Published:
Last updated:
A free AI due diligence toolkit for commercial real estate, and what it doesn’t do
CRE-AI-Skills packages 11 free AI prompts for commercial real estate due diligence, covering rent rolls, PSAs, title, and lease-to-system checks. Here is what each one does, and where a system of record has to take over.
AI due diligence tools for commercial real estate used to mean a vendor platform with a login and a contract. CRE-AI-Skills is a free, open-source library of 11 prompts that do the same job inside whatever AI tool a team already has open. It was built for Claude Code but works as a pasted system prompt in ChatGPT, Gemini, or any other chat-based model, and it doesn’t ask for an API key or a subscription.
The maintainers, publishing under the cre-ai-skills organization on GitHub, describe the goal plainly: standardize the document review that every acquisitions and due diligence team does by hand, so a rent roll audit or a title review starts from a consistent checklist instead of whatever the analyst remembers from the last deal.
What’s actually in the library
Each of the 11 skills targets one recurring due diligence task. They follow a standard folder structure, a SKILL.md entry point, a README, and reference material, which is what makes them portable across tools in the first place.
| Skill | What it checks | Best used for |
|---|---|---|
| Deal document intake parser | Classifies and normalizes a full deal package into structured data | The first pass on a new data room |
| Rent roll auditor | Unit-level exception report against a multifamily or mixed-use rent roll | Catching rent, occupancy, and lease-term mismatches |
| Operating expense auditor | Line-by-line adjusted expense schedule reconciled to underwriting basis | Reconciling seller-reported opex to what underwriting assumed |
| Property condition and environmental review | Reads PCA/PCR and Phase I ESA reports for capital needs | Flagging deferred capex and environmental exposure |
| Submarket study | Structured analysis of rents, comparables, supply, and demand | Framing a market memo |
| Tenant credit and exposure analyst | Evaluates creditworthiness, concentration, and lease durability | Sizing tenant risk across a rent roll |
| PSA reviewer | Clause-by-clause purchase and sale agreement review with issue annotation | A first pass before counsel |
| Title and survey reviewer | Reviews title commitments and ALTA surveys, classifies exceptions | Triaging title exceptions before the closing checklist |
| Loan document reviewer | Compares loan documents against term sheets and covenant summaries | Checking loan docs match what was negotiated |
| Estoppel and insurance tracker | Tracks estoppel collection status and verifies insurance programs | Keeping a closing checklist current |
| Lease-to-system auditor | Compares executed leases against system exports (Yardi, TRIRIGA, MRI) | Confirming what’s in the system of record matches what’s signed |
Why a checklist like this shows up now
Corporate real estate teams have moved past the pilot-or-skip decision on AI. JLL’s 2025 Global Real Estate Technology Survey, which polled more than 1,000 senior CRE decision-makers across 16 markets, found that 92% of corporate real estate teams have started piloting AI or plan to this year, up from under 5% just two years earlier. The same survey found only 5% report having achieved most of their program goals, and 81% report at least three existing systems underperforming expectations.
Read those two numbers together and the gap is obvious. Adoption isn’t the bottleneck anymore. What happens to the output after the pilot is. A checklist tool that runs once on a data room is a low-risk way to test that gap without touching a system of record, which is likely why CRE-AI-Skills isn’t the only entrant: similar collections of AI due diligence prompts, including ahacker-1/cre-agent-skills and sasha-deneux/claude-skills-cre, have appeared on GitHub around the same time. Standardized, shareable due diligence prompts are turning into a small category of their own, not one team’s side project.
Where the checklist hands off to a system of record
Ten of the eleven skills are one-time reviews: read a document, produce a report, move to the next deal. Lease-to-system-auditor is different. It compares an executed lease against what a system like Yardi, TRIRIGA, or MRI already has on file, which is the same reconciliation problem lease abstraction software already exists to solve on an ongoing basis.
That difference matters more than it looks. A due diligence skill runs the check once, at acquisition, and hands back a report. But a lease doesn’t stop generating events after closing: renewal windows open, escalations kick in, CAM reconciliations land, options lapse if nobody exercises them on time. Reading every lease before the renewal clock runs out is a standing job, not a one-time audit, which is exactly what CRE-AI-Skills’ own documentation acknowledges when it calls its outputs educational and informational resources, not production software, and tells users to have a licensed professional review anything before acting on it.
These outputs are educational and informational resources, not production software. A licensed professional should review anything before anyone acts on it.
That’s not a knock on the toolkit. It’s an honest description of what a checklist is for. REAL’s lease administration platform exists because the reconciliation lease-to-system-auditor runs once needs to stay true for the life of the lease, not just on the day someone remembered to run the prompt. The distinction is AI agents vs an IWMS: a well-built prompt can compress a review task most people do slowly by hand. It can’t be the system of record that has to be right every day the lease is in force.
How to try it
Installation is copy-and-paste. For Claude Code, clone the repo and copy the skills folder into ~/.claude/skills/ for every project, or into a single project’s .claude/skills/ folder for one deal at a time. For any other AI tool, open the relevant skill’s SKILL.md file and paste the contents into a custom instruction or system prompt. Nothing to configure, nothing to pay for.
Treat the output the way the maintainers themselves describe it: a fast, consistent first pass, not a substitute for the professionals who sign off on a deal.
Frequently asked questions
Is it safe to use CRE-AI-Skills on a live deal?
- It’s safe to use as a first-pass review, the way the maintainers describe it. It isn’t a substitute for a licensed attorney, accountant, or due diligence professional signing off on the deal, and the repo says so directly. Treat its output as a starting checklist, not a closing memo.
Does it cost anything or need an API key?
- No. It’s a set of text prompts (SKILL.md files) that run inside whatever AI chat tool a team already uses. There’s nothing to install beyond copying the files and nothing to license.
Does this replace lease administration software?
- No, and it isn’t trying to. The toolkit’s skills run once, at acquisition or during diligence. Lease administration is the ongoing job of keeping critical dates, rent, and obligations accurate for as long as the lease is in force, which is a different problem with a different failure mode: a missed date years after closing, not a flawed data room review on day one.
Who maintains CRE-AI-Skills?
- It’s published under the cre-ai-skills organization on GitHub. The repository itself doesn’t name the maintaining individuals or organization in its README.
See REAL run end to end.
Watch a demoRelated posts
AI for commercial real estate is more than a chatbot
Most AI for commercial real estate is a chatbot with a label. The work needs more: a system that reads your leases, proves its answers, and recovers cost.
AI agents vs IWMS: what enterprise occupiers actually need from each
An IWMS organizes your real estate data. An AI agent acts on it. Those are different jobs, and most enterprise occupiers currently have neither working well.
Why replacing your IWMS is the wrong question
The enterprise debate isn't IWMS or AI agents. It's what sits on top of your existing system to handle the work the IWMS was never designed to do.