Last week at Wealth Management EDGE, Subatomic CEO & Co-Founder Sam Sova sat down for an in-depth interview exploring how AI coworkers are transforming wealth management firms from the inside out. The conversation covered everything from foundational data strategy to the real costs of running agentic AI at scale.
Here are the key takeaways from the interview — and why they matter for RIA principals evaluating AI adoption.
AI Coworkers, Not Just Software
Sam opened the conversation with a provocative question: Are you going to keep buying software, or start hiring AI?
The distinction matters. Traditional software automates a specific task with rigid rules. AI coworkers are different — they are agents trained on your firm’s actual workflows, data, and best practices. They learn how your team works and operate alongside your people, handling the repetitive work that drains advisor time.
This is not about replacing people. It is about giving every advisor on your team an AI-powered Chief of Staff that handles the operational overhead so they can focus on what they do best: building client relationships.
Observability: You Can’t Trust What You Can’t See
Before deploying agentic AI, Sam stressed that firms need to be able to observe three critical areas:
- What data the AI accesses — Understanding which client records and systems the AI touches is foundational to compliance and trust.
- How outputs are generated — Black-box AI is a non-starter in a regulated industry. Firms need visibility into how the AI reaches its conclusions.
- How to track and optimize token-related costs — Every AI interaction has a cost. Without tracking, expenses can spiral quickly.
This is exactly why Subatomic built Deep Lens — a monitoring layer that gives firms real-time visibility into what their AI agents are doing, what data they are accessing, and what it is costing them. Think of it as compliance and cost control built directly into the AI infrastructure.
Token Costs and the Case for LLM Flexibility
One of the most practical segments of the interview addressed tokenization costs and model risk. Sam argued that firms should avoid locking themselves into a single large language model. The AI landscape is evolving rapidly — what is the best model today may not be tomorrow, and pricing structures are shifting constantly.
Subatomic takes an LLM-agnostic approach, routing different tasks to different models based on the optimal balance of quality and cost. A simple data lookup does not need the same horsepower as a complex portfolio analysis. By matching the right model to the right task, firms can dramatically reduce their AI operating costs without sacrificing output quality.
Data Has to Come First
Perhaps the most important takeaway from the entire interview: successful AI deployment depends on foundational data work.
Sam explained that most firms have their data scattered across CRMs, portfolio management systems, custodial platforms, and email inboxes. Before AI can work effectively, that data needs to be normalized — structured and unstructured sources reconciled so the system can correctly identify clients across every touchpoint.
Subatomic uses a medallion data architecture to handle this: raw data is ingested, cleaned, enriched, and organized into a unified layer that AI agents can reliably work from. Without this foundation, even the most sophisticated AI will produce unreliable results.
Think Org Chart, Not Tech Stack
In what may have been the most memorable moment of the interview, Sam reframed how firms should think about AI adoption entirely. His advice: stop treating AI as a technology decision and start treating it as an organizational one.
When you evaluate a new hire, you think about what role they will fill, what skills they bring, and how they will integrate with the existing team. The same framework applies to AI coworkers. Which workflows need support? What does the AI need access to? How does it fit into your existing operations?
This organizational mindset is what separates firms that get real value from AI from those that end up with expensive shelfware.
The One Question Every RIA Principal Should Ask
Sam closed with a challenge to every RIA principal in the room: the question is not whether to adopt AI — it is whether you are building the foundation to do it right. That means getting your data house in order, choosing partners who give you visibility and control, and thinking about AI as a long-term team member rather than a quick technology fix.
Jump to a Topic
- 0:00 — AI Agents Explained
- 0:43 — The AI Coworkers Concept
- 1:12 — Tracking Agents and Costs
- 1:34 — Deep Lens Monitoring
- 2:19 — Token Costs and Model Risk
- 3:55 — Data First for AI Wins
- 5:16 — Medallion Data Architecture
- 6:10 — Organizational Chart, Not Tech Stack
- 7:50 — The One Question to Ask
Ready to see how AI coworkers can work for your firm? Get in touch with the Subatomic team to learn more.