The Six Transformations

IThe Inversion Nobody Planned For

For a century, investment products moved mostly by push: manufactured by issuers, warehoused by banks, sold through channels whose economics rewarded placement. The emerging model is pull: an investor — or increasingly, an investor's AI — requests an exposure, and the exposure is assembled on demand, in fractional size, at any hour, from a universe that no longer requires an intermediary's permission to enter.

The industry's response has been to debate which products survive this inversion. Will ETFs be disintermediated by direct indexing? Will structured products be decomposed by machines that price their components transparently? Will funds still exist when anyone with a conversational interface can replicate an index from single stocks starting at one dollar?

These are the wrong questions, because they assume the product was ever the point. It wasn't. A fund is not a thing investors need; it is a bundle of transformations, some of which investors need and some of which existed only to be sold.

IIThe Replication Mirage

Start with the seemingly strongest case for total disintermediation: replicating a global equity index with fractional single stocks. The FTSE All-World holds roughly 4,300 constituents, with tail weights measured in ten-thousandths of a percent (FTSE Russell, 2026). At one-dollar minimum order sizes, holding the index at proportional weights requires a portfolio well into six figures before the smallest position clears the minimum. Below that threshold, the investor is not replicating; they are sampling, and absorbing tracking error that the fund's pooling would have eliminated.

Suppose the portfolio is large enough. The investor now owns 4,300 positions, each generating its own dividend events, corporate actions, withholding-tax claims, and proxy materials. An AI can administer this. But notice what has been built: a fund administrator, replicated per capita, without the pooling economics that made administration cheap. The treaty reclaim a UCITS fund executes once for a million investors now executes a million times. The securities-lending revenue a pooled vehicle harvests is forfeited entirely. And the index weights themselves are licensed intellectual property — self-indexing at retail scale is a legal confrontation with the benchmark administrators, not merely a technical exercise.

Self-replication is possible. It is also, for most investors, a machine for converting fee savings into administrative cost, tax leakage, and tracking error. Which raises the productive question: if assembly is not the need, what is?

IIIThe Six Transformations

Strip away everything that exists because it can be sold, and the residual list of what an investor actually needs is short. Six transformations, in descending order of how rarely they are discussed relative to their importance.

• Claim quality. Ownership that survives the insolvency of every intermediary between the holder and the asset, enforceable in a court that will actually hear the claim. This is Gate 1 of the Inheritance Test: does the holder have a direct proprietary claim, or a contractual claim on someone who has one? Most investors never examine this until the moment it is tested, at which point it is the only thing that matters. Every subsequent transformation is worthless if the claim underneath it is a claim on a claim.
• Exposure conversion. Savings must become claims on productive assets at the risk level the investor can actually carry — which is rarely the level they report on a questionnaire, and almost never the level implied by the products they are shown. The exposures themselves are commodities; beta is nearly free. Getting the level right, and keeping it right as circumstances change, is the need. Factor tilts, drawdown constraints, and exclusion policies belong here too — they are instruments of level-setting, not tailoring. The industry has historically sold exposure selection — which securities, which themes — precisely because selection is differentiable and level-setting is not.
• Time transformation. The portfolio's liquidity profile must match the investor's biography, not the market's calendar. Liquidity when life demands it; term when liabilities allow it. Most permanent investor damage is a maturity mismatch with one's own life — forced liquidation at the bottom because the portfolio's duration ignored the holder's. Funds performed this transformation crudely, through redemption terms and the sheer inconvenience of exit. The transformation remains essential even as the crude mechanism dissolves.
• Tax passage. The least frictional legal path between gross and net return. Over the compounding horizons that define investment outcomes, tax architecture is the largest controllable variable — larger than security selection, larger than fees in most regimes — and the least discussed, because nobody earns distribution revenue on it. The wrapper's tax chemistry is genuinely irreplicable by self-assembly: a US ETF's in-kind creation and redemption mechanism launders embedded capital gains (Poterba & Shoven, 2002); an Irish UCITS accesses treaty withholding rates an individual cannot; a self-built basket triggers a taxable event on every rebalance. Whether the fund or direct holding wins is regime-specific. That the answer is never neutral is universal.
• Cost floor. Every basis point of intermediation must purchase one of the transformations above. Any basis point that does not is extraction. This is the test most of the historical product landscape fails — not because intermediation is illegitimate, but because the fee was priced against the bundle while the value sat in two or three of its components. The pull inversion enforces this test mechanically: when the investor's AI can decompose the bundle, only the load-bearing components can sustain a price.
• Protection from oneself. The behavioural wrapper — whatever stands between the investor's impulses and their execution. Historically this was assembled from accidents: the fund's boring quarterly statement, the redemption notice period, the adviser who didn't answer the phone on the worst day of the crash. The persistent gap between fund returns and fund investor returns — documented for decades in studies of investor timing (Barber & Odean, 2000; Morningstar, 2025) — is the cost of this need going unmet. It is the only transformation on the list that friction provided and frictionlessness deletes.

Personalisation is load-bearing where it shapes the risk level — tax lots, currency of liabilities, time horizon, drawdown tolerance — and pure vanity at the level of security selection, where it adds tracking error with a personal signature.

Notice what is absent from the list: tailoring. The pull scenario assumes the investor needs a bespoke portfolio. This is mostly false. The industry sold customisation because customisation justifies fees. The AI era risks selling it because customisation justifies engagement. Neither motive maps onto a need.

IVThe Access Spectrum, and What Each Tier Silently Deletes

Set the six transformations against the three ways investors now reach markets, and a pattern appears: each generation of access solved the previous generation's extraction and deleted one of the needs in the process.

Advised access — the traditional intermediated model — delivered exposure conversion, time transformation, and behavioural protection through a human relationship, and overcharged for them by bundling in assembly fees and smuggling conflicts through the product shelf. The regulatory architecture of suitability under MiFID II assumes this world: an intermediary who proposes, a client who disposes, and a duty attached to the proposal. The model's defect was never its function. It was that the compensation flowed from manufacturers rather than from the investor, a structural asymmetry examined at length in The Hidden Price.

Non-advised self-execution — the neobroker generation — deleted the conflict and the cost, and deleted behavioural protection with them. Appropriateness testing replaced suitability; a warning replaced a counterparty with a duty. The empirical record of this trade is visible in the widening behaviour gap of the commission-free era: as the friction of execution fell towards zero, the cost of impulse fell with it, and the damage migrated from fee statements — where it was at least disclosed — to timing decisions, where it is invisible. The harm was not removed. It was transferred from extraction the investor could see to extraction the investor performs on themselves.

Direct access without intermediary — the DeFi endpoint — completes the progression. Here the deletion is total, and so is the clarity. On-chain bearer instruments solve the first transformation better than traditional finance ever has: settlement finality without a custody chain, ownership that survives every intermediary because there is no intermediary to survive. And the same architecture deletes tax passage entirely, renders time transformation an illusion — twenty-four-hour liquidity is twenty-four-hour impulse execution — and removes the last vestige of behavioural protection, because there is no counterparty left to refuse anything.

DeFi is the first financial system in which claim quality and behavioural protection are perfectly inversely provisioned. It maximises the transformation investors never think about and eliminates the one whose absence destroys them.

VFrictionlessness Is the Product Defect

The pattern across all three tiers admits a blunt summary. The historical fund's worst features — its boredom, its opacity of daily pricing, its slow redemption — were unpriced behavioural assets. Each access generation removed friction and called the removal progress. Sometimes it was. But the removal was also, each time, a harm transfer: from visible fee extraction to invisible behavioural extraction, from a conflicted intermediary who at least owed duties to no intermediary owing anything.

The terminal form of this progression is now assembling itself: AI-mediated distribution whose operator monetises engagement, attached to execution rails with one-dollar minimums and no closing bell. Every component is individually defensible. The composite is a machine for converting impulse into position at a latency no previous generation of investors has survived contact with.

—The Needs Survive the Products

The fund as distribution technology is ending. The fund as legal technology — tax, access, unitisation, pooled scale — survives, and arguably gets purer, because everything that existed only to be sold gets stripped away. The six transformations are indifferent to the wrappers that carry them.

Which is why the sixth need — the one every generation of access has treated as disposable — is about to invert from an unpriced by-product into the scarcest product in finance. What the investor needs, in a world where everything is available, is an entity that can refuse.