Disclosure
Pacific Software Ventures (PSV) runs the CRE AI Institute and trains CRE teams, so we have a stake in how you learn this. The free steps below work on their own, and we say where something is ours.
How do you learn AI for CRE?
To learn AI for commercial real estate, start with one general assistant, Claude or ChatGPT, and use it on the one task that costs you the most time each week, like a rent roll, a T-12 or an OM. Practice on a deal where you already know the right answers, so you learn where the AI is wrong, not just where it is fast. Then save what worked as a reusable instruction, add the workflows for your seat (broker, acquisitions, lender, asset manager or developer), and connect it to your own files. You don't need to code.
The path, in order
Five steps, from your first prompt to work you'd put your name on. Each links to the guide for that step.
- Pick one assistant and learn how it reads a document. Upload a real OM or lease and ask it questions you can check. You're learning how it follows instructions and where it guesses. Use one paid plan, not five free ones. AI in 15 minutes, in CRE terms
- Run it on the one task that costs you the most time. Underwriting a first pass, abstracting a lease, summarizing a rent roll, drafting owner outreach. One task, finished on a live deal, teaches more than a month of reading about AI. How to use Claude for CRE
- Check its work on a deal where you know the answers. This is the step most people skip. A model can look right and still carry the wrong rent, miss a concession or invent an assumption. Practice on a deal with an answer key and you'll see exactly where to look. How to check Claude's work
- Save what worked as a reusable instruction. Turn the job you just did into a Project or a Skill, so it runs the same way next time, in your template, with your rules. Claude Projects for CRE
- Connect it to your own files and systems. Once the work is repeatable, point the AI at your deal folders, CRM and models, so answers come from your data and show where each number came from. Connecting Claude to your data
What to learn first, by role
The path is the same. The first task changes with your seat.
- Brokers. Start with owner and tenant prospecting: researching who owns a building and writing a first touch worth replying to. Then BOVs and OM drafts. AI prospecting for CRE
- Acquisitions and investors. Start with the first underwriting pass: reading the rent roll and T-12 into your own model, with every number traced to its source. AI underwriting for CRE
- Lenders. Start with the credit memo: pulling the borrower's documents into your memo template and flagging what doesn't tie out. AI for CRE lenders
- Asset managers. Start with variance reports and investor updates, built from the operating statements you already receive. AI for asset managers
- Developers. Start with feasibility: a yield-on-cost check and the sensitivities, from the site facts and your cost assumptions. Claude for real estate developers
Mistakes to skip
- Learning tools instead of tasks. A list of 40 AI tools teaches you nothing. One task you can now do in a fraction of the time teaches you everything.
- Trusting a number you didn't trace. If you can't point to the page or cell a figure came from, it isn't underwritten. Make the AI cite its source every time.
- Putting deal documents somewhere you shouldn't. