Running at 25%

There is a pattern I have observed consistently across CRE PE firms, and it is uncomfortable to name directly. The most capable people on your team — the analyst who can build a complex waterfall model from scratch, the asset manager who knows every tenant situation across the portfolio, the IR associate who can distill a 40-page LP report into the three sentences a sophisticated allocator actually cares about — are spending the majority of their time on work that has nothing to do with those capabilities.

Status updates. Reforecast coordination. Meeting prep. Email chains where five people are aligned on something two of them could have decided in ten minutes. The operational scaffolding that holds a firm together but produces no investment return on its own.

A reasonable estimate, consistent with research on knowledge worker time allocation, is that your best people are operating at roughly 25% of their actual capacity. The other 75% is coordination overhead — the tax the firm pays for being organized, not for being excellent.

AI is not primarily a content generation tool. It is a coordination overhead elimination machine. Understanding that distinction is what separates firms that are building durable operational advantage from firms that are running a more expensive version of the same process they had five years ago.

Where the 75% goes

In a CRE PE firm of ten to thirty people, the coordination overhead is not abstract. It has specific line items.

Deal screening and pipeline management. A typical mid-market firm running active deal sourcing will move 50 to 150 properties through some level of initial screening in a given month. The analyst time consumed by pulling broker packages, building preliminary models, formatting deal memos for IC review, and updating the pipeline tracker is substantial — and most of it is process execution, not judgment. The judgment moment is the go/no-go decision at the end of the screen. Everything before it is infrastructure.

Quarterly portfolio reporting. The IR associate who produces quarterly reports for LPs typically spends two to three weeks per cycle on data collection, formatting, variance explanation, and coordination with the asset management and finance teams. The work that actually matters to the LP — the narrative judgment about what is happening in the portfolio and why — might represent four hours of that cycle. The rest is production overhead.

Lender and counterparty management. An active financing pipeline requires tracking dozens of lender relationships simultaneously: term sheet status, extension requests, covenant compliance windows, rate lock timing, closing conditions. The asset manager or CIO who holds this manually is doing database work with their brain. That is a misallocation of the scarcest resource in the firm.

Internal coordination. Every meeting that could have been a structured written update. Every email chain that exists because there is no single source of truth for deal status, asset performance, or portfolio metrics. Every reforecast cycle that requires the same data to be pulled, reformatted, and re-entered by three different people in three different systems. This is the invisible overhead — no single item is large enough to flag, but the aggregate consumes weeks of senior time per quarter.

The firms operating at $1M revenue per head are not staffed differently than firms operating at $300K per head. They have stripped the coordination tax from the work, which means their people spend their time on the work that only they can do.

AI's first job at your firm

The instinct at most firms is to deploy AI where the outputs are most visible: IC memo drafting, market commentary, LP report language. Those are reasonable places to start. But they are not where the structural leverage lives.

The structural leverage is in the workflow layer below the visible output — the infrastructure that produces the inputs those documents require. Deal screening pipelines that run continuously against market data rather than requiring analyst time to initiate. Portfolio monitoring systems that flag variance to the business plan automatically rather than waiting for the quarterly cycle to surface it. Reporting templates that pull live data and generate first drafts on a schedule rather than requiring a two-week production process per cycle.

A Harvard Business School field study at Procter and Gamble found that individuals working with AI were three times more likely to produce ideas in the top ten percent of quality as judged by independent experts — not because the AI generated the ideas, but because removing the coordination overhead around ideation freed the person's capacity for the work that actually required their judgment. One person with AI matched the output of a two-person team without it. The coordination cost that required two people to synthesize perspectives was absorbed by the tool instead of the organization.

The same dynamic applies to a CRE PE investment process. The analyst's judgment about whether a deal meets the investment criteria is valuable. The two hours it takes to pull the broker package, format the preliminary model, and update the pipeline tracker before that judgment can be rendered is coordination overhead. Strip the overhead. Preserve the judgment. That is AI's first job at your firm.

The mistake hiding inside the efficiency narrative

Here is where most AI deployment conversations in CRE go wrong. The framing is almost always about doing more with less: reducing headcount, handling higher deal volume with the same staff, replacing junior roles that have been automated. That framing produces bad decisions.

