Chris MihosA poor strategy tends to lose money gradually, but a poor structure can lose all of it in a day.
Trading risk is the amount a strategy can lose through its positions, which shows up as drawdown, leverage, volatility and concentration. Structural risk is the amount you can lose because of the way the capital is held, and it depends on who has custody, who can move the assets, which counterparties are involved and whether the numbers you’re shown are real. Most people only measure trading risk, because that’s what appears on a factsheet, whereas family offices check the structure first because structural failures are usually total. Managing risk well means choosing your trading risk deliberately, on top of a structure that can’t let you down.
Structural risk is the one nobody puts on a factsheet
In November 2022, FTX went from being one of the largest crypto exchanges in the world to bankruptcy in roughly a week. Plenty of its customers had chosen their positions carefully and held assets they considered conservative, and they still found that billions of dollars of customer assets had gone missing. Their trading risk had been perfectly reasonable, but the structure they were relying on had failed underneath them.
That’s what structural risk looks like in practice. It never shows up as a drawdown, and you can’t diversify it away by spreading your capital across different strategies on the same platform. It really comes down to a handful of questions: where the assets physically sit and with whom, who is allowed to trade them and who is allowed to move them out (and whether that’s the same party), which brokers, venues and lenders sit in the chain, who calculated the value you’re looking at, and whether you could still prove what you own and get it back if the manager or the platform disappeared tomorrow.
The question of permissions is becoming more important every year. In May 2026 Robinhood opened live stock trading to AI agents such as Claude and ChatGPT, using a separate account so that the agent can only touch whatever has been placed in it. A few months earlier, an experimental AI agent with its own crypto wallet was asked for a small tip on X and sent away somewhere between $250,000 and $440,000 instead, depending on which report you read. Whether it’s an agent, a fund manager or a platform, the structural question is the same: can it move the money, or can it only trade it?
Institutional capital settled that question a long time ago. When a pension fund or family office allocates to an outside hedge fund, the manager is given the right to trade, while the assets stay with a custodian and can only be moved with the owner’s signature, often more than one. That separation is the reason the owner still knows where the assets are even if the manager turns out to be reckless or dishonest, and it’s the right model for AI agents as well.
Trading risk is more familiar, but it’s often measured too narrowly
Drawdown matters more than return, because losses compound against you. A 50% loss needs a 100% gain to get back to where you started, so the depth of a strategy’s worst historical drawdown and the time it took to recover usually tell you more than its best year does. Leverage matters too, because the same strategy run at three times leverage isn’t simply a bigger version of the original but a genuinely different risk.
Concentration is the risk that many newly wealthy people carry without ever having chosen it. SpaceX listed at $135 a share this year, climbed as high as $226 and was trading at around $127 within a month, so anyone holding most of their net worth in it watched more than 40% come off the peak in a matter of weeks.
The kind of trading risk that family offices worry about most, though, is hidden overlap. Three managers with three different strategy names can turn out to be running the same trade on the same venue, and a fund described as diversified credit can carry a heavy exposure to a single sector. By the end of 2025, around 19% of private credit’s direct lending, more than $500 billion, had gone to software companies, according to the Bank for International Settlements. When AI began to threaten the software industry, that exposure showed up for holders as redemption queues rather than as anything they could have read on a factsheet.
How a family office approaches it
What matters most about the family office approach is the order in which things are done. The structure comes first: before any capital moves, the family office confirms custody, separates the right to trade from the right to move assets, and arranges for an independent administrator to produce the valuation, and if the structure doesn’t pass, the strategy isn’t even considered.
The strategy comes next. The family office looks at the drawdown history, the leverage limits and how the strategy has behaved under stress, and then decides how much of that risk it actually wants, which is a choice the family makes rather than one the manager makes for them. After that it looks at the portfolio as a whole, rolling exposures up across every manager to find the overlaps that would never be visible one fund at a time.
Finally, it keeps watching. Risk isn’t something you check once at the point of allocation, so custody balances are reconciled against what managers report, drawdowns are tracked against the limits the family has set, and changes are flagged as they happen rather than at the next quarterly meeting.
What AI changes
Almost every part of that process involves reading, reconciling and watching, which is exactly the kind of work AI can now do reliably and at very low cost. An AI layer can reconcile custody balances against a manager’s reports every day instead of every quarter, read offering documents for leverage limits, side pockets and key-person clauses, track a strategy’s drawdown against a threshold you’ve chosen and let you know when it gets close, and roll positions up across managers to show you that two of them are effectively the same trade.
What it shouldn’t do is decide your risk for you. How much drawdown you can live with, how much concentration you’re willing to accept and which managers you trust are all decisions for you to make, and the job of the AI is to make sure you’re making them with the full picture in front of you.
Questions worth asking about any arrangement
On the structural side, it’s worth asking who holds the assets, whether the manager, the platform or an agent can move them or only trade them, how many signatures it takes to move them, who calculates the valuation, and whether you could verify all of this yourself without having to ask anyone. On the trading side, ask about the maximum historical drawdown and how long recovery took, what leverage the strategy is allowed to use, and what else in your portfolio is exposed to the same thing. If the answers to the structural questions are vague, there’s little point in moving on to the trading questions until they aren’t.
How MakeBanc handles both layers
On the structural side, your assets sit with an institutional custodian in a vault that you sign for, and the independent manager you choose can trade them but can never move them, while MakeBanc itself never holds the keys. You can require two or three named signers before anything changes, and you can verify custody on-chain without needing to ask anyone. The NAV is calculated by a third-party administrator and reconciled daily.
On the trading side, strategies are classified by type, whether market neutral, arbitrage, directional or multi-strategy, and historical performance and drawdown are shown as reported by the manager, labelled as self-reported, observed or verified, with the period always stated. The trading risk you take on is your choice, and MakeBanc neither assesses it for you nor suggests a strategy.
Frequently asked questions
What is the difference between structural risk and trading risk? Trading risk is the risk of losses from a strategy’s positions, such as drawdown, leverage and concentration. Structural risk comes from the way the capital is held, including custody, who can move assets, counterparty failure and unverified figures, and because it can lead to a total loss it’s usually checked first.
What does it mean that a manager can trade but not move assets? It means the manager has the authority to buy and sell within the account, but the assets stay with a custodian and can’t be transferred out without the owner’s signature. This is standard practice for institutional allocators and family offices working with external managers.
Why does drawdown matter more than returns? Because losses compound against you. A 50% loss requires a 100% gain to recover, so the depth and duration of past drawdowns say more about a strategy’s risk than its best returns do.
Sources
- Wikipedia, Bankruptcy of FTX
- Robinhood Newsroom, Robinhood is now open to agents, 27 May 2026
- The Crypto Times, AI agent accidentally sends $450K, February 2026
- WealthManagement.com, Wealth firms urge patience on Anthropic IPO (SpaceX share price), 30 October 2026
- BIS Quarterly Review, Private credit’s software lending meets AI disruption, March 2026
Past performance is not indicative of future results. Performance and drawdown data on MakeBanc are reported by the asset manager and labelled by source. MakeBanc is a technology platform that connects investors with independent asset managers and does not provide investment advice, custody services or banking services.