Sava wants trust administration to run itself

Nimit Maru had just done the thing every founder dreams about. He sold his company, Fullstack Academy, for $55 million after raising only $1 million along the way. Then he tried to do the responsible next step: set up trusts for his family.

What he found was an industry that manages more money than most countries' GDP and still runs on spreadsheets, email threads, and PDF attachments. Opening a trust meant weeks of paperwork. Understanding what was inside one meant calling someone and waiting. The infrastructure holding trillions of dollars of American family wealth turned out to be a filing cabinet with a phone number.

That experience became Sava, the company Maru co-founded in 2025 with Rush Sadiwala. Sava calls itself an agentic trust company: a real, regulated trust company in the making, run on AI agents that handle the administrative work humans currently grind through by hand. The pitch convinced Y Combinator, which backed Sava in its Fall 2025 batch, and a group of investors including Gradient Ventures, Instacart co-founder Max Mullen, and SV Angel, who together put in $10 million.

The $6.5 trillion filing cabinet

Trusts are one of the main ways wealth moves between generations in the United States. More than $6.5 trillion sits inside them today, administered by independent trust companies and bank trust departments. Sava estimates the administration work alone is a $25 billion a year revenue opportunity.

The work itself is not glamorous. A trust is a legal wrapper around assets, and someone has to keep that wrapper intact: collect statements from custodians, track what the trust owns and what it's worth, distribute money to beneficiaries on schedule or on request, file the right tax documents by the right deadlines, and document every decision well enough to survive an audit or a family dispute.

Today that someone is a trust officer juggling dozens or hundreds of accounts. The tooling is generic: email for communication, spreadsheets for reconciliation, PDFs for everything official. Fees are high because the labor is expensive, and service is slow because the labor is scarce. A beneficiary who wants to know what's in their trust often gets a quarterly statement and not much else.

None of this is because trust companies are lazy. The work is genuinely fiddly, every trust document is a little different, and the penalty for getting it wrong is legal, not just financial. The industry stayed manual because the work seemed to require careful humans reading carefully. That assumption is the thing Sava is testing.

What an agentic trust company actually is

Sava's answer is not software sold to trust companies. Sava intends to be the trust company, with its charter application pending regulatory approval, and to run the administration layer on AI agents from day one.

The division of labor is specific. Agents do the grunt work: they collect data from custodians and counterparties, chase down signatures, assemble document packages, and reconcile positions across accounts. People do the judgment work, the decisions where a fiduciary's accountability actually matters. The agents' job is to make sure that when a human decision is needed, everything required to make it well is already gathered, checked, and in one place.

A routine distribution request shows the shape of it. When a beneficiary asks for money today, a trust officer has to pull up the trust document, confirm the request is allowed under its terms, check balances across custodians, gather whatever supporting paperwork the file requires, get the approval recorded, and then actually move the money. Each step is short. The waiting between steps is where the weeks go, because every step depends on a busy person remembering to do it. In Sava's model, agents run that sequence continuously: the document terms are already extracted, the positions are already reconciled, the paperwork assembles itself, and the human fiduciary is presented with a decision rather than a to-do list.

That last part is worth underlining, because it is where the "agentic" framing earns its keep. The fiduciary duty stays with people. What changes is that the people stop being the workflow engine. The failure mode of trust administration is rarely bad judgment. It is usually a dropped thread or a missed deadline, and agents are good at exactly that class of problem.

Around that core, Sava is building what the modern version of a trust account should look like: one dashboard to open a trust, fund it, track it in real time, make distributions, and stay compliant. The company describes it as Stripe for fiduciary work, and the comparison is doing real work. Stripe's insight was that payments, a regulated and miserable-to-build capability, could be wrapped in a clean API and offered to every company that didn't want to become a payments expert. Sava's bet is that trust administration is the same kind of capability.

Trusts as an API

That is where the second half of the business comes in. Because Sava treats the trust charter as programmable infrastructure, other companies can build on top of it.

For fintechs, that means embedding trust and estate services directly into their own products. A wealth platform or a banking app could offer customers the ability to create and fund a trust without sending them to a law office and a separate trust company, the same way fintechs today embed payments or brokerage through API providers instead of getting their own licenses.

For estate attorneys, Sava is a faster back office: onboard clients quickly and hit tax deadlines without chasing documents across email. Attorneys still design the estate plan. Sava handles the administrative machinery that follows.

For financial advisors, it means real-time API access to client positions and valuations inside trusts, information that today lives in someone else's quarterly PDF.

Each of these groups touches trusts constantly and none of them wants to run a trust company. If Sava's rails work, the trust stops being a place where information goes to sit and becomes something closer to a live account: queryable, embeddable, and current.

Full stack, on purpose

The strategic choice worth dwelling on is that Sava is not selling this system to incumbents. YC group partner David Lieb described Sava as a startup going full stack, using AI to compete directly with an enormous legacy industry rather than selling software to it.

The logic mirrors what's happening in other professional services. An incumbent trust company that licensed Sava-style automation would be buying a tool that undercuts its own pricing model, which is a hard internal sell. A new trust company built around the automation has no such conflict. Every hour of labor the agents absorb goes straight to margin or to lower fees, and the company can compete on the two things clients actually feel: speed and price.

Going full stack also means taking on the burden incumbents hide behind. A trust charter is a serious regulatory undertaking, and Sava's is still pending approval. That is a real gate, and it is also the moat. Anyone can write software that reads trust documents. Becoming a chartered fiduciary that can legally administer them is the part that takes years, and it is the part that makes the API worth building on.

Founders who match the problem

Sava is a three-person company in New York, so the founders are most of the story right now.

Maru has built and exited a company through YC before. Fullstack Academy, which he co-founded in 2012, went through YC's S12 batch and grew into one of the best-known coding bootcamps before its acquisition. He has also lived the customer problem personally, which is where the company came from in the first place.

Sadiwala, Sava's COO, spent the first years of his career on the buy side in investment management before founding two startups: Zolt, a news app Apple featured among its best of 2016, and Framework.so, a creator platform backed by Kevin Rose that grew past three million end users. The combination matters for this market. Trust administration sits at the intersection of financial operations and software, and Sava's founding team has spent years on each side.

Why trusts fit AI agents unusually well

There's a version of this story that could be told about almost any paperwork-heavy industry, so it's fair to ask why trusts in particular.

The honest answer is that trust administration is close to a best case for current AI agents. The work is document-heavy, rule-bound, and deadline-driven. Most of it is reading, extracting, cross-checking, and assembling, exactly the tasks language models have become reliable at when the workflow around them is engineered carefully. And the industry's service expectations are so low that even modest automation shows up as a dramatically better product. When the status quo is a quarterly PDF, a live dashboard feels like science fiction.

The timing argument is demographic. The largest wealth transfer in American history is underway as trillions of dollars move from baby boomers to their heirs over the next two decades, and a large share of that money will move through trusts. The generation inheriting it grew up with software and will not accept fax-era service as the price of an inheritance.

Sava is small, early, and still waiting on its charter. But it has picked a target with the qualities that tend to reward patience: an enormous pool of assets, entrenched competitors with no incentive to modernize, and customers who have never once enjoyed the current experience. If the agents can do the filing, the $6.5 trillion filing cabinet is up for grabs.