Editor’s note: This article is based on insights from our podcast series. The views expressed in the podcast reflect the views of the speakers and do not necessarily represent the views of this publication. Readers are encouraged to explore the entire podcast for additional context.
Artificial intelligence has gone through many hype cycles, but this time is fundamentally different, according to Stuart Feld, senior vice president at Raymond James. In a recent episode of the CAIO Connect podcast hosted by Sanjay Puri, Feld shared his grounded, empirically based perspective on what it actually takes to deploy AI responsibly and at scale, especially in highly regulated industries like financial services.
What set this conversation apart was not futuristic speculation, but practical insights from someone who is actively leading AI transformation within large enterprises.
Feld makes a compelling point. AI was the first truly revolutionary technology to be adopted by the general public before businesses. Unlike cloud, mobile, and algorithmic trading, AI tools like ChatGPT went mainstream almost overnight, forcing companies to react rather than lead. With billions of users engaging with AI every day, organizations can no longer afford to treat it as an experiment.
This reality shaped Raymond James’ approach and ultimately led to the creation of a dedicated Chief AI Officer role in early 2025.
AI was nothing new to Raymond James. Feld and his team have a long history of delivering machine learning solutions where AI brings tangible benefits, particularly in the areas of back-office compliance and oversight. The difference was reliability. Feld did not advocate AI theoretically. He was using it to solve real business problems.
Read: Inside New York’s AI vision with Shreya Amin (
This track record made the CAIO role a natural evolution rather than an afterthought. Its mission was not about hype, but about scale, governance, and accountability.
One of the most relatable moments in our conversation was Feld’s reframing of the common fear that everything we build today will be obsolete in 12 months. “What we’ve been doing with machine learning since we started five years ago. We’re doing it differently now. It’s just getting better… Proven and true machine learning, some of the things we’re doing with generational AI are going to get even better. Now as a 5.0 model, in a year’s time, we might be in a situation where we’re now at an 8.0 model. ”
What was his reaction? Outdated doesn’t mean useless.
If properly designed, AI systems will improve over time. The core machine learning foundations and capabilities of GenAI will evolve, not be discarded. Experimentation layers such as agent-to-agent communication and model context protocols can change rapidly, but the underlying architecture becomes more valuable.
At Raymond James, Feld says, “We’re not looking to replace advice or judgment. We’re putting humans in the loop. What I mean is, we’re not doing anything that an advisor can’t do. We’re just doing it at lightning speed.”
AI needs to do things that humans are already doing, faster. Advisor completes hours-long tasks in seconds, freeing you up to do more valuable work. AI brings information to the surface, highlights opportunities, and relieves menial administrative tasks, but the final decision always remains with humans.
This “human-involved” approach is non-negotiable, especially in financial services where explainability, traceability and accountability are as important as accuracy.
Feld outlines a simple but powerful trajectory for enterprise AI adoption.
- General surface information
- Surface personalized insights
- Recommended action
- Perform the action with final human approval
Read: Treat AI like a “junior employee”: Lessons from Mark Carmish (
Most organizations, including Raymond James, are now firmly in the first two stages. Once trust, governance and accountability are firmly established, action will take a backseat.
Despite the buzz, Feld points out that very few organizations are running truly autonomous AI agents in production. Most are conducting experiments to test security, scalability, and integration with existing APIs.
For mission-critical functions, Feld has yet to see autonomous agents operating without human supervision. And that vigilance is not just regulatory, but also philosophical.
AI remains undervalued. There are big low-hanging fruit that can deliver immediate value without a moonshot. Feld’s advice is simple and powerful. Focus on actual implementation rather than proof of concept.
Boards don’t want to hear about what AI will one day do. They want to know what you’re doing now.
And as this “CAIO Connect” podcast episode makes clear, responsible, human-centered AI is not slow, it’s the fastest path to sustainable scale.
