Lak and I started Obin because we are witnessing a fundamental shift in how financial institutions operate. The marginal cost of executing complex workflows is collapsing, and in financial services that changes the game: the bottleneck is no longer “people and hours”, but “judgement, governance, and speed.” When you can summon expert capacity on demand, firms can finally pursue the high-value opportunities they’ve been forced to shelve for years, not because they weren’t profitable, but because there simply wasn’t bandwidth to execute.
Think about the real bottlenecks across your institution. Whether you’re managing a private credit portfolio, diligencing investment opportunities, adjudicating insurance claims or underwriting commercial loans, capacity constraints lead to missed opportunities: deals you couldn’t pursue, claims backlogged for weeks, or covenant breaches discovered too late. What if you could expand your effective capacity tenfold without adding headcount? You would compress cycles from days to hours, pursue deals that were previously uneconomic and redeploy your team from process work to high-judgement decisions. The result is a financial institution - asset manager, insurer, or bank - that can deploy capital faster, price risk more precisely, and operate with a level of real‑time insight that used to be invisible when capacity was the constraint. This is the potential of AI in financial services – and it’s no longer theoretical.
But realizing this potential requires solving the hardest problem in financial AI: edge cases. In finance, the "happy path" has little value. The real work lives in the long tail - a distressed credit with both a PIK toggle and a springing lien, a commercial insurance claim involving three jurisdictions and a disputed exclusion clause, or a covenant compliance calculation where the borrower's EBITDA definition doesn't match the credit agreement. Generic AI models hit an "80% Wall" because they're trained on common patterns, not the exceptions that define financial services.
We build AI agents that decompose ambiguous workflows into machine-verifiable steps, then systematically learn each edge case through a combination of domain expertise and architectural precision. When we solve a complex covenant calculation for one client, that logic becomes immediately available across our entire platform.
Financial institutions can't run on black-box models. You need auditability, compliance, and the ability to trace every decision back to its source data. This is where most AI vendors fail. Obin is architected differently from the ground up. Our AI agents operate within a multi-layered governance framework that ensures every output is traceable, auditable, and reversible. But we go further: we built a "Financial Compiler" - a deterministic validation engine that checks agent outputs against ground truth (accounting identities, regulatory rules, covenant definitions) before a human ever sees the result.
To deliver these highly capable and highly secure AI co-workers reliably to our customers, we rely on rigorous context engineering. Every Obin agent operates within a precisely defined context: the documents it can access, the tools it can use, the rules it must follow, and the outputs it can produce. This "context boundary" acts as both a security perimeter and a performance optimizer; agents reason only over relevant information and can't hallucinate facts from outside their scope. The hard part is building this context infrastructure at scale. For an investment fund, that means mapping historical deal precedents, firm-specific underwriting policies, and portfolio company performance data into a structured knowledge graph.
We've industrialized this process through our platform, Robin Studio. Instead of hand-crafting prompts for every edge case, we systematically build the taxonomy, extraction logic, validation rules, and evaluation functions required for each domain. This is the "long pole" of enterprise AI - the hard infrastructure work that most vendors skip - and it's why we can deploy production-ready agents faster than anyone else.
Crucially, our AI agents continue to learn and get better, expanding your firm’s capacity with every deployment.
When we solve a complex corporate action logic or a new document type for one client, that architectural improvement becomes immediately available to the entire ecosystem. You benefit from edge cases encountered across the industry, but how you use that capability - your underwriting criteria, risk appetite, and decision frameworks - remains your proprietary alpha. This creates a flywheel where our infrastructure becomes more robust with every interaction, continuously expanding what your team can handle without adding headcount.
Our client base accelerates this compounding effect. By working with top-three institutions across banking, insurance and asset management, we are able to encounter more of these edge cases faster. While an internal team only learns from its own friction, we see the aggregated complexity of the market across dozens of clients. This allows us to accelerate our learning loop and deliver a level of domain intelligence that no single institution can build alone.
This is the future we’re building.
And we’re just getting started.
Obin AI Founders
Apoorv Saxena and Lak Lakshmanan
