Wilson Chan, Founder and CEO of Permutable AI – Interview Series

Wilson Chan is the Founder and CEO of Permutable AI, a London-based company specialising in real-time global data and sentiment intelligence for financial institutions. With a background in AI, financial markets and data engineering, Wilson builds systems that translate vast information flows into explainable, actionable insights. He is a frequent commentator on AI innovation in financial markets, market sentiment and the future of data-driven decision-making, advocating for technology that enhances human expertise rather than replaces it.

You began your career at Merrill Lynch and Citi before founding Permutable AI. What inspired you to transition from traditional finance to building an AI company focused on market intelligence?

I began my career as a quant trader at Merrill Lynch on the fixed-income derivatives desk, where even then we were experimenting with early machine-learning techniques to speed up arbitrage across the yield curve. Most of finance was still in a “software 1.0” mindset – hand-coded rules and manually tuned models.

Over the past decade, we’ve seen a rapid progression: first to “software 2.0,” where neural networks replaced explicit logic, and now into “software 3.0,” where large language models can directly reason over huge volumes of unstructured data. Seeing those shifts happen from the inside made it obvious that markets would eventually be shaped by AI-native systems capable of interpreting global information faster and more holistically than any traditional stack.

Permutable AI was created to capture exactly that opportunity – building modern LLM and multi-agent systems that anticipate market shifts in real time instead of reacting after the fact.

Permutable AI’s mission to “anticipate, not just react to, market change” is compelling. How did this philosophy shape the early design of your vertical LLM architecture?

We’ve built reasoning models designed to understand relationships and causality, not just correlations. The architecture adapts continuously to new macro conditions, news flows, and geopolitical developments. That adaptive layer is core to our philosophy: if you want to anticipate markets, the system has to evolve as fast as the world itself.

Can you explain how vertical large language models differ from general-purpose LLMs like GPT, and why they’re better suited for financial and commodities markets?

At Permutable, we build multi-agent stacks designed to work collaboratively and perform end-to-end workflows traditionally handled by entire teams. We’re confident they already outperform much of the market (see our 1 year results), but we also believe the best results come from combining these agents with highly skilled engineers and domain experts.

A vertical LLM embeds financial ontology – entities, flows, supply chains, macro-drivers, correlations.  Its outputs aren’t just text, but structured decisions: drivers, impact levels, confidence scores, so purpose-built for markets.

Many institutions are struggling to modernize their analytics infrastructure. How does your reasoning-based, adaptive architecture replace legacy systems in a practical way?

We’ve built LLM systems with self-evaluation and continuous monitoring at their core. The aim is to keep human oversight focused and minimal, while still ensuring reliability. The reality is that innovation inside large institutions is incredibly difficult because cultural and structural blockers often get in the way.

Where organisations have the right culture and leadership, the transformation is dramatic. Ultimately, those that embrace this shift will accelerate; those that don’t may struggle to stay relevant.

The Trading Co-Pilot is an exciting development. How does it leverage real-time sentiment and macroeconomic data to give traders an edge?

Our system scans hundreds of thousands of articles in real time and surfaces the ones that actually matter – complete with analysis generated almost instantly. It goes far beyond what a standard search-enabled LLM can produce. It’s essentially a live reasoning engine that sits beside every trader, constantly updating its read on the world.

Explainable AI is becoming increasingly critical in regulated industries. How does Permutable ensure transparency and accountability within its models and decision outputs?

Our key differentiator is that every model output is fully traceable down to the exact article, timestamp, and source. We reduce hallucinations by tightly controlling the task boundaries of each model. Everything comes with a built-in audit trail with  transparency embedded into the core architecture.

Your partnerships now span data providers, trading platforms, and analytics firms. What does an ideal strategic partnership look like for Permutable, and how do these collaborations enhance your global reach?

We look for partners aligned with Permutable’s long-term vision: bringing AI-driven, real-time market intelligence into the core decision flows of global markets. The ideal partner has international reach, strong ecosystem credibility, and the ability to help scale our intelligence and insights across multiple regions and asset classes.

You’ve mentioned the goal of building a “world model for capital markets.” What would such a model look like in practice, and what challenges must be overcome to achieve it?

A world model effectively maps and understands how tradable asset prices interact and influence one another – whether that asset is sovereign debt, FX, commodities, or even something as specific as the price of coffee. It’s a unified representation of global market dynamics.

How do you see adaptive AI transforming the speed and accuracy of decision-making for asset managers, hedge funds, and other financial institutions over the next five years?

Whenever I’m invited to a board-level meeting, I know within the first minute whether that organisation is capable of going through an AI transformation. Culture is everything.

It’s a fact that multi-modal AI will unify reports, news, images, flow data, and price signals into a single reasoning layer. And that hedge funds will move even further ahead because they adapt faster. But many large trading organisations alongside the brightest and best still have teams resisting innovation – they simply don’t realise how quickly the landscape is shifting.

Finally, as AI continues to reshape trading and analytics, what excites you most about the next frontier for Permutable AI and the broader fintech ecosystem?

We believe multi-agent systems will become the dominant framework, though the tooling is still maturing. What excites me most is that the winners in trading and analytics will be the institutions most willing and able to adapt – and that’s exactly where Permutable’s vision sits.

Thank you for the great interview, readers who wish to learn more should visit Permutable AI.