Navigating the Dynamics of Artificial Intelligence

John Chan is a Director of Technology at Raymond James Financial running the Carillon Labs – the innovation labs specializing in AI/ML. His passion is to promote AI adoptions and implement machine learning solutions in the financial sector. He has over 20 years of experience leading and implementing technology solutions from FinTech startups to top-tier banks and consulting firms. Before Raymond James, John was an artificial intelligence (AI) strategist and engineering lead at Gamma Lab of OneConnect Financial, Morgan Stanley data science team and KPMG Cognitive Technology Lab. He is active in NLP research focusing on Generative AI, Conversational AI, Document Understanding and risk and compliance technology. He is a frequent speaker at AI events.

1. Can you share some insights into your professional journey and the key experiences that led you to your current role at Raymond James?

I started my professional journey as a data analyst. About 10 years ago, I reinvented myself in data science and AI. I believe AI is the most significant revolution in human history, surpassing what the Industrial Revolution has brought to us.

Solving business problems and finding ways to deliver more effective solutions, I have had the privilege of collaborating with smart, talented individuals and forming strong teams capable of tackling challenges. It is important to combine art and science, understanding the complex dynamic of people and change while aligning visions with strategies and pursuing engineering excellence. These experiences and passions have collectively shaped my path, leading me to my current role.

2. What about the most significant trends and advancements in AI/ML that you believe will impact the financial industry, particularly in wealth management and investment banking?

The new wave of the AI revolution is just starting. There is no doubt that it is going to be more impactful to the financial industry or any industry that requires using language to make decisions or create content. Last year, we saw industries flooded with GenAI topics. Presently, we have RAG added to the mix, targeting some shortcomings of GenAI or LLM in general.

  To achieve AI execution excellence, we need rapid strategies and tactical combat decision-making mindsets, similar to the venture mindsets to win business

The next year is likely to be Agentic AI, where many RAG/LLMs will be working together to push for better results. With vibrant activities in the research community and abundant funding pouring into AI, I believe the AI model accuracy will soon blow our minds. Not only the wealth management and investment banking industries but virtually any industries that require timely communication and are heavy on documents will be transformed by AI/ML, specifically by GenAI.

 

3. What do you see as the biggest challenges for AI/ML in your industry over the decade, and how are you preparing to address them?

I think the hardest thing is to navigate with agility over bureaucracy. We are experiencing unprecedented shear velocity challenges. Highly regulated industries are used to moving slowly. The lack of speed to cope with the speed of AI changes will expose vulnerabilities, giving competitors opportunities to steal market shares.

The old model is Try-Slow, Fail-Slow, and does a lot of storytelling, driven by multi-year roadmaps and thus cannot keep up with the speed no matter how much tweaking on plans. To achieve AI execution excellence, we need rapid strategies and tactical combat decision-making mindsets, similar to venture mindsets to win business. Concretely, we need organizational commitment to change, starting from executive sponsorship and moving from top down on fast tracks.

 4. What does your current AI/ML team look like in terms of roles and expertise? How do you ensure the team has the necessary skills to stay ahead of the curve?

We have the AI solution team actively interfacing with businesses or clients to align our technologies with the business objectives, and the data teams gather, cleanse, analyze, understand and augment data. The AI engineer and AIOps team are responsible for coding, training, finetuning, testing and producing AI services. We also have the teams for the model risk, including explainability, privacy, governance and AI ethics.

To ensure the team stays ahead of the curve, I start with a comprehensive screening when building teams, ensuring they can demonstrate immediate technical skills and be culturally fit. They have to be passionate about using AI/ML to solve real-life problems and want to have fun while working hard, and have the motivation to stay ahead of the curve.

AI is advancing fast to the point that it is quite difficult to keep up. I encourage my team to learn new things specific to their domain expertise, read blogs and technical papers often, and share their new knowledge with their teammates to sharpen each other.

 5. Can you describe the AI/ML technology stack currently in use at Raymond James? What considerations went into selecting these tools and frameworks?

I am quite agnostic to technology stacks. Many stacks can make things work. I just make sure the broader team can agree on the tools and frameworks that we can stick with for the long term. Many cloud providers have comprehensive AI solution stacks. Most of my work is utilizing the AWS cloud ecosystem. I am sure AzureAI, GCP, etc. are as good. For development, I am a python/torch person. I like Nvidia DGX, especially for deep model training.

6. What advice would you give to other financial services firms looking to adopt AI/ML technologies? What are the key considerations and potential pitfalls they should be aware of?

Many firms have initiated AI/ML for quite some time, some firms are pretty mature, especially in classic ML. However, GenAI requires different skills and approaches to apply as it targets knowledge work automation and making data-informed decisions. Organizations need to have a defined AI Strategy that can support long-term digital transformation with strong executive sponsorship.