The research is clear on what happens when firms deploy AI primarily as a headcount reduction strategy. Gartner's analysis in early 2026 projected that by 2027, half the companies that cut staff for AI would rehire workers performing similar functions. Forrester's data showed 55% of employers already expressing regret over AI-driven layoffs. The pattern is consistent: firms that automate task execution and reduce the people who performed those tasks discover, too late, that the people were doing something more than the tasks.

A study of 62 million American workers across 285,000 firms between 2015 and 2025 found that companies adopting generative AI saw junior employment drop 8% relative to non-adopters within 18 months. Senior employment kept rising. The surface reading is that AI replaces junior workers. The accurate reading is more specific: AI replaces task execution. What it cannot replace is the accumulated institutional knowledge that senior people carry — the mental model of how the firm's investments work, what the load-bearing assumptions are, and what decisions made three years ago constrain the options available today.

For a CRE PE firm, this distinction matters more than it might appear. The junior analyst is not primarily valuable because they can build a preliminary underwriting model. They are valuable because, over time, they are building the pattern recognition that eventually becomes the judgment that makes your firm's investment decisions better than the market's. If you automate the task and eliminate the role, you also eliminate the learning path. The judgment that would have existed five years from now never develops.

The right frame is not "what can AI replace?" It is "what overhead can AI eliminate so your people can focus entirely on the work that builds the judgment your firm needs?"

What the senior people on your team are actually worth

There is a second way firms get this wrong, and it tends to affect senior roles more than junior ones.

The asset manager who has managed 18 transactions through a full cycle carries something that is not on any org chart. They know which assumptions in the current business plans are optimistic and why. They know which GP relationships will hold under stress and which ones require active management. They know the decisions made at acquisition that constrain what is possible today, because they were there when those decisions were made. They know, in a way that cannot be fully documented, what this portfolio needs and what it can tolerate.

When AI-assisted analysis enters an asset management process, the temptation is to measure it against the outputs the asset manager has historically produced: reforecasts, variance reports, lender updates, board presentations. AI handles the production of those outputs reasonably well. The conclusion drawn too quickly is that the role has been substantially automated.

What has actually been automated is the production overhead. The work the asset manager was doing between the judgment calls — the data collection, the formatting, the coordination — that can be automated. The judgment calls themselves — when to push back on a GP's optimistic leasing timeline, when to have a difficult conversation with a lender, when a business plan deviation requires an LP update versus a quiet course correction — those cannot be automated. They require someone who has been wrong before and learned something from it.

This is what I call contextual stewardship: maintaining the mental model of how the firm's investments actually work, representing that model in ways that help AI systems perform better, and exercising judgment about when technically correct analysis is organizationally wrong. It is not a technical skill. It is not about learning a specific AI tool. It is about becoming the person in the firm who holds the context that keeps the machines from going sideways.

The firms that build durable operational advantage in the next several years will be the ones that are clear about this distinction. They will use AI to eliminate coordination overhead aggressively. They will not use it as a rationale to eliminate the people who carry institutional context. Those are not the same decision, but they are easy to conflate when the efficiency narrative dominates the conversation.

The path to $1M per head

The north star metric I use is $1M in revenue per employee. For a mid-market CRE PE firm managing $1B in assets, that implies a team of eight to twelve people operating at a level of productivity that would have required thirty to forty people a decade ago. That is not a headcount reduction story. It is an overhead elimination story.

The firms getting there are doing two things simultaneously. First, they are automating the coordination layer: deal screening pipelines, portfolio monitoring systems, reporting infrastructure, lender tracking, LP communication workflows. The people who were spending 75% of their time on this work are now spending that time on judgment, relationships, and the hard problems that require someone who actually understands the assets.

Second, and less visibly, they are encoding what their senior people know into the infrastructure that supports the firm. Judgment Filters that encode IC experience into systematic checks. Deal criteria that encode underwriting philosophy into screening logic. Asset management protocols that encode business plan management experience into early warning systems. When the institutional knowledge lives only in people's heads, it is fragile. When it is encoded into infrastructure, it compounds.

Your best people are not running at 25% because they lack ability. They are running at 25% because the firm has not yet built the infrastructure that strips the coordination overhead away and leaves them with the work they are actually here to do.

That infrastructure is buildable. The question is whether you build it deliberately or wait until the firms that already have it make the gap visible.

Back to all insights