The book

Algorithmic Roulette

How Machines Decide Who Gets Funded

The investment committee is no longer where the decision begins. By the time a company reaches the room, machines have already helped decide what is visible, what counts as evidence and what looks fundable.

Trade nonfiction in development

The decision before the decision.

An investment committee believes it begins deciding when the papers arrive. By then, much of the judgment has already happened. Databases have determined which companies are visible. Categories have shaped how they are understood. Models have elevated some signals and discarded others. Language tools increasingly compress uncertainty into recommendations that look more conclusive than the evidence beneath them.

The final decision may remain human. The reality from which that decision is made increasingly is not.

Machines increasingly shape who is found, trusted and funded. But the odds are not set by code alone. They are shaped by the data, categories, incentives and institutional choices embedded around it — often long before a human decision-maker enters the room.

A precise answer to the wrong investment question is accelerated error, not insight.

The governing idea

The Investability Machine.

Before a company can be chosen, it must become visible, legible and investable. Algorithmic Roulette calls the system that performs this upstream work the Investability Machine: the interconnected data, classifications, models, incentives and institutional habits that shape the decision before the decision.

The book follows that machine from discovery to accountability.

Find

Who enters the field of vision — and who disappears before evaluation begins.

Frame

How categories and peer sets determine what evidence is allowed to count.

Score

How signals become judgment, and convenience can masquerade as prediction.

Fund

How ownership, time horizon and capital structure reshape the company being selected.

Govern

Who remains accountable when decision support begins to sound like the decision itself.

What is at stake.

This is not a book about whether AI will replace investors. It is about something more immediate: how machine-mediated selection changes what capital can see, what institutions learn to trust and which companies ever receive the chance to become consequential.

That makes capital allocation a question of institutional power. What becomes measurable is not always what matters. What looks familiar is not always what is good. And the company that fails to fit a category can disappear long before anyone believes a judgment has been made.

What capital can see shapes what the world gets to build.

Inside the book

Twelve chapters. Four movements. One argument.

Part I — The Engineered Wager
1
The Decision Before the Decision

Why the investment committee sees the end of a selection process, not the beginning.

2
Who Becomes Visible

How databases, sourcing systems and network effects determine which companies enter the field of vision.

3
The Familiarity Premium

Why legibility is repeatedly mistaken for quality and unfamiliar advantage is easy to classify as risk.

Part II — Inside the Investability Machine
4
The Category Is Already a Decision

How labels, peer sets and narratives determine what evidence can count.

5
What Counts as Evidence

Which variables genuinely predict outperformance, which are proxies and how false precision enters the process.

6
When Analysis Starts Sounding Like Judgment

How automated summaries compress uncertainty and alter the psychology of institutional challenge.

7
Capital Is Part of the Product

Why funding structure, ownership and time horizon can create or destroy investability.

Part III — Durable Advantage
8
Beyond the Model

What remains scarce when information, analysis and code become abundant.

9
When Regulation Becomes Advantage

When regulation is a cost, when it is a barrier and when it becomes a source of durable advantage.

10
The Wrong Capital Can Kill the Right Company

Why exceptional science can fail under an inappropriate capital architecture.

Part IV — Governing the Machine
11
The Test Before the Decision

A discipline for examining what a model can answer, what it cannot and what the institution must still own.

12
The Human Right of Refusal

Why a human in the loop is not governance — and why dissent, authority and responsibility still matter.

Why I am writing it.

I have spent more than twenty-five years across technology, capital and regulation — building companies, evaluating opportunities, allocating capital, serving in governance roles and working with institutions across markets.

Across those roles, one pattern has become increasingly difficult to ignore: the decision presented to a committee is rarely the beginning of the decision. It is the output of a system.

Algorithmic Roulette is an attempt to make that system visible: where it improves judgment, where it distorts it and where human beings must retain the knowledge, authority and courage to refuse its frame.

About Leesa Soulodre

Leesa Soulodre is Founder & Managing General Partner of R3i, an investor and board director working where frontier technology meets institutional capital and governance. Her career spans investment, company building, technology commercialisation, board and advisory work, regulated markets and more than a decade of executive education.

Her public record includes work with 400+ corporations across 19 industries and experience across Europe, Asia-Pacific and North America.

Follow the work.

Selected essays, research notes and publication updates related to Algorithmic Roulette.

Join the reader list