Scaling commercial real estate teams: AI or a new hire?

Scaling your commercial real estate team with AI is mostly a sequencing problem. Automate the mechanical work first, then hire against what is left. Run it in the other order and you pay salary for tasks a model handles in ninety seconds, while the work that actually needed a person keeps slipping.

Using AI and scaling with it are different things

Adoption in commercial real estate is close to universal now. Capacity gains are much rarer. A 2026 CRE Analyst survey found that roughly three quarters of CRE professionals use AI as a basic assistant, while only about five percent have used it to automate a workflow end to end.

Those two numbers describe an operating problem rather than a technology one. Drafting an email faster does not change how many deals an acquisitions team can carry, or how many properties a manager can cover. Automating rent roll intake does.

The pressure behind that gap is not evenly distributed, but most teams feel a version of it. Loan maturities are heavy through 2026, transaction activity has been uneven rather than absent, and the operational side of the business has not gotten quieter. Whatever your volume looks like, it is being absorbed by roughly the headcount you already have.

What breaks first when volume grows

Ask a director of acquisitions or a regional property manager where the week goes and the answers cluster into four buckets.

• Document intake. Offering memoranda, T-12s, rent rolls, leases, estoppels, CAM reconciliations. CBRE research cited by Commercial Observer puts abstraction of a single complex lease at four to eight hours and $150 to $350. One acquisition can generate 800 to 1,200 pages before anyone opens a model.
• Coordination. Tour logistics, broker and vendor follow-up, work order triage, chasing the document that never came back.
• Reporting. Budget versus actuals, variance narratives, quarterly LP updates, owner reports. Recurring, formatted, and almost entirely mechanical.
• Follow-up. The prospect who went quiet, the lease expiring in twenty-six months, the tenant inquiry sitting in an inbox since Thursday.

Very little of that requires judgment, and all of it grows in step with volume. Every additional deal or door buys you more of it. That is the real constraint on a growing team, and it explains why a senior hire so often fails to relieve the pressure. You hired someone for their judgment and then handed them intake.

Automate first, hire second

The test is usable in a Monday meeting. Automate a task when it is repetitive, document-driven, and produces an output someone can verify against a source. Hire when the task carries judgment, a relationship, or accountability for a decision.

Automate first:

• Extraction and normalization from leases, rent rolls, and operating statements
• First-pass market, comp, and ownership research
• Recurring reports with a fixed format
• Any drafting that follows a template

Hire for:

• Setting the exit cap and defending the rent growth assumption
• Reviewing and signing off on AI output before it reaches a model or a committee
• Owning a relationship with a broker, lender, vendor, or tenant
• Everything that goes wrong in a way nobody wrote a procedure for

Sequencing this way changes the hire itself. The job description moves from producing the work to verifying it and owning the exceptions, which is a more capable role at similar cost. It also changes what you screen for. Someone who can catch a mis-mapped T-12 line is worth more to you than someone who can build the schedule from scratch but takes the output on faith. The roles we place have shifted in exactly this direction over the past year.

The headcount question, honestly

There is a real argument on the other side of this, and it deserves a straight answer. JLL has noted that AI agents are starting to break the link between company growth and employee headcount in knowledge work, which is part of why office demand forecasting has gotten harder. If that holds broadly, growing capacity without growing teams becomes the norm rather than the exception.

The current evidence in real estate points the other way, with a caveat worth naming. AppFolio's 2026 Property Management Benchmark Report, which surveys property management operators and skews residential and multifamily, found that firms with broad AI adoption expected average portfolio growth of 31 percent for the year against 12 percent for firms that had not implemented, and that 34 percent of adopters planned to increase headcount versus 25 percent of non-users. JLL's own 2026 Future of Work survey of more than 2,200 executives found around 60 percent of companies still planning to expand headcount over the next three to five years, with the difference showing up in composition rather than direction.

The reading that fits both: AI raises the ceiling on what a team can carry, and firms that raise the ceiling take on more. JLL's published lease abstraction work is the cleanest example anyone has put numbers to, reporting roughly 60 percent less manual review labor, three times the volume without added headcount, and over a million dollars in previously missed escalation clauses. A team running three times the volume does not get smaller. It takes on more, and it needs people who can hold that volume to a standard.

Two things that break AI in CRE before staffing ever matters

Both came up repeatedly at Realcomm 2026 and in First American's industry research this year, and both sit upstream of any hiring decision.

The first is data. Commercial real estate runs on fragmented ownership records, inconsistent property information, and systems that were never built to talk to each other. A model reading clean, consistent source documents performs very differently from one reading whatever is in the shared drive. If your rent rolls are formatted five ways across six properties, normalization is the project, and the AI is the second step.

The second is review. The model assembles inputs. A person owns the conclusion. An extraction error in a rent roll does not announce itself, it propagates into an underwriting model and comes out the other side as a number somebody defends in an investment committee. Regulators and capital partners have both been pushing toward more defensible documentation, and an output nobody traced back to source is hard to defend.

So make review a named job with a name attached rather than a step everyone assumes someone else took. Every automated workflow should ship with a defined output format, a source reference for every extracted figure, and one person accountable for the verification pass. Skip the third and you buy speed at the cost of trust.

How to run a capacity audit

Before buying a tool or opening a requisition, get the picture. Setup takes an hour or two. The logging runs two weeks.

Step 1. Log two weeks of hours

Have each person tag time against four buckets: document work, coordination, reporting, and judgment or relationship work. Rough estimates are fine. You want the proportions, not precision.

Step 2. Total the first three buckets

That is your automatable load. Our rule of thumb is that under roughly a fifth of team hours, AI is not your bottleneck yet and the honest answer is a hire or a process fix.

Step 3. Rank by hours times repeatability

A four-hour task you run weekly beats a twelve-hour task you run twice a year. Pick one to start. Not three.

Step 4. Baseline before you change anything

Record current time per unit and current error rate. Without a baseline you cannot tell real automation from the feeling of it, which is how these projects tend to stall.

Step 5. Write the reviewer into the workflow

Name the person, define what they check, decide what happens when the output is wrong. Run it thirty days and compare to the baseline.

A worked example of what step 1 tends to surface:

• Rent roll normalization. Document work, roughly six hours per deal. Automate, with review.
• Broker and vendor follow-up. Coordination, roughly four hours weekly. Support hire.
• Owner and LP reporting. Reporting, roughly nine hours per cycle. Automate the assembly, keep the narrative and the sign-off with a person.
• Exit cap assumption. Judgment, a few hours per deal. Keep it senior.

Most teams that run this find two things at once: the automatable load is bigger than they assumed, and the coordination work still needs a person. That combination is what a scaled team looks like in practice. Models carry document and reporting volume, support professionals own coordination and verification, and senior people spend their hours on decisions.

Key takeaways

• Almost everyone in commercial real estate uses AI. Few have automated a workflow, and that is the only version that changes what a team can carry.
• Automate what is repetitive, document-driven, and verifiable. Hire for judgment, relationships, and accountability, and make review a named job.
• Data consistency comes before automation. Messy source documents produce confident, wrong output.

Start with the capacity audit. Once you know which hours are automatable and which need a person, the hire gets easier to define.

If you want support professionals who already work this way, read how our placement process works. Our guides go deeper on individual workflows, and the frequently asked questions cover how engagements are structured.