Why I deciced to start a consultancy combining Finance & Risk knowledge with GenAI
From Elegant Models to Messy Reality
When I studied economics—and later quantitative finance—we were trained to believe that nearly every economic phenomenon could, in principle, be modeled. We began with beautifully simple frameworks: clean equations, explicit assumptions, and the familiar comfort of ceteris paribus. As our coursework advanced, the models became richer. Assumptions were relaxed, stochastic terms were added, and calibration techniques improved. Each new generation of models claimed to inch closer to “reality.”
Yet as my career unfolded, the distance between modeled worlds and operating reality became impossible to ignore. Macroeconomic indicators and market prices are not exogenous inputs handed down from a neutral universe; they are emergent outcomes of millions of human decisions—some strategic, some reactive, many emotional. Behavioral finance has reminded us repeatedly: people are not consistent utility maximizers. Incentives differ. Information is uneven. Attention is scarce. One human mind is complex; a market made of millions is an adaptive, shifting ecosystem.
What Traditional Quantitative Methods Miss
Early in my consulting work I leaned heavily on quantitative toolkits. They are powerful—indispensable, in fact—when the problem is well-posed, the data structured, and the decision surface reasonably stable. But I saw diminishing returns when clients faced problems that were:
Human-in-the-loop by design (credit exceptions, policy overrides, complex approvals).
Document- and conversation-heavy, where context lived in emails, memos, meeting notes, or regulatory correspondence.
High-variance edge cases that rule-based systems either rejected, routed incorrectly, or escalated endlessly.
I shifted my approach. Instead of forcing problems into models, I started with people—talking to process owners, tracing decision paths, and pinpointing where data could support human judgment, not replace it. Models stayed in the mix, but as part of a broader, human-focused decision framework. This consistently outperformed rigid quant-only solutions, even if it ruffled feathers among model purists.
Enter Generative AI: A Different Kind of Toolkit
My perspective changed dramatically once I began working hands-on with generative AI models and, later, agentic frameworks. What struck me first was not the novelty of chat-style interfaces but the models’ ability to work with unstructured information at scale. Risk and finance functions are dense with text: policies, model documentations, regulatory guidances, client correspondence, audit findings, product term sheets, and board materials. Historically, much of this knowledge was trapped in silos and difficult to merge together.
More advanced models go further. They can classify cases, draft analyses, check against policies or regulations, and suggest next steps with minimal human input. This isn’t just about chatbots—it’s about supporting complex tasks like exception triage or decision-making, areas once reserved for experienced analysts.
Where the Value Lies
The biggest gains aren’t in flashy, fully automated systems. They’re in streamlining high-friction, repetitive tasks that bog down finance and risk teams. Examples include:
Credit & Counterparty Risk Review: Extract key exposures, collateral terms, and covenant triggers from heterogeneous documents; flag inconsistencies before committee.
Regulatory Change Mapping: Link new guidance to internal policies, procedures, and control owners; highlight required updates and ownership gaps.
Model Governance Documentation: Summarize validation reports, track assumptions, and auto-generate change logs tied to regulatory templates.
Limit Exceptions & Escalations: Classify requests, compare to precedent, and draft decision memos with traceable reasoning.
These are not science projects but real, everyday pain points in corporates, banks and insurers.
Why I Started AIBreaker
Knowing the challenges my consulting clients face, and understanding the opportunities GenAI offers, I see an extraordinary amount of value waiting to be unlocked. This realization is what led me to start AIBreaker, focused on consulting for finance and risk, with generative AI at its core. Our mission is simple: to help finance and risk functions work smarter by combining deep domain knowledge with GenAI-driven tools.
Looking Forward
We are early in the adoption curve. LLMs will continue to evolve, new technological frameworks will appear. That is the journey I want to partner on. I’m excited to work with others who see the same gaps between how processes are running today and what new technology can offer to optimize them. If your share this approach, I’m sure we’re going to have a great time working together.