The operating-leverage case for AI when fees compress

There is a quiet arithmetic problem running underneath the institutional real estate management business, and it has been getting worse for years. On one side, management fees are under steady downward pressure, from larger and more sophisticated LPs, from competition among managers, and from an institutionalization of the asset class that has made fee levels a negotiated, scrutinized number rather than a given. On the other side, the books those fees support keep getting bigger, as capital concentrates with larger managers and each is asked to do more per dollar. Lower fees on bigger books is not a temporary squeeze. It is the structure of the business now, and it changes which strategies actually work.

The instinct, when a book grows, is to staff up to it: hire analysts, hire asset managers, hire on the finance and IR side to keep pace with the reporting load. That was fine when fees were generous enough to carry the incremental headcount. It is dangerous now, because the fee per asset has fallen and the cost per hire has not, so the manager that hires in proportion to the book quietly converts a growing top line into a shrinking margin. The way out is not to stop growing. It is to break the link between the size of the book and the size of the team.

The vise: fees down, books up

Start with the direction of fees. Across institutional real estate, the long-run trend is compression: base management fees that once sat comfortably are negotiated down, fee breaks at scale are expected, and LPs push for terms that align cost with performance. It is the same fee pressure that has reshaped every other corner of asset management, arriving in real estate with a lag. The number a manager earns per dollar of AUM is lower than it was, and the path of least resistance points further down.

Now layer in scale. Industry productivity benchmarks, the kind compiled in the long-running surveys of real estate investment managers, show the per-person load rising over time: asset managers carrying meaningfully more properties each than a decade ago, on the order of sixteen versus thirteen, with net operating income overseen per asset manager climbing alongside. That sounds like good news, and at the margin it is, but it is also the symptom. The load per person is rising because it has to: the book grew, the fee per asset shrank, and so far the amount of work each person absorbs is what has given.

Put the two together and you have the vise. Revenue per asset is falling while assets per manager are rising, so the firm is asked to administer more for less, indefinitely. That cannot be relieved by working the existing people harder forever: there is a ceiling on hours and a floor on quality below which an institutional manager cannot go without losing the LP. The pressure has to be released somewhere other than the people.

Why you cannot scale headcount in lock-step

The reason hiring cannot be the answer is visible in the cost structure of a manager. At the median firm, compensation runs at roughly 42% of revenue. That single figure tells the whole story: people are not a line item, they are the line item, the largest controllable cost in the business by a wide margin. So any strategy that scales labor in proportion to the book scales the biggest cost in proportion to the book, while the fee per unit is falling. The margin then compresses from both ends, and no business holds its margin while its largest cost grows at least as fast as revenue and its revenue per unit shrinks. The lock-step hire feels safe because each one is justified by real work. But the aggregate of justified hires, against a compressing fee, is a margin walking quietly toward zero.

There is also a quality cost the headcount number hides. More people doing the same manual work means more handoffs, more versions of the model, and more places for a number to diverge, so past a point the marginal analyst returns less than the one before. The firm pays more to produce work that is harder to keep consistent, which is the worst of both directions.

The only lever left

If fees are set by the market and the book is something the firm wants to grow, then neither the top line per asset nor the size of the book is the lever. What remains is the cost of producing the work. Not the cost of judgment, the underwriting call, the hold-sell decision, the negotiation, which is what the firm is paid for, but the cost of the production around it: assembling the data, building and updating the models, reconciling the systems, and generating the reports. That layer is where the hours go, and it is the one input the firm fully controls.

This reframes the AI question entirely. The point of AI here is not novelty and not a chatbot. It is that the production layer, the part of the work that is rote, repeatable, and rule-governed, is exactly the part an AI-native operating system can take over, leaving the judgment with the people. Lower the cost of production and you relieve the vise without touching fees, without degrading service, and without hiring in lock-step. You change the slope of the cost line relative to the book, the only structural fix available. Industry research backs the size of the prize: McKinsey, examining where AI creates value in real estate, points to large potential gains concentrated in exactly these production-heavy functions.

The manual production layer

It helps to be concrete about what the production layer is, because that is what gets automated and where the leverage comes from.

Data assembly

Before anyone analyzes anything, someone has to gather it: pull the rent rolls, reconcile them against the leases, line up the ledgers, chase the operating statements, and stitch the pieces together. This is a large fraction of analyst time and it produces no judgment, only inputs. An AI-native system that reads the systems of record and reconciles them continuously removes most of this work, for every asset at once rather than one spreadsheet at a time.

Modeling and updating

Models do not build and maintain themselves. The cash-flow model has to be constructed, refreshed each period, re-tied when a lease changes, and re-run when an assumption moves. Much of that is mechanical, applying known rules to updated inputs, and it scales badly because each asset carries its own model to keep current. Automating the rote maintenance, on a deterministic engine so the output is reproducible, frees the analyst to spend time on the assumption rather than the keystrokes.

Reporting

Then the work has to be packaged: the IC memo, the asset review, the lender update, the quarterly investor report. Assembling these by hand, pulling figures, formatting, checking ties, is a recurring tax that grows directly with the number of assets and investors. It is also the most templatable work in the firm, and therefore the most automatable. Generating the reporting from the same reconciled, cited data, for a human to review and approve, collapses the recurring cost.

None of this production is the thing the firm is paid for. LPs do not pay a manager to assemble data, refresh a model, or format a report; they pay for the judgment those activities support.

Operating leverage, made real

Frame all of this in the language the CFO uses and it becomes operating leverage: revenue grows faster than the cost base, so margin expands as the firm scales. In a labor-bound services business that is structurally hard to get, because the cost base is people and people scale with the work. Automating the production layer is what creates the leverage a manual firm cannot have: the book grows while the cost base grows more slowly, because new assets no longer require a proportional unit of labor to administer.

• Grow the book without growing the team in step. Adding assets adds revenue without a proportional headcount, so the per-asset cost of administration falls as the book grows.
• Hold the margin while fees compress. With compensation decoupled from the asset count, the firm absorbs lower fees per asset without watching the margin walk to zero.
• Redeploy people onto judgment. Analysts freed from data assembly and report formatting move to the work that earns the fee: underwriting, asset strategy, and investor relationships.
• Turn scale into an advantage instead of a tax. A manual firm pays a coordination penalty as it grows; an AI-native firm gets cheaper per asset, the only version of scale that compounds.

This is why the AI case under fee compression is an operating case, not a technology case. The technology is the means; the end is a cost structure that can survive lower fees on larger books. The firms that come through compression in good shape will be the ones that decoupled the book from the team, automated the production layer, and kept their best people pointed at the judgment LPs actually pay for.

Fee compression is not a problem you out-hire and not one you out-work; it is a problem you out-engineer, by lowering the cost of producing the work so the book can grow faster than the cost base. Built AI is the AI-native operating system that automates the production layer, the data assembly, the modeling, and the reporting, on a deterministic engine with every figure cited, so the firm gains real operating leverage instead of converting a growing book into a shrinking margin. To see what that does to your own cost per asset, explore how the platform works or book a walkthrough.