Claude for Commercial Real Estate: Models, Skills, and Workflows

Short answer: what Claude is, and why CRE firms are standardizing on it

Claude is Anthropic's family of AI models, and it has become the default choice for commercial real estate firms doing document-heavy AI work: underwriting, IC memos, lease abstraction, off-market sourcing, LP reporting. The reason is not brand preference. CRE work runs on long, messy documents, offering memoranda, rent rolls, T-12s, leases, and Claude's long context window lets a model read the entire document in one pass instead of chunking it and losing structure. Claude also tends toward honest analysis: it flags when a rent roll does not reconcile or an OM's numbers look optimistic, rather than smoothing over the gap to produce a clean-looking answer. For underwriters and asset managers, that bias toward flagging inconsistencies over papering them over is worth more than a marginally higher benchmark score.

The practical result is that a firm can now build a real AI system, not a chatbot bolted onto email, on top of Claude: a copilot that reads a deal package and drafts the underwriting, an agent that triages off-market signals, a reporting layer that turns portfolio data into an LP letter. JLL has reported that 88% of investors are piloting AI while only 5% are hitting their AI goals, which tracks with what we see: the model is rarely the constraint. The gap is almost always the system around the model, how it is wired into a firm's actual documents, data, and workflow.

The Claude model family for CRE work

Claude is not one model, it is a family, and CRE firms get the most value when they route different tasks to different models rather than defaulting to whichever one is open. At NextAutomation we run this routing in production: Claude Opus 4.8 for heavy execution work like full underwriting builds and multi-document IC memo drafting, Claude Sonnet for balanced day-to-day work like lease summaries and market write-ups, Claude Haiku 4.5 for fast, cheap tasks like classifying inbound leads or tagging documents, and Fable 5 for the planning and judging layer that checks the other models' work before it reaches a person. The economics matter here: a firm running every task through the heaviest model burns budget on work that a cheaper model handles just as well, while running everything through a cheap model produces underwriting mistakes that cost far more than the model bill.

We go deep on exactly which model fits which CRE task, and how to think about the cost-versus-quality tradeoff, in the dedicated guide.

What you can build with Claude for CRE

The pattern that works is Claude plus Skills plus MCP: Skills give Claude a repeatable, firm-specific instruction set for a given job, and MCP (Model Context Protocol) gives Claude live, governed access to your actual systems, your data room, your CRM, your accounting platform, instead of copy-pasted context. Put together, that combination is what turns Claude from a smart chat window into a system a firm can run deals through. The recurring builds we see across CRE:

• Underwriting: a copilot that ingests an OM, rent roll, and T-12, normalizes the numbers, and drafts the model, flagging anything that does not reconcile before an analyst spends hours on it.
• IC memos: turning a completed underwriting file and market data into a first-draft investment committee memo in the firm's own format and voice.
• Off-market sourcing: agents that watch for distress and ownership-change signals, score them against a firm's buy box, and hand a ranked list to acquisitions instead of a wall of unfiltered leads.
• LP reporting: turning portfolio and asset-level data into investor letters and quarterly reports on a schedule, with the numbers pulled from source rather than re-typed.

Each of these is a real system, not a demo. For the underwriting build specifically, see the live example.

For the mechanics of the two building blocks, Skills and MCP, we cover them in full elsewhere so this hub does not repeat them.

The open claude-skills-cre repo

NextAutomation is Claude-native, and we publish that work in the open rather than keeping it behind a paywall. Our claude-skills-cre repo on GitHub is a curated, growing set of Claude Skills built specifically for commercial real estate: underwriting helpers, lease abstraction, market analysis, LP communication, and more, written the way we write them for paying clients. It is free to pull, free to modify, and a fast way to see what a well-written CRE skill actually looks like before you write your own or hire someone to.

If you want the packaged, install-in-minutes version rather than the raw repo, we also ship a free downloadable pack.

Build it yourself vs bring in an implementation partner

Both paths are legitimate, and which one fits depends on what you are trying to do. Installing a Claude Skill and running your own Claude Project is a real, valid starting point: it costs nothing beyond a Claude subscription, it takes an afternoon, and for a solo broker or a small team it can cover a meaningful share of the recurring work, lead qualification, listing copy, first-pass lease summaries. That JLL number is worth sitting with here: 88% of investors are already piloting AI. The 5% hitting their goals are not the ones with the fanciest model, they are the ones who turned a pilot into a system that survives contact with a real deal pipeline, real data quality problems, and real compliance requirements.

That is usually where the do-it-yourself path runs out of road, not because Claude cannot do the work, but because production reliability, live data connections via MCP, human-review gates on anything investor-facing, and monitoring when a document format changes are a different job than writing a good skill. NextAutomation exists at exactly that line: we are an AI implementation partner for CRE, we build the custom AI system, wire it into your CRM and data room, and hand you something your firm owns and runs, not a vendor lock-in. We are Claude-native by design, which is why we route across the full Claude model family rather than defaulting to one model for every task, and why we publish claude-skills-cre in the open instead of treating skill-writing as the moat.

If you want to see what that looks like for firms like yours before talking to us, our case studies are anonymized but real.

Go deeper

This page is the hub. For the comparisons and the practitioner tools that sit underneath it:

• Which Claude Model for Commercial Real Estate: Opus vs Sonnet vs Haiku, mapped to CRE tasks.
• Claude vs ChatGPT for Commercial Real Estate: the CRE-specific comparison.
• ChatGPT vs Claude for Business: the general business comparison.
• Claude Code for Commercial Real Estate: for firms with technical teams building their own tools.
• Claude Projects for Commercial Real Estate: organizing recurring work without writing code.
• Claude Skills for Real Estate: The Complete Guide: the full skills reference.
• MCP for Real Estate Firms: connecting Claude to your live systems.