI and his reaction to AI-native financial planning platforms.

“When Claude first rolled out its solution for financial advisors, I felt fear and excitement. How complicated will it be? How will it work?” Nott said.

He does not describe himself as a software developer. “I’m not a software developer, but I have some basic tech skills,” he said. Large language models helped him explore coding that he had avoided because learning it required a lot of work.

“LLMs have allowed me to dig into coding, which I’d been avoiding learning because it’s a lot of work,” Nott said. He now uses that knowledge to automate more of his process, while still checking whether the results meet his needs.

“I know enough to know what I want, how it should look, and how to check the outputs, and now I can automate more of my process,” he said. For Nott, AI connects systems and tasks that once sat apart: “AI has become the glue between the various systems we use and the different workflows we have.”

That shift matters because financial work can contain large amounts of repetitive reporting. Nott started his career while still in college in the late 2000s doing data reporting work for a wealth firm. That work amounted to “750 to 1,500 hours a year, most of it crammed into the few weeks after each quarter closes.”

The promise of agentic finance is not to remove the advisor from the process. It is to move routine tracking, analysis, and coordination into systems that can keep working across longer time periods, while people retain responsibility for judgment and money movement.

Stay connected via Google News

Follow us for the latest AI News