Mission
Handle,
The Last Back Office
The largest industry on Earth runs on paperwork. We are building the layer that takes it.
The opportunity
Insurance is the largest industry on Earth that nobody thinks of as an industry. More than seven trillion dollars of premium moves through it every year.
If insurance were a country it would have a larger economy than every nation on the planet except the United States and China. It touches every car, every building, every cargo, every hospital bed, every human life that anyone ever bothered to protect. It is the machine that lets civilization take risk.
And it runs on paperwork.
Three million people work in insurance in the United States alone, 1.4 million of them inside agencies and brokerages. Globally the brokers and agents who intermediate this thing collect around $642 billion a year in commissions. Behind that number is a vast, invisible workforce of people who spend their days logging into carrier portals, downloading PDFs, retyping policy numbers, chasing receipts, and reconciling spreadsheets against other spreadsheets.
Roughly a quarter of every premium dollar the industry collects gets spent on the cost of running itself. Call it a trillion and a half dollars a year, globally, paid to move documents from one system to another.
- Premium / year
- $7T+
- US insurance workers
- 3M1.4M in agencies
- Commissions / year
- $642B
- Cost of running itself
- $1.5T~25% of premium
We are going to take that number to zero.
Insurance is a document business with a balance sheet attached.
This is the thing most people miss about the industry, including most people inside it.
Insurance doesn't sell money. It sells a recorded promise. A policy is a document. An endorsement is a document. A claim is a document. A premium is a receipt against a document.
The entire seven-trillion-dollar apparatus is, mechanically, an information-processing industry that happens to hold capital.
It started that way. Underwriters wrote their names beneath the terms of a voyage in a London coffee house in the 1680s, and the fundamental unit of the business — a document that says who owes what to whom when something goes wrong — has not changed in three hundred and forty years. Only the volume has. The volume went up by a factor of a billion, and the industry's answer to that was to hire more people to handle it.
The largest information industry in the world never got the information technology.
Why nobody has fixed this
Not for lack of trying, and not for lack of money. Broker management systems and policy admin systems have existed for decades. Brokers buy them, load them, and within a year the data inside is wrong.
Here's why, and it's the single most important thing to understand about this market: the hard part of a system of record for insurance is not the record. It's the ingestion. The truth doesn't live in the broker's database. It lives in the carrier's portal — eighty-plus of them per market, no APIs, credentials that rot, layouts that change without notice, documents that are images of tables.
Every company that came before us built a database and asked humans to fill it. They shipped a very expensive place to type things twice.
So the industry did the only rational thing available. It hired people. And it couldn't even outsource them properly — the global insurance BPO market is $7.7 billion, a rounding error against the problem, because this work can't be handed to a stranger. It requires knowing the account.
Fig. 1 — Outsourcing was never a match for the size of the problem: the BPO market is 0.5% of what the industry spends running itself.
The work stayed in-house, invisible, and enormous. Until now the only technology capable of reading an unstructured document, understanding what it meant for a specific client, and acting on it was a human being.
That is no longer true. It stopped being true about three years ago. Most of the industry responded by putting a chatbot on their website.
Records to Agents
Our strategy is one line: you cannot automate what you cannot see. Everything we build follows from that, in four stages, in this order:
- 01
One. The system of record.
Not a database, a pipe. Handle connects into every carrier portal in a market and pulls policies, receipts, endorsements and commissions continuously, then reconciles them against what the broker believes is true. The first time we ran full reconciliation on a large book we found more than 40,000 records the broker's own system had never contained. Not stale. Absent. Policies they were servicing that their database did not know existed, and commission gaps in the tens of millions of pesos. Their record had been lying to them for years and there was no way they could have known, because checking would have required a human opening every portal, every day, forever.
- 02
Two. Copilots.
Once the record is true, you let people talk to it. Ask your book a question in plain language and get an answer that's actually correct. This stage looks modest and it is the most important one strategically, because it is where trust gets built. A broker will not let software touch their client relationships until the software has demonstrated, boringly, over months, that it knows their business better than they do.
- 03
Three. Agents.
Then you put labor on top.
- a
- Quoting: the agent takes a risk, works every carrier in parallel, comes back with the comparison — the thing that used to take a person a day and eight browser tabs.
- b
- Collections: the agent chases money the broker has already earned, over WhatsApp and email, supervised, every action traceable.
- c
- Claims: the hardest one, the one that matters most to the end customer, gated until the data says we can do it right.
- d
- And many more to come. Every agent is only as good as the record underneath it. That's why the order is the order.
- 04
Four. Autonomy.
The back office runs itself. Humans handle exceptions and clients. Nothing else.
And then, on the other side of all four
Nobody in the history of this industry has ever held a clean, continuous, cross-carrier dataset of what policies exist, what they cost, who pays and who doesn't. Not the brokers. Not the carriers. Not the regulators. We will. And once you know how insurance actually behaves rather than how it was priced, you are no longer just automating the back office. You are in position to reprice risk itself.
What we're actually going for
Mexico is where we prove it — a market with over a trillion pesos of premium, ninety thousand agents, and three percent insurance penetration, which is to say a market where the cost of operating is precisely what keeps coverage out of reach of the people who need it most. Every point of expense we take out is a point of coverage that becomes affordable. That is not a marketing line, it's arithmetic.
Then Brazil, R$760 billion and the same portal problem in Portuguese. Then the rest of Latin America. Then the United States, where the same fragmented middle exists in English and pays better. Then everywhere, because insurance is sold everywhere and it is broken in exactly the same way everywhere.
| Stage | Market | Premium | Why |
|---|---|---|---|
| 01 | Mexico | >$1T MXN | Where we prove it. 90,000 agents, 3% penetration. |
| 02 | Brazil | R$760B | The same portal problem in Portuguese. |
| 03 | Rest of Latin America | — | Same fragmentation, same playbook. |
| 04 | United States | — | The same fragmented middle, in English, paying better. |
| 05 | Everywhere | $7T | Insurance is sold everywhere and broken the same way everywhere. |
The goal is not a good software company for brokers. There are several of those and they are all small.
The goal is that within a decade, the phrase "insurance back office" describes a piece of infrastructure rather than a floor full of people. That a brokerage runs five times the book with the same team. That a claim gets paid the same day because no one had to find the file. That Handle sits underneath a meaningful fraction of the world's seven trillion dollars of premium as the layer where the work actually happens.
The final boss
Nothing about this is going to be given to us.
The enemy is entropy — an industry whose data decays faster than anyone can maintain it, where every portal change breaks something and every carrier has its own definition of a policy number. The enemy is three hundred and forty years of accumulated process. The enemy is regulation written for paper, systems that were never built to be read by anything but a person, and a thousand small manual habits that each individually make sense and collectively cost the world a trillion and a half dollars a year.
Getting through that requires a specific set of things, and I want to be explicit about them:
- Agency
- Speed
- Taste
- Obsession with the boring
- Systems thinking
- A tolerance for ugly work
- Domain humility
- Stubbornness
The ability to sit inside someone else's operation until you understand it better than they do, and then build, and then go sit there again.
We were early to this. We intend to be the ones who finish it.
What are you still doing by hand that a machine should have taken from you already?