Commercial real estate underwriting is due for a reset.
For years, "automating underwriting" meant spreadsheets with macros, Word templates, and analysts copy-pasting data from one system to another. Teams told themselves it was temporary. It wasn't.
In 2026, the firms doing 15–20 deals per week aren't the ones with the most analysts. They're the ones that cut the manual loops entirely.
This post is about what's actually working — not what vendors claim, not what's in a pitch deck. What institutional CRE teams are running in production today, what they're dropping, and what the benchmarks actually look like.
The Current State of CRE Underwriting
Most institutional CRE teams are still underwriting deals the same way they did in 2015.
The typical workflow: an analyst pulls rent rolls from a data room, inputs them into a Argus or Excel model, runs sensitivity analyses manually, writes a summary memo, and sends it up the chain. For a single deal, this takes 2–4 hours minimum. For a 10-deal week, it's the entire team's bandwidth.
And it's not just the modeling. The real bottleneck is data aggregation: rent rolls, operating expense histories, lease abstracts, market comps, debt quotes, and exit assumptions all live in different places. The analyst spends 60–70% of their time gathering and formatting data, not analyzing it.
A 2025 survey of institutional real estate investment managers found that 67% of CRE firms were still using Excel as their primary underwriting tool, despite the availability of purpose-built platforms. Of those, 44% described their workflow as "highly manual" with significant copy-paste dependencies.
The firms that have moved past this aren't running magic AI. They're running smarter workflows.
Why Manual Workflows Fail at Scale
The math is simple: manual underwriting doesn't scale.
At 5 deals per week, manual workflows are manageable. At 15–20 deals per week — the deal volume institutional teams target to stay competitive — manual workflows collapse in three predictable ways.
1. Analyst bandwidth becomes the ceiling.
Every deal requires the same overhead: data extraction, format normalization, model input, sensitivity runs, memo writing. Adding analysts adds headcount cost but doesn't change the per-deal time. You can't out-hire a broken process.
2. Consistency degrades as volume increases.
When analysts are rushed, they skip sensitivity scenarios, use outdated market assumptions, and make ad-hoc adjustments that don't follow the firm's underwriting standards. A 2024 analysis of underwriting variance across 12 institutional firms found that deals reviewed under time pressure showed 30% higher deviation from the firm's stated investment criteria than those reviewed under normal conditions.
In other words: when you're moving fast, you're making worse bets.
3. Data becomes siloed and stale.
Manual workflows depend on analysts pulling data on-demand. This means the data is only as good as the analyst's last pull. Rent roll updates mid-deal, revised market comps, or changed debt terms often don't make it into the model until the next version — if at all.
The result: institutional teams are flying partially blind on deal volume, making inconsistent decisions under time pressure, and burning analyst time on formatting instead of analysis.
The Anatomy of an Automated Underwriting Workflow
Firms that have solved this — and there are enough of them now to see the pattern — build automation in four layers.
Layer 1: Document ingestion and data extraction
This is where most automation starts. Modern CRE underwriting tools can ingest rent rolls, T12 operating statements, lease abstracts, and market data feeds and extract the relevant fields without manual re-entry.
The critical capability here isn't just OCR. It's contextual extraction: understanding that a line item labeled "RE Tax" in one rent roll format is the same as "Real Estate Taxes" in another, and mapping it to the right field in the underwriting model.
Firms running this layer report 40–60% reduction in data aggregation time per deal.
Layer 2: Model generation and normalization
Once data is extracted, it needs to flow into the underwriting model — and that model needs to match the firm's assumptions, sensitivity frameworks, and output format.
The key to making this work: model templates that are structured, not just formatted. Firms that export a Word template with some Excel formulas don't get automation — they get fancier manual work. The model needs to accept structured data input and produce outputs in a consistent format that feeds directly into the deal memo or investment committee package.
Layer 3: Market data integration
Automated workflows pull market comps, cap rate benchmarks, and debt quotes from integrated data sources rather than requiring analysts to search and input manually. This keeps assumptions current and reduces the variance between deals reviewed under different conditions.
The 2026 benchmark for top-tier firms: market data refreshes on a weekly cadence, with automated alerts when a comparable property's value changes by more than 5%.
Layer 4: Deal memo and IC package generation
The final layer: automated output generation. The underwriting model produces a deal memo in the firm's standard format — investment thesis, sensitivity table, key risks, recommendation — without the analyst re-writing anything.
This is where the time savings compound most visibly. Firms that have fully implemented this layer report saving 1.5–2.5 hours per deal on memo writing alone. At 15 deals per week, that's 22–37 analyst hours. Per week.
What to Look for in CRE Underwriting Software
Not all underwriting tools are built for institutional workflows. Here's what matters when evaluating platforms in 2026.
Integration with existing tools. If the platform can't ingest data from your data room provider (Intralinks, ShareFile, Box) and output to your IC reporting format, you'll spend as much time managing the tool as you do underwriting deals.
Assumption consistency controls. The software should enforce your firm's underwriting standards — not just produce a model that looks like yours. Look for tools that let you configure sensitivity ranges, required scenarios, and output fields at the firm level, so every analyst is working within the same framework.
Audit trail and version control. CRE underwriting involves legal and fiduciary responsibility. Every data input, assumption change, and model adjustment should be logged with a timestamp and user ID. If the platform doesn't have this, it's a compliance risk.
Lease abstracting capability. This is becoming a differentiator. Tools that can parse and abstract lease terms — rent steps, renewal options, expense stop amounts, TI allowances — directly from the data room save the most time in the current workflow. Most platforms still require manual lease abstracting.
Speed and deal volume. Ask the vendor what their typical deal turnaround is. Firms running 15+ deals per week need a platform that can process a full deal — from data ingestion to memo output — in under 45 minutes. If the vendor can't give you a clear answer, that's a signal.
Key Benchmarks for 2026
| Metric | Industry Average | Top Quartile Firms |
|---|---|---|
| Hours per deal (full underwriting cycle) | 3.5–4.5 hrs | 1.5–2 hrs |
| Data aggregation time per deal | 2–2.5 hrs | 45–60 min |
| Sensitivity scenarios modeled per deal | 2–3 | 5–7 |
| Weekly deal capacity per analyst | 2–3 deals | 6–8 deals |
| Time from data room upload to IC-ready memo | 4–6 hours | 60–90 minutes |
| Assumption variance across deal reviews | 25–35% | 8–12% |
The gap is significant. Top quartile firms aren't working twice as hard — they're running on automated workflows that handle the repetitive work. The result is more deals underwritten at higher quality, with analysts spending their time on judgment calls instead of copy-paste.
What Firms Are Doing Wrong
If the benchmarks are this favorable for top quartile firms, why are most teams still below them?
The most common failure modes in 2026:
Treating automation as a data problem. Buying a new platform and migrating historical data doesn't solve the workflow. Automation works when it's embedded in the deal process — starting at data room ingestion and ending at IC memo output. Piecemeal automation (just the rent roll, just the model) still requires analyst time and creates new handoff bottlenecks.
Holding out for the "perfect" platform. Several firms have spent 12–18 months evaluating underwriting tools without implementing any of them. In that time, they've lost hundreds of analyst hours to manual work. An 80% automated workflow that exists beats a 100% ideal workflow that doesn't.
Underestimating the memo problem. Most firms focus on model automation and underinvest in memo generation. But memo writing is the most time-intensive part of the underwriting cycle and the hardest to automate without a well-structured model template. Firms that don't build the memo layer into their automation strategy end up with a solution that saves less than half the expected time.
Ignoring lease abstracting. If your analysts are still manually abstracting leases, that's 20–30 minutes per tenant, per deal. For a 15-tenant asset, that's 5–7 hours of manual work per deal. This is the single highest-ROI automation target for most institutional teams, and most platforms still treat it as optional.
Where AI Fits in the 2026 CRE Underwriting Stack
A brief note on AI, since every vendor is promising it and most buyers are confused about what it actually means for underwriting.
AI works in CRE underwriting in two confirmed use cases: document extraction and lease abstracting (reducing manual data entry time) and scenario sensitivity generation (running structured stress tests across a defined range of assumptions without analyst input).
What AI doesn't do — yet — is replace underwriting judgment. The investment thesis, market context, deal-specific risks, and relationship-informed assumptions still require an analyst. The best automation stack handles the data and the model. The analyst handles the story.
Firms that use AI as a layer within a structured workflow — not as a standalone replacement for the process — are reporting the most consistent ROI. For a deeper look at which AI capabilities are delivering that ROI today, see AI in Commercial Real Estate Underwriting — What Is Actually Working in 2026.
The Bottom Line
CRE underwriting automation is not theoretical in 2026. There are firms running 15–20 deals per week with 4-person analyst teams, producing IC-ready memos in under 90 minutes per deal.
The path there isn't complicated:
- Automate document ingestion and data extraction first — highest ROI, fastest to implement.
- Build model templates that enforce your firm's underwriting standards — this is the consistency layer.
- Layer in lease abstracting — single highest untapped time savings for most teams.
- Build the memo generation last — it's the most complex but delivers the most compounding value.
The firms still running manual workflows aren't behind because they lack resources. They're behind because they haven't made the decision to move. That's the only real blocker.