From Companies to Contracts

Securitising the real economy’s contracts, the AI way: HAT over MASS.

Every few decades, finance finds a new thing to securitise. Mortgages, then credit-card receivables, then music royalties, then the future gate receipts of football clubs. Each time, the underlying insight is the same: there is a cash flow investors would have wanted but had no way to reach — and once the standards, the verification and the market machinery exist to make it ownable, a return stream that sat in no one's portfolio before becomes investable.

I founded Micro Connect, which runs NUMA, a new market built on one such conviction: the next cash flow is the humblest one of all — the commercial contract. Not the company; the contract. The agreement under which a hotpot restaurant shares its daily takings, a computing centre bills for hosting, a concert tour splits its box office. Wall Street spent a century building the machinery that makes companies financeable: admission standards, measurement, distribution. This year we have published three white papers arguing that the same machinery can now be rebuilt, contract by contract.

One distinction matters, and I want to make it early. Nothing in this model is bundled or repackaged — not by us, not by anyone. Investors compose their own portfolios of contracts and enjoy the benefits directly. The obvious question is why none of this happened long ago. The answer is not that anyone doubted the value of these cash flows. It is arithmetic.

What we are building

If one sentence must survive this essay, let it be this: we are building the operating system for contract securitisation — the layer on which contracts become financeable the way companies became financeable a century ago. An operating system does not own the applications that run on it, and it does not play the users' games for them; it makes the running possible. So here: the standards by which a contract is admitted, the measurement by which it is read, the records by which ownership is kept — that is the layer we build. On top of it, investors compose their own portfolios and keep the returns; nothing is pooled or repackaged by us. AI is what makes this layer buildable at all — and we are, so far as we know, the first to build it.

Why now

Understanding a business costs roughly the same whether the business is worth a billion or a million — the analysis, the verification, the monitoring. Below a certain size, the cost of understanding exceeds the value of the thing being understood. That is why a pension fund can own a sliver of Apple but not a sliver of the noodle shop downstairs: not policy, not prejudice — the price of being known.

AI has moved the line. AI can now read, verify and score thousands of contracts a day at a cost approaching zero — work that once required rooms full of analysts, lawyers and monitors. If that holds, cash flows that were always real become legible to capital for the first time, and a market that never existed becomes buildable.

"The question was never whether these cash flows deserved capital. It was whether anyone could afford to find out."

Three white papers, published on the third of August, September and October, lay the foundation of this market together. The first asks which contracts can really work — what a contract must be, and show, before it may enter. The second asks how a contract, once inside, can be read — how its cash flows and risks are measured, day after day. The third asks who does all this reading and checking — and answers: an army of AI, under a discipline humans designed. The third white paper is the newest, and in many ways the heart of the matter; more on it below. The white papers are written for specialists. This essay is for everyone else.

Who this is actually for

Let me be precise about who benefits, because this is easily misread as philanthropy. It is not. The corner shop is one beneficiary — a meaningful one — but only a small part of the picture. Think instead of the large company that does not wish to dilute its equity, whose equity market is closed to it for now, or whose balance sheet cannot prudently carry more debt. For that company, financing against contracts is simply one more way of doing what Wall Street has always done: raising money in the most cost- effective manner available. The instrument is new; the purpose is the oldest in finance. And once the machinery exists, the large users carry the small ones along — scale is what makes the small economic.

For investors, the point is sharper still. A market of contract cash flows is a blue ocean of returns that do not move with the existing markets — the daily takings of a restaurant do not care what the Nasdaq did yesterday. Differentiated, quality return streams existed all along; what never existed was a way to own them. That is the opportunity these white papers describe: not a charitable programme for the little guy, but a new and largely uncorrelated territory for anyone with capital and judgment.

Is finance ready for AI?

Yes — absolutely, and emphatically. AI can do things in finance that were never before possible: reading a million small businesses the way a research department reads fifty large ones; watching cash flows daily instead of seasonally; making the invisible legible at a cost approaching zero. Done right, it will make finance broader, cheaper and fairer.

And no — not if it is done carelessly. AI makes things up. In a chatbot that is embarrassing; in market infrastructure it is catastrophic. A scoring engine that invents a cash flow is not a tool; it is a fraud machine. So the real question is not whether AI belongs in finance, but what it takes to make AI trustworthy enough to run something finance depends on.

Our answer — and the third white paper is essentially a long meditation on this — is not a smarter model. It is a stricter division of labour. Thousands of AI agents, each assigned a task so short and simple that getting it wrong is hard and checking it is easy; no single agent trusted with anything that matters. Above them sits a structure we call, with some fondness for the acronym, HAT over MASS: Human-commanded, AI-operated, Trust over Micro-Agents with Short and Simple tasks. Every agent has a named human accountable for it. Rules are written, changed and retired by people. AI operates; humans answer.

Whether this scales is an empirical question, and I will not pretend it is settled: the rulebooks now cover hundreds of industries, and validating them against real businesses is the work ahead. But I believe the shape of the answer is right: do not try to make AI trustworthy in the abstract. Make each agent too small to lie convincingly and put a named person behind every one.

Fitting into the regulatory landscape

A fair question for any new market is how it will fit into the regulatory landscape — and it deserves an honest answer rather than guesses about how those conversations will go. AI in finance is already everywhere; the genuine question for the industry, and for those who oversee it, is what level of trust a machine-run infrastructure must earn, and how. That conversation will take the time it takes, and it should.

Our own posture is simple. We design for accountability first — named humans, written rules, full records — and we publish our progress monthly, warts included. There is a cultural root to this that an English reader may miss. In Journey to the West, the novel every Chinese schoolchild knows, the monkey king Sun Wukong — capable of nearly anything — is allowed on the pilgrimage only because he wears a golden band the monk can tighten. The band is not the enemy of the magic; it is what makes the magic employable. That sentence is, more or less, our entire theory of AI governance.

Why you might care

If this works, the consequences reach well beyond small business. A market that can price a contract can, in principle, price any contract: the farmer's pre-sold harvest, the guesthouse's season, the tour's box office. Wall Street's century was about deciding which companies deserved capital. The next century will be about deciding which cash flows deserve it — a bigger, more granular, and frankly more interesting question.
 

About the destination, I have no doubt at all. AI will deliver this market; Wall Street will come to Main Street. The only honest uncertainty is about ourselves. We are the first to attempt this seriously, and we have spent five years bumping into walls — long enough to have a fairly good idea of how it should work, and we are working diligently to prove that we are on the right track. There is no guarantee itsucceeds in our hands. That it succeeds is, to my mind, not in question.

> Access to Industry Compass Deliverables <

To explore the Industry Compass frameworks covering 500 industries, please scan the QR code below and follow the instructions to provide the required information and complete verification. You will then be able to browse indicator and factor structures, available rule documentation, and version notes by industry.

Please refer to official disclosures for details of the materials currently available.

Note: Online viewing only; downloads are unavailable

Industry Compass website: https://microconnect.com/industry-compass

This essay tells the story around three white papers; the substance lives in the originals — White Paper One (August 3), White Paper Two (September 3), White Paper Three, HAT over MASS (October 3). The next progress report follows on November 3.