Lets Connect
At 11:47 p.m., somewhere between Mumbai and New Jersey, a mid-level analyst is still staring at a spreadsheet.
Her day job is simple to describe and impossible to automate – at least that’s what everyone around her believes. She chases missing documents, rekeys data from PDFs into policy systems, reconciles three versions of the same report, and writes emails whose main purpose is to ask other humans to fix things in other systems.
On paper, she works in “operations” for a large insurance company. In reality, she works inside the invisible factory that keeps the global financial system from stalling.
Multiply her by a few million, stretch the work across every bank, insurer, asset manager, payments processor, and broker, and you start to see the real shape of financial services: a $36 trillion revenue machine that runs with a cost-to-income ratio of 40–50%—the highest of any major industry.
A huge chunk of that cost is not branches or marketing or fancy trading floors.
It’s people in the middle and back office.
This is the part of finance we’re obsessed with.
Not the glossy app on your phone, not the shiny card in your wallet, but the messy, manual, institutional memory behind it all.
And we believe that in the next ten years, that invisible factory will be run by an AI workforce.
The financial industry’s biggest secret: the hidden factory
If you walk through a large insurer’s operations floor today, you won’t see robots or assembly lines.
You’ll see:
• Claims handlers jumping between five systems to approve a simple claim.
• Underwriters reading 200-page engineering reports and manually copying a few fields into rating engine.
• Finance teams downloading CSVs from three different systems, reconciling them in Excel, then pasting the final numbers into PowerPoint for the monthly board pack.
• Compliance analysts scanning PDF regulatory notices and trying to guess which policies or customers need to be updated.
All of this is work about work. It’s not pricing risk, serving customers, or designing new products. It’s gluing together brittle systems that grew like coral reefs over decades.
It’s “copy-paste capitalism.”
The financial system is, in theory, code and capital.In practice, it’s humans shuttling information between systems that can’t talk to each other.
Every time a regulator changes a rule, or a new product is launched, that invisible factory gets more complicated. More spreadsheets. More exceptions. More manual checks. More headcount.
For years, the answer has been:
• “Let’s add another system.”
• “Let’s hire another BPO vendor.”
• “Let’s throw more people at it.”
Now we have an industry where IT spend is ~$1Tn a year and majority of that goes not into building the future, but into maintaining legacy systems that cannot evolve fast enough.
We’re not just paying to run operations. We’re paying to remember how the system works.
That’s the problem we’re building for.
Why “AI copilot” is not enough
It’s tempting to look at this mess and say, “We’ll give everyone an AI copilot.
They can type questions in chat, and the bot will summarize documents.”
That’s not even close to what’s needed.
Summaries don’t move money.
Summaries don’t clear exceptions, file bordereaux, reconcile ledgers, or push updates into the core policy admin system.
The real bottleneck is not knowing what to do. It’s doing it – across dozens of systems, governed workflows, and audit requirements.
What we need is not one more chatbot.
We need an AI workforce:
A network of specialized AI agents that can understand human intent, plan multi-step workflows, take actions across systems, and leave a clean audit trail for regulators and risk teams.
Think of them as autonomous colleagues who:
• Read documents and extract what matters.
• Log into systems (through secure APIs).
• Push and pull data across the stack, and take actions autonomously.
• Ask for help only at the moments where human judgment is truly required.
Not a “copilot” that rides shotgun but a workforce that can drive.
Why now: AI as the new substrate for operations
Every few decades, a new layer of technology forces its way into the calcified layers of financial infrastructure – mainframes in the 1970s, relational databases and Excel in the 1990s, cloud and APIs in the 2010s. AI – specifically large language models and agentic systems are different. It’s not just faster coding or better search. It’s software that can:
• Read unstructured information like a human.
• Reason across messy, partial data.
• Orchestrate multi-step workflows across tools.
• Learn and adapt from feedback.
Language is now an interface.
Systems can finally “understand” the policy wording, the email thread with the broker, the engineer’s report, and the regulator’s circular – without months of manual rule-writing.
Agents can act, not just answer.
We now have the primitives to let AI systems plan, execute, and verify work across multiple tools and data sources, not just generate text.
This unlocks a fundamentally new question for financial institutions:
“What if our internal operations were run by a digital workforce that never forgets, never loses context, and scales with compute instead of headcount?”
Why we’re starting with insurance
Insurance is the purest expression of this problem.
The product is a promise: we’ll be there when something bad happens.
To price that promise and keep it, insurers run some of the most complex middle and back-office workflows in the world:
• Commercial property underwriting that requires combining historical loss runs, SOVs, catastrophe models, engineering surveys, and broker submissions.
• Claims journeys that touch loss adjusters, third-party administrators, repair networks, lawyers, and reinsurers.
• Finance, actuarial, and reinsurance operations that must reconcile premiums, claims, and reserves across multiple entities and treaties.
Every step is document-heavy, governed, and audited. Every step is also filled with humans doing work that looks like it could be automated – but historically never has been.
In an $8 trillion market, where more than $350 billion is spent on middle and back-office people cost, the AI workforce is not a “nice-to-have.” It’s a margin and solvency question.
That’s why we’re starting here.
If we can build an AI workforce that can safely run property underwriting, claims operations, and finance for insurers, we can generalize it across the rest of financial services. Insurance is our proving ground.
The Economic Prize: Compressing the cost-to-income ratio
When you compress middle and back-office cost, strange things happen to an income statement.
At the industry level, if we can move financial services from a 40–50% cost-to-income ratio toward 25–30%, we are unlocking trillions of dollars:
Higher profitability and ROE.
Lower prices and better coverage for customers
Capital freed to invest in new products, new markets, and resilience.
In a world of compressed yields and rising regulatory expectations, this is not a “nice margin improvement.” It’s existential. Because in finance, efficiency is margin. And margin is market cap.
The Operating System for Money Is Being Rewritten
We have to stop thinking of AI as “a feature” inside existing systems and start thinking of it as a workforce layer that sits on top of all systems.
AI won’t replace finance professionals. But institutions that deploy AI-first systems will replace those that don’t.
A Note to Our Partners
We know this sounds ambitious.
But so did cloud in 2005.
At that time, betting that the world’s largest banks and insurers – conservative, regulated, risk-averse – would eventually run their core systems on other people’s computers looked insane.
Today, it looks inevitable.
We believe the same will be said, a decade from now, about AI workforces in financial services.
Ten years from now:
• No serious institution will run operations without an AI workforce.
• The cost-to-income ratios we accept today will look archaic.
• The line between “technology company” and “financial institution” will be meaningless. Finance will run on AI by default.
We’re building the company that makes that inevitable future practical, safe, and adoptable – starting with the most complex, most operationally intensive corner of finance: insurance.
If you’re a reading this, here’s the simple version:
We’re going after a multi-trillion-dollar cost line that no one can afford to keep paying, with technology that is finally ready to replace manual work, not just assist it.
And we’re doing it in a way that compounds:
• Every workflow automated becomes reusable IP.
• Every institution onboarded deepens our understanding of real-world operations.
• Every AI agent deployed becomes a node in a growing AI workforce for global finance.
We’re not just betting that the future of finance is autonomous. We’re building the workforce that will run it.