Essay · Markets & enterprise
The conflicted intermediary
Every layer between a client's money and where it needs to go is a place a conflict of interest can hide — and a place that boxes in both the client's outcome and the advisor's own judgment. Those layers were once the only way to serve people at scale. Agentic AI is what finally makes removing them possible, not just imaginable.
The tax nobody names
The complexity tax shows up everywhere institutions touch ordinary people's lives — but nowhere more quietly than in the layers standing between a person's money and the investment it's actually going into. Every fund wrapped inside another fund, every share class priced differently for the same underlying holding, every referral agreement that routes a client to whichever provider pays the best override, is complexity doing what it always does: moving value from the person who can't see the mechanism to the party who built it.
This is not a story about bad actors. It's a story about a structure built on dated technology and outdated regulation — one that rewarded adding a layer over removing one, because for most of the industry's history, adding a layer was the only way to serve more people with fewer people, and the advisor working inside that structure had no tool that let them serve people any other way. The constraint is gone. Most of the rules are not.
Same fund, different loyalty
Consider two share classes of the same mutual fund. The underlying holdings are identical. The only difference is what each pays the advisor who recommends it — and in several share classes, that difference comes out of the client's return in the form of a 12b-1 fee or a revenue-sharing arrangement between the fund company and the firm placing the trade. Regulators have scrutinized exactly this mechanism in recent years, bringing enforcement actions against advisers who placed clients into fee-bearing share classes when identical, cheaper share classes of the same fund were available to them. The product didn't change. The loyalty did.
Multiply that single mechanism across the products that make up a typical portfolio — proprietary funds, commission-bearing annuities, referral arrangements between firms — and the pattern repeats: a layer exists whose economics point toward the firm collecting it, not necessarily toward the client holding it. The advisor recommending it rarely built that structure either — they inherited a compensation architecture and did their best inside it.
The layers were once the only way to scale
None of this happened because anyone set out to build a system that worked against clients. Active management, fund packaging, and distribution networks were genuinely expensive to build one at a time — analysts, trading desks, compliance infrastructure only paid for themselves if they served thousands of accounts through a shared product. Wrapping a strategy in a fund and selling shares of it was, for decades, the only way to deliver professional money management to someone without millions of dollars to invest directly. The layer was the price of admission, not a scheme.
What kept the layer in place well past its usefulness wasn't the intent behind it. It was the software problem underneath it: nobody could deliver the individualized version — each client holding their own securities directly, tax-managed to their own situation, with no packaged product standing between them and the market — at a cost that made sense for anyone but the wealthiest.
What removes the layer now
Personalized indexing changes that math. An agent-run engine can hold hundreds of individual securities per household directly, harvest tax losses at the position level, and rebalance around each client's specific circumstances — the exact customization a packaged fund could never offer — without the packaging cost that justified the fund in the first place. What used to require a fund company's full operations, sold to thousands of anonymous shareholders, now runs at the individual-household level.
The gap right now is adoption, not capability. By Cerulli Associates' count, only 18% of financial advisors currently use direct or personalized indexing — up from 16% two years earlier — while 26% have it available to them and still choose not to, and another 12% don't know what it is. That gap is not a demand problem. It's the same lag every new infrastructure runs through before the people who'd benefit from switching learn that switching is possible — and the advisor stands to gain from closing it as much as the client does: the tool finally lets them build the practice many wanted to build all along, no longer boxed in by what a packaged product could offer.
Not every layer is a conflict
This case has a real limit, worth stating plainly. No advisory model is entirely free of incentive — the fee-only advisor charging a percentage of assets under management has a version of this problem too: a structure that rewards gathering more assets isn't automatically aligned with a client who'd be better served paying down a mortgage or starting a business. Removing a layer doesn't remove incentive. It just changes whose incentive is in the room.
Packaging itself isn't inherently the enemy, either — a fund can be the right vehicle for a strategy too small or too illiquid to hold directly, and plenty of advisors on commission-based or hybrid models serve their clients well despite a narrower standard of care than a fiduciary owes. The test was never whether an intermediary exists. It's whether the intermediary's economics point the same direction as the client's.
Fewer, aligned layers
The reform isn't zero intermediation. It's exchange as close to the source as a client's situation allows, with as few layers between them and their money as the strategy actually requires — and each layer that remains built so its economics move with the client, not against them. That's a different industry structure than the one built around packaged products sold at scale. It isn't a smaller industry. It's a less taxed one — and the people who gain the most from that are the ones who were paying the tax without ever being told its name. None of this required a mandate or a subsidy. It required the tool getting good enough to do the work — and now it has.