Category: CRE AI Assistants & Copilots

Every CRE analyst who has spent an afternoon buried in a 47-page triple-net retail lease knows the feeling. The document is dense with nested definitions, buried termination clauses, and rent escalation schedules that reference other sections, which reference other exhibits. The abstract has to be done before the investment committee call. The deadline is real. The work, however, is almost entirely mechanical — locate, extract, format, repeat. It is the kind of task that makes talented people feel like expensive photocopiers.

The industry has recognized this problem for years. Professional lease abstraction services charge between $90 and $250 per lease. A trained analyst takes four to eight hours to produce a clean abstract from a single commercial document. Yardi, MRI, Prophia, and a dozen other platforms have built purpose-specific AI tools to automate parts of the workflow, and they have delivered real compression — getting initial data extraction down to as little as seven minutes on straightforward documents. JLL has projected that AI applied to these administrative tasks can free roughly 20 percent of asset managers’ time for higher-value work. The ROI math is not subtle.

What those numbers obscure is the access problem. Purpose-built lease abstraction software carries enterprise pricing, integration requirements, and implementation timelines that put it out of reach for the boutique acquisition shop, the family office analyst, the mid-market property manager running a 30-asset portfolio in Excel. There is a gap between “the big platforms have solved this” and “I personally have this solved.” Claude Skills, launched by Anthropic in October 2025 and quietly emerging as the most flexible AI workflow tool in CRE, closes that gap in a single afternoon. The following guide shows exactly how to build one — and why the architecture matters more than the specific commands.

This article sits within BestCRE’s CRE AI Assistants & Copilots coverage, part of our broader analysis of how AI is reshaping the 20 sectors of commercial real estate. The lease abstraction workflow described here is one of the most immediate, high-ROI applications of AI in CRE operations — and it requires zero software budget to implement.

What Claude Skills Actually Are — and Why They Are Not Just Fancy Prompts

Before getting into the build process, it is worth being precise about what a Claude Skill is, because the distinction matters for how you design one. A Skill is not a saved prompt. It is not a chatbot. It is a structured folder containing a SKILL.md file — written in simple Markdown — that encodes procedural knowledge, formatting standards, domain context, and output specifications. When you reference a Skill in Claude, it reads those instructions before processing your document, then applies them consistently across every subsequent use. Anthropic introduced the Agent Skills open standard on October 16, 2025. By December 2025, the company had added organization-wide Skill management and a directory of partner-built Skills. In January 2026, Anthropic published a 32-page guide to building Skills — covering design, testing, and distribution.

The operational implication for CRE practitioners is significant. A one-off prompt that says “please summarize this lease” produces variable results. The output quality depends on how you phrased the request that day, what context was already in the conversation, and whether Claude happened to emphasize the right sections. A Skill inverts that dynamic entirely. The Skill is where your standards live — which fields to extract, in what order, formatted how, with what level of analytical commentary on unusual clauses. Every lease abstract produced through the Skill reflects the same playbook. That consistency is what makes it genuinely useful at the portfolio level, not just for one-off requests.

Skills are available to Claude Pro subscribers ($20 per month), as well as Max, Team, and Enterprise plan users. They work across claude.ai, Claude Code, and the Claude API — meaning the same Skill you build in the browser interface can eventually be deployed programmatically across a deal pipeline. For individual analysts and small shops, the browser-based workflow described here is the fastest path to value.

The Skill Creator Skill: How to Build Your Tool Without Writing a Single Line of Code

Anthropic ships Claude with a pre-built “skill creator” Skill — a meta-tool that helps you build other Skills. This is the fastest starting point for CRE practitioners who want to create a lease abstraction workflow without writing technical documentation from scratch. The process takes roughly five minutes and produces a deployment-ready Skill file. Here is the exact sequence.

First, open a new Claude conversation and invoke the skill creator. You can find it in the Skills directory within your Project settings, or invoke it directly. Tell Claude what you are trying to build: “I want to create a Skill that automates commercial lease abstractions for CRE. The output should be a professionally formatted Word document with clean tables, organized by section, following my firm’s standard abstract template.” Claude will then ask a series of clarifying questions — the purpose of the Skill, the output format, the level of analytical commentary required, and whether you want the Skill to flag unusual or potentially adverse clauses for human review. Answer these as specifically as you can. The quality of those answers determines the quality of the Skill.

Second, feed Claude a sample lease abstract template. If your firm has a standard template — even a rough one in Word or Excel — paste it into the conversation or upload the file. Claude will reverse-engineer the structure, identify the fields your template captures, and build the Skill’s extraction logic around your actual format rather than a generic one. If you do not have a template yet, this is a good moment to build one by telling Claude what categories matter to your analysis: key dates, tenant and guarantor information, base rent and escalation schedule, expense reimbursement structure (gross, NNN, modified gross), renewal and termination options, co-tenancy clauses, permitted use restrictions, and any assignment or subletting provisions.

Third, let Claude run research. The skill creator will proactively identify CRE-specific terminology, common lease structures by asset class (retail, office, industrial, multifamily), and the fields most likely to affect underwriting. This research pass is what separates a generic document summarizer from a Skill that actually understands why an anchor co-tenancy clause in a grocery-anchored retail lease matters differently than the same clause in a neighborhood strip center. Watch what Claude identifies and push back where its interpretation does not match your analytical priorities.

Fourth, review and save the SKILL.md output. Claude will generate the complete Skill file. Read through it before deploying. The best Skills are specific about output format, explicit about which fields to prioritize when the lease language is ambiguous, and direct about what constitutes a “flag for review” versus a standard provision. If your Skill is too vague, the abstracts it produces will be too generic to be genuinely useful. If it is too rigid, it will struggle with unusual lease structures. The right level of specificity comes from a short back-and-forth during the build.

Running Your First Lease Abstract: The Live Workflow

Once the Skill is saved, the operational workflow is minimal. Upload the lease document — PDF is the standard format, though Claude handles scanned documents with reasonable accuracy when the scan quality is adequate. Reference the Skill and give a single directive: “See attached lease. Please prepare abstract per the lease abstraction Skill.” Claude reads the Skill first, then processes the document against those instructions. The output arrives formatted and structured, not as a wall of prose that still requires manual reformatting.

For a standard commercial lease of 30 to 50 pages — the typical length for a single-tenant net lease or a mid-sized office or retail document — Claude will produce a clean, structured abstract in under five minutes. The output includes tables for the financial terms (base rent, escalation schedule, CAM caps if applicable), a plain-language summary of the critical dates (commencement, expiration, rent commencement, option exercise deadlines), and a flagged section for any provisions that deviate from standard market terms. A Taco Bell ground lease with partial redactions, as a concrete example, still yields a usable abstract — Claude notes where information was redacted and marks those fields accordingly rather than inventing data to fill gaps.

The Word document output — triggered by Claude’s built-in docx Skill, which runs automatically when document creation is requested — arrives with proper formatting: section headers, clean tables, consistent font treatment. It is ready to drop into a deal file or share with an investment committee without post-processing. That last point is worth emphasizing. The hours lost in traditional lease abstraction are not just the reading time — they are the reformatting time, the “make this look like our standard template” time, the back-and-forth between analysts using slightly different conventions. A Skill eliminates that variation by design.

What to Extract: The Anatomy of a CRE Lease Abstract Worth Using

The value of a lease abstract is determined entirely by whether it captures the information that actually affects underwriting, portfolio management, and risk assessment. Generic abstracts that log basic dates and rental rates are operationally useful but analytically thin. The best abstracts — and the best Skills — are built around what you would actually want to know before making a capital allocation decision. Here is the field architecture worth encoding in your Skill.

The financial terms block should capture base rent in absolute dollar and per-square-foot terms, the escalation schedule (fixed percentage, CPI-tied, or step-up at specific dates), the full expense reimbursement structure with caps, and any percentage rent provisions for retail leases. Critically, the Skill should be instructed to calculate implied yield on rent at the stated cap rate range your firm uses — a simple instruction that turns a data extraction into a preliminary underwriting check.

The lease term block should include commencement date, rent commencement date (these are frequently different), expiration, all renewal option periods with notice requirements and rent reset mechanics, and any early termination rights with the associated penalty calculation. This section is where most manual abstraction errors occur — escalation schedules and option deadlines buried in exhibit language are commonly missed.

The tenant and guaranty block should capture the legal entity name of the tenant (not just the trade name), the guaranty structure and guarantor creditworthiness indicators, and any carve-outs or limitations on the guaranty. For net lease investors analyzing single-tenant assets, this section is the credit underwriting foundation. A Taco Bell franchise operated by a 50-unit operator carries meaningfully different credit risk than one operated by the company-owned entity — the lease abstract is where that distinction should be visible.

The risk flags section is where a well-built Skill adds its highest value. Instruct Claude to identify and summarize any co-tenancy provisions, exclusivity clauses, prohibited use restrictions, assignment or change-of-control provisions, audit rights, and ROFO or ROFR provisions. These are the clauses that affect a property’s value to a future buyer and its vulnerability to adverse tenant actions. Attorneys catch them during due diligence, but abstracting them early — before a deal is fully committed — gives the investment team a structural read on risk before significant capital is deployed.

Expanding the Skill Library: Beyond Lease Abstracts

The lease abstract Skill is the fastest demonstration of what this architecture can do, but it is not the ceiling. The same build process — invoke skill creator, specify the output, feed it a template or framework, let it research the domain, save the Skill — works for any repeatable CRE analytical task. The skills worth building next follow directly from where the most analyst time is currently consumed.

An offering memorandum generation Skill encodes your firm’s OM format, deal narrative conventions, and financial summary structure so that a new OM starts from a 70 percent complete draft rather than a blank page. A market analysis Skill can be built around a specific market intelligence framework — defining which data sources to synthesize, which metrics to prioritize, and how to structure the forward-looking thesis. Investment framework Skills that encode specific decision-making approaches — capital allocation criteria, risk weighting models, portfolio construction logic — turn each deal analysis into a structured evaluation against explicit standards rather than an ad hoc judgment call. The consistency those Skills produce is valuable both for individual analysts developing their discipline and for investment committees evaluating submissions from multiple team members.

One practical constraint to note: Skills are token-intensive. A comprehensive lease abstraction Skill loaded with domain context, formatting instructions, and flag criteria consumes meaningful context window before the actual lease document is even processed. Claude Pro’s usage limits will be hit faster when Skills are in active use — something Anthropic has acknowledged as a design tradeoff between capability and compute. For firms processing high volumes of leases, the Max or Team plan is worth evaluating against the time savings. Even at the Pro tier, the math favors the Skill: at $20 per month for unlimited Skills usage within the usage cap, the break-even against a single outsourced lease abstract at $90 to $250 is immediate.

The Strategic Argument: Why Workflow Automation Is Now a Competitive Differentiator

The instinct among CRE practitioners has been to treat AI workflow tools as efficiency plays — things that make existing processes faster. That framing underestimates what is actually happening. When a boutique acquisition shop can process lease abstracts at the same speed as an institutional platform running enterprise software, the speed advantage that platform enjoyed narrows to near zero. When an analyst can build a decision-framework Skill that applies consistent underwriting logic across every deal, the consistency advantage that large shops gained from having senior oversight on every transaction extends to smaller operations. The gap between institutional-grade analysis and solo-practitioner analysis is not closing gradually — it is collapsing on specific tasks where AI automation has reached deployment-ready quality.

This is the broader dynamic BestCRE has been tracking across its coverage of AI’s impact on CRE business models. The $12 billion that Wall Street erased from CBRE’s market cap during record earnings was not a verdict on CBRE’s fundamentals — it was a read on the labor-intensive components of brokerage and advisory services that AI is directly displacing. Lease abstraction is one of those components. The practitioners who build workflow automation now are not just saving time on individual tasks — they are redefining what a lean, high-output CRE operation looks like.

The sophistication ceiling for Claude Skills has not yet been reached. Anthropic’s January 2026 Skills guide describes multi-Skill workflows where one Skill hands off structured output to another — a lease abstract Skill feeding a portfolio analytics Skill, which feeds a reporting Skill. That architecture is not hypothetical. It is buildable today by any practitioner willing to spend an afternoon on setup. The question for CRE operators is not whether AI will automate the administrative layer of their workflows. It is whether they build that automation themselves, on their terms, with their standards embedded — or whether they wait for a vendor to deliver a packaged version at enterprise pricing and integration overhead.

Step-by-Step Build Checklist

For practitioners ready to build immediately, here is the compressed build sequence. Open Claude on a Pro, Max, Team, or Enterprise plan. Navigate to your Projects and open the Skills section — create a new Project if needed, as Skills are project-scoped by default. Invoke the skill creator by searching the Skills directory or typing “@skill-creator” in the conversation. Tell it you want a CRE lease abstraction Skill with Word document output. Answer its questions about your output preferences, field priorities, and flagging criteria. Upload your existing abstract template if one exists, or describe your preferred structure. Allow Claude to complete its domain research pass — do not skip this; it materially improves the Skill’s handling of asset-class-specific lease language. Review the generated SKILL.md file, make any adjustments, and save. Test the Skill on a real lease. Iterate on the field priorities based on what the first output gets right and what it misses.

The setup time is genuinely under an hour. The time savings begin on the first lease you run through it.

Skills Are the Starting Point. A Full CRE AI Agent Team Is the Destination.

A lease abstraction Skill is a single agent doing a single job. It is a powerful demonstration of what AI can execute on your behalf when given the right instructions — but it operates in isolation. The lease gets abstracted. Then you take that output and manually feed it into the next step: the underwriting model, the investment memo, the lender package, the asset management report. The workflow compression is real, but the handoffs between steps are still manual, still slow, still yours to manage.

The logical next layer is not more Skills. It is a coordinated team of AI agents — each one specialized, each one operating on your firm’s specific standards, and each one passing structured output to the next agent in the chain. A lease abstract agent feeds a deal screening agent. A market research agent informs a risk assessment agent. An investor reporting agent assembles everything into a formatted deliverable. The individual tasks collapse from hours to minutes. The connected workflow collapses from days to hours. That is not a hypothetical architecture — it is what a purpose-built CRE AI Agent Team looks like when deployed against a real deal pipeline.

Building that kind of system requires more than an afternoon with Claude’s skill creator. It requires understanding how agents communicate, how to structure handoffs without data loss, and how to encode your firm’s judgment and standards into each agent’s operating logic rather than defaulting to generic outputs. That is precisely the problem 9AI was built to solve.

9AI designs and deploys custom CRE AI Agent Teams — built around your asset classes, your underwriting framework, your deal process, and your reporting requirements. Not packaged software. Not a chatbot with a CRE skin on top. A configured team of specialized agents that executes the analytical and operational work your firm does every day, at the speed and consistency that manual workflows can never match. If you have seen what a single Skill can do and want to understand what a full agent team looks like against your specific workflow, that conversation starts at 9AI.co.

BestCRE is the independent authority on commercial real estate AI, covering the 20 sectors of CRE through institutional-quality analysis for practitioners, operators, and allocators. Our coverage tracks the AI tools and workflow architectures reshaping how CRE professionals source, underwrite, and manage assets — from lease abstraction to data center infrastructure to the AI tools transforming healthcare real estate investment strategy.

Frequently Asked Questions

What is a Claude Skill and how does it differ from a regular prompt for lease abstraction?

A Claude Skill is a structured instruction file — written in Markdown and stored in a SKILL.md format — that encodes procedural knowledge, formatting standards, and domain-specific logic that Claude loads before processing any document. Unlike a one-off prompt, which produces variable results depending on how it is phrased and what context is active in the conversation, a Skill applies the same standards every time it is invoked. For lease abstraction, this means the same fields are extracted, the same flags are raised, and the same output format is produced whether you run one lease or one hundred. Anthropic launched the Agent Skills standard in October 2025 and it is available to Pro, Max, Team, and Enterprise plan subscribers. The practical distinction matters: one-off prompting is ad hoc experimentation; a Skill is a deployed workflow asset that compounds in value across every document it processes.

How does a Claude Skills-based workflow affect the time and cost of commercial lease abstraction?

Manual commercial lease abstraction takes four to eight hours per document, with outsourced services costing $90 to $250 per lease. Purpose-built AI platforms have reduced initial data extraction to as little as seven minutes for straightforward leases. A Claude Skill-based workflow operates in the same speed range — typically under five minutes for a standard 30- to 50-page commercial lease — with no per-lease cost beyond the Claude subscription. At $20 per month for a Pro plan, the break-even against a single outsourced abstract is immediate. JLL estimates that AI automation of administrative tasks like lease abstraction can free roughly 20 percent of asset managers’ time for higher-value work. At the portfolio level, that figure compounds quickly: a 50-asset portfolio with annual lease reviews represents 200 to 400 analyst hours at current manual rates, collapsible to a fraction of that with a properly built Skill.

What information should a CRE lease abstract capture, and what makes a Skill better at extracting it than generic AI?

A professionally useful CRE lease abstract captures five core categories: financial terms (base rent, escalation schedule, expense reimbursement structure, percentage rent); lease term (commencement, rent commencement, expiration, renewal options with notice deadlines and rent reset mechanics); tenant and guaranty (legal entity name, guaranty structure, guaranty carve-outs); critical risk provisions (co-tenancy, exclusivity, prohibited use, assignment restrictions, ROFO/ROFR); and property-specific terms (permitted use, alterations rights, signage, parking). What makes a Skill materially better than generic AI querying is asset-class specificity. A Skill built for net lease retail understands why a co-tenancy provision tied to an anchor tenant’s occupancy creates different risk than one tied to occupancy percentage — and flags it accordingly. Generic AI treats all clauses equally. A well-built Skill treats them the way an experienced asset manager would.

What other CRE workflows can be automated with Claude Skills beyond lease abstraction?

The same Skill architecture applies to any repeatable analytical task in CRE. High-value Skills in active development among CRE practitioners include offering memorandum generation (encoding deal narrative conventions and financial summary structure), market analysis reports (defining data source hierarchy, key metrics, and forward-looking thesis structure), investment decision frameworks (encoding capital allocation criteria and risk weighting logic), and due diligence checklists (ensuring consistent documentation across deal teams). Multi-Skill workflows — where one Skill’s structured output feeds into another — are architecturally possible today and enable sequences like lease abstract → portfolio analytics → investor reporting. The practical constraint is token consumption: complex Skills loaded with domain context consume meaningful context window before the task document is processed, which affects usage limits at lower subscription tiers.

Who can access Claude Skills and is this workflow practical for smaller CRE operations?

Claude Skills are available on Claude Pro ($20 per month), Max ($100 to $200 per month), Team, and Enterprise plans. They are not available on the free tier. For individual analysts and small CRE shops — boutique acquisitions teams, family offices, mid-market property managers — the Pro tier is the practical entry point. The workflow is particularly well-suited to smaller operations precisely because they lack access to enterprise lease abstraction platforms at $500 to $2,000 per month. A Skills-based workflow on Claude Pro delivers institutional-quality output consistency at a subscription cost that breaks even against a single outsourced abstract. The build time is under one hour. The operational lift afterward is minimal — upload a lease, reference the Skill, receive a formatted abstract. For high-volume operations processing dozens of leases per month, the Max or Team plan avoids hitting usage limits on the Pro tier, and the ROI against outsourcing or purpose-built software is even more pronounced.

Related Reading

Best CRE AI Barometer: Cushman & Wakefield Just Built One. Here’s How It Scores.

AI Erased $12 Billion from CRE Brokerage Stocks. Here’s What That Actually Means.

Best CRE Sectors: The 20 Categories of Commercial Real Estate AI in 2026