The real cost of screening deals by hand is not the hours your team spends reading offering memorandums. It is the deals they never got to. At a 4-to-20-person firm, screening capacity is a fixed pipe, and every deal that arrives above the line your team can process gets skimmed or ignored — including, some weeks, the best deal of the quarter. The visible cost is a number on a timesheet. The real cost is invisible, larger, and shows up as deal flow that quietly never becomes deal volume. This piece prices manual screening across five cost lines a principal can estimate for their own firm, then draws the exact line where AI belongs in screening and where it must not go.
Why the hours are the smallest part of the cost
Ask a principal what manual deal screening costs and you get an hours answer: an analyst spends most of a day a week opening deals, reading the OM, pulling the T-12 and rent roll, running a back-of-envelope return, and deciding what is worth a full underwrite. Multiply the hours by a loaded rate and you have a number. That number is real, and it is the least important part of the cost.
The reason is that screening is not a task you complete; it is a throughput constraint you run into. A lean team can screen a fixed amount per week by hand. Below that line, everything gets looked at. Above it, deals get triaged on the subject line, the broker’s reputation, or whichever email happened to be on top. The cost of manual screening is dominated by what happens above that line — the deals that never got a fair read — and none of it appears on a timesheet.
That reframing matters because it changes what you are actually buying when you make screening faster. You are not buying back a day of analyst time. You are raising the throughput ceiling, so more of your real deal flow converts into underwritten deals, and fewer good deals die in the unopened pile. Price the status quo honestly and the case for changing it looks different than the hours suggest.
The five cost lines of manual screening
Manual screening costs a lean firm on five lines, only one of which is visible:
- Direct labor — the hours to open, read, and sanity-check each deal.
- Throughput ceiling — the deals that arrive above your weekly capacity and never get opened.
- Speed-to-first-look — the access you lose when a faster firm engages the broker before you do.
- Decision quality — the buy box applied inconsistently by a tired human, killing good deals and advancing weak ones.
- Senior-time misallocation — principals and acquisitions leads doing extraction instead of the judgment and relationship work only they can do.
Lines 2 through 5 are where the money is. Work through each one for your own firm and the total is almost always a multiple of the labor line you started with.
Line 1: Direct labor — the visible number
Start with the number everyone already counts. For each deal that gets a genuine screen, someone opens the package, reads the OM for the story, pulls the operating statement and rent roll to check whether the story is true, runs a quick return against the buy box, and writes a one-line verdict. On a clean, well-formatted deal that is a modest block of time. On the scanned, inconsistent, typo-ridden files that fill a real inbox, it is more.
Multiply the per-deal time by the deals you actually screen in a month and by a loaded hourly cost, and you have the visible line. For most lean firms it is meaningful but not alarming — which is exactly why it understates the problem. The labor line only counts the deals you chose to open. It is blind to the ones you did not.
Hold onto the per-deal figure, though. It is the input that makes the next line concrete.
Line 2: The throughput ceiling — deals you never open
This is the largest and least-measured cost. Count the deals that hit your inbox in a month, then count the ones that got a real screen rather than a two-second skim. The gap is your throughput ceiling, and it is not a rounding error at a lean firm — brokers blast, and a small team’s weekly screening capacity fills fast.
The cost is not the un-opened deals as a group; most of them deserved to die. The cost is that your kill rate above the line is random. A deal that fits your buy box exactly can land in a crowded week and get skimmed off on the subject line, while a worse deal in a quiet week gets a full read. Over a year, the expected value of the deals lost to a random ceiling is large, because a single good acquisition your firm never underwrote dwarfs a year of analyst hours.
You cannot recover a deal you never opened, and you will never know which one it was. That is what makes this line easy to ignore and expensive to keep ignoring. The whole point of raising the ceiling is to make the kill decision deliberate again — every deal screened against the same standard, so the ones that die, die for a reason.
Line 3: Speed-to-first-look — access you lose to faster firms
In a competitive process, the firm that engages first often gets the look. A broker with a hot deal and ten calls to make remembers the principal who came back within the hour with three sharp questions, not the one who surfaced four days later after the deal was under contract. Slow screening is not just a throughput problem; it is an access problem.
Manual screening is structurally slow at the moment speed matters most. When a good deal lands during a busy week, it waits in the queue behind whatever arrived first, and the delay is invisible until the broker has already moved on. The cost is the deals you would have won access to if you had come back first — relationships and looks that compound, because brokers route their next deal to the firm that treated the last one seriously and fast.
This line is hard to quantify and real anyway. Ask yourself how many times in the last year a broker’s “you’re a little late on this one” was a deal you would have chased if you had seen it Monday instead of Thursday.
Line 4: Decision quality — the buy box you apply unevenly
A buy box is only as good as its consistent application, and a human screening deals under fatigue applies it unevenly. The tenth deal of the day gets a different standard than the first. A number in the wrong column of a messy T-12 gets missed. A deal advances on the strength of a compelling narrative that the operating statement would have contradicted if anyone had checked line by line.
The cost has two directions. False negatives — good deals killed for the wrong reason — feed back into the throughput problem. False positives are worse per event: a deal that should have died in screening instead consumes a full underwrite, senior review, maybe a site visit, before it dies later at greater cost. Inconsistent screening pushes cost downstream, where it is highest.
The fix is not heroics; it is a standard applied the same way to every deal regardless of how tired the screener is or how good the story sounds. That kind of consistency is exactly what a machine is good at and a fatigued human is not — provided the machine is doing the reading and ranking, not the deciding. The discipline that keeps that line clean is the same one covered in the comp-selection framework for an AI-assisted world: let the tool assemble and rank, keep the judgment human.
Line 5: Senior-time misallocation — judgment spent on data entry
At a lean firm, the person screening deals is often the person who should be negotiating them. When a principal or acquisitions lead spends hours a week extracting numbers from PDFs, the firm is paying its most expensive judgment to do its least judgment-heavy work. The extraction gets done; the calls that build the pipeline do not.
This is an opportunity cost, and it is steep because senior CRE time has a much higher and-best use than data entry. Every hour a principal spends transcribing a rent roll is an hour not spent on a broker relationship, a negotiation, or the two live deals that actually deserve their attention. The firm feels productive — work is happening — while its scarcest resource is spent on the most delegable task in the building.
The tell is simple: if your best deal-maker could not name three broker relationships they let cool last quarter because they were buried in packages, they probably could not, because they were. That is the cost, and it does not appear anywhere in the books.
Pricing your own screening cost
You can put a rough number on all five lines in an afternoon. Take your visible labor line as the anchor, then estimate the rest against it:
| Cost line | How to estimate it | Typically |
|---|---|---|
| Direct labor | Per-deal screen time × deals screened/month × loaded rate | Visible, moderate |
| Throughput ceiling | Deals received − deals genuinely screened, valued at the expected value of one missed acquisition/year | Largest, hidden |
| Speed-to-first-look | Deals/year lost to a competitor who engaged first | Real, hard to measure |
| Decision quality | Full underwrites spent on deals that should have died in screening | Moderate, downstream |
| Senior-time misallocation | Principal/lead extraction hours × the value of their best alternative use | High, invisible |
The exercise is not about precision to the dollar. It is about seeing that the labor line you started with is the small one, and that the throughput ceiling and senior-time lines — the invisible ones — carry most of the cost. A firm that prices only the hours is deciding whether to change based on the least important number it has. For where screening sits inside the full analysis workflow, the deal-analysis playbook for lean teams puts the stages in sequence.
Where AI removes cost — and where it must not
AI attacks four of the five lines directly, because the expensive lines are all downstream of one thing: how much your team can read and rank in a week. A general assistant like ChatGPT, Claude, or Gemini, pointed at a deal package, will pull the key terms into structured fields, check them against your buy box, and draft a one-line verdict in a fraction of the manual time. Raise that per-deal speed and the throughput ceiling rises with it, the queue that slows your first look shrinks, and the principal gets out of the extraction seat.
The consistency line improves too, but only in the right configuration. A model applies the same buy box to the tenth deal as the first, which is precisely the fatigue problem it solves — as long as it is reading and ranking, never deciding. The safe division of labor is narrow and firm:
| Screening task | AI does this well | A human must own this |
|---|---|---|
| Read the package | Extract terms from OM, T-12, rent roll into fields | Verify any number that will price an offer |
| Rank against the buy box | Score and sort deals, explain each ranking | Make the go/no-go call on the short list |
| Summarize | Draft the first-pass deal one-liner | Sign off before it drives a decision |
| Flag gaps | Surface missing or inconsistent figures | Judge whether the gap kills the deal |
The line that keeps this safe is the same one vendor pitches blur: screening is triage, not underwriting. Screening tells you what to open; underwriting tells you what it is worth. A model can rank your inbox all day, but it has no proprietary transaction data and will state a wrong number with the same fluent confidence as a right one. Every figure that prices an offer gets verified against its source, and the go/no-go stays with a person. Cross that line and you have not removed cost — you have moved it downstream to where a fabricated number does the most damage.
There is a low-cost way to test the boundary before buying anything. Getting the team fluent enough to run their own deal packages through a general assistant — prompting it well, verifying its output, knowing what never to trust it with — is a workshop-scale investment, market rates for that kind of AI fluency training run roughly $2–15K, and it often proves whether a dedicated platform is even needed. A custom-built screening workflow is a larger project, typically in the tens of thousands to low six figures depending on scope. Both decisions get easier once you have priced the status quo honestly.
What changes when screening stops being the bottleneck
When screening is no longer the constraint, the firm’s whole deal posture changes. More of your real deal flow converts to underwritten deals, so the funnel widens at the top without adding headcount. You come back to brokers first more often, so you win more looks. The buy box gets applied the same way every time, so the deals that die, die for a reason, and the ones that advance earned it. And your best deal-maker is back on the phone instead of in a PDF.
That is the actual return on removing manual screening cost — not a day of analyst time recovered, but a firm that punches above its headcount because its scarce judgment is pointed at judgment work. It is the same pattern that lets a lean shop out-operate a larger one across the board, the through-line of our broader work on how small CRE firms out-operate institutional competitors. For a concrete look at how the pieces fit into a week, inside a small shop’s AI-augmented deal pipeline walks the whole cadence stage by stage.
The first step is not buying a tool. It is pricing what you already spend — all five lines, not just the hours — so the next decision rests on the real number instead of the smallest one.
FAQ
What is the real cost of screening deals by hand?
The real cost is the deals you never underwrote, not the analyst hours you spent. Manual screening costs a lean firm on five lines: direct labor (visible), the throughput ceiling of deals that arrive above your weekly capacity and never get opened, the access you lose when a faster firm engages a broker first, decision-quality erosion from applying the buy box unevenly under fatigue, and the misallocation of senior judgment time to data entry. The labor line is usually the smallest of the five. The invisible lines — throughput and senior-time — carry most of the cost, which is why firms that price only the hours understate the problem badly.
How do I calculate what manual deal screening costs my firm?
Anchor on the visible labor line — per-deal screen time times deals screened per month times a loaded rate — then estimate the four hidden lines against it. Value the throughput ceiling at the expected value of one good acquisition you miss per year, because a single missed deal dwarfs a year of hours. Add the deals lost each year to a firm that engaged first, the full underwrites wasted on deals that should have died in screening, and the principal’s extraction hours priced at their best alternative use. The number does not need to be exact; it needs to show you that the hours are the small line.
Isn’t manual screening just a few hours a week?
The hands-on-keyboard time is a few hours a week; the cost is not. Screening is a throughput constraint, not a task you finish, so the real expense is everything above the line your team can process — the deals skimmed off on a subject line during a busy week, including occasionally the best deal of the quarter. A few hours of visible labor can sit on top of a much larger invisible cost in missed and mis-killed deals. Counting only the hours is the most common way lean firms underestimate what their screening process actually costs them.
Can AI screen commercial real estate deals for me?
AI can screen — extract terms, rank against your buy box, draft a first-pass summary — but it cannot underwrite or make the go/no-go call. Screening is triage: it tells you what to open. A general assistant like ChatGPT, Claude, or Gemini, or a dedicated platform, does that reading-and-ranking work in a fraction of the manual time and applies your criteria consistently. What it cannot do is verify a number against reality or decide what a deal is worth — it has no proprietary transaction data and will state a wrong figure as confidently as a right one. Keep the go/no-go and every offer-pricing number human.
What is the difference between AI deal screening and AI underwriting?
Screening ranks deals so you know what to open; underwriting models the cash flows so you can price an offer. Screening answers “is this worth my time,” underwriting answers “what is it worth.” Vendors often market the two interchangeably, which leads firms to buy a ranking tool when they need a model or a heavy modeling platform when they only need triage. A lean firm almost always needs the screening layer first, because triage — not modeling — is the binding constraint on how many deals reach a real underwrite.
Will AI make my firm miss good deals or advance bad ones?
Used correctly, AI reduces both errors; used as a decision engine, it can create new ones. A model applies the same buy box to every deal regardless of fatigue, which cuts the false negatives a tired human produces late in the day. The risk is the false positive from an unverified number — a figure pulled from the wrong column that the model reports with full confidence. The safeguard is grounding and review: every figure that prices an offer links to its source and gets checked, and a person owns the go/no-go. AI narrows the pile; a human decides.
How much does it cost to fix manual screening with AI?
It ranges from a modest training investment to a full build. Getting your team fluent enough to run deal packages through a general assistant themselves is workshop-scale — market rates for that kind of AI fluency training run roughly $2–15K — and often proves whether you need more. A custom-built screening workflow is a larger project, typically tens of thousands to low six figures depending on scope and integrations. The right choice depends on your deal volume and how standard your buy box is; price the status quo across all five cost lines first, so you are comparing against the real number.
Do I still need an analyst if AI screens the deals?
Yes — the analyst’s job shifts from extraction to judgment. The tool reads packages, ranks them, and drafts summaries; the person verifies the numbers that matter and decides what to underwrite and what to offer. The value of a lean team’s talent was never in transcribing rent rolls; it was in the go/no-go, the negotiation, and the broker relationships. Moving the mechanical work to AI frees that judgment for the work only it can do. The risk to manage is complacency — a good screening tool works well enough that people stop checking, so build the human verification in deliberately.
Is it safe to put confidential deal packages into an AI tool?
Only if the contract says so. Deal packages are confidential and often under NDA, so before uploading anything, get three answers in writing: where the data is stored, whether your uploads train shared models, and how you delete and export your history. “Your data is secure” is marketing; a data-processing addendum that commits the vendor not to train on your data and to delete on request is enforceable. A firm without an IT department relies on the contract as its main safeguard, so treat vague answers as a reason to walk away rather than a detail to sort out later.
Key takeaways
- The real cost of screening deals by hand is the deals you never underwrote, not the analyst hours — the visible labor line is usually the smallest of five.
- Manual screening costs a lean firm on five lines: direct labor, the throughput ceiling of un-opened deals, lost speed-to-first-look, uneven buy-box application, and senior judgment spent on data entry. The invisible lines carry most of the cost.
- Price all five against your own deal volume before deciding anything; a firm that counts only the hours is making the call on its least important number.
- AI removes cost by raising the screening throughput ceiling — it extracts, ranks, and summarizes consistently — but screening is triage, not underwriting. The go/no-go and every offer-pricing number stay human.
- Fixing manual screening ranges from workshop-scale fluency training to a custom build; the right level depends on deal volume and how standard your buy box is.
Not sure what manual screening is actually costing your firm, or where AI belongs in your deal flow without introducing risk? A short assessment maps that against your real deal volume, property types, and buy box faster than any tool comparison. Book your free AI-readiness assessment →
Arthur Wandzel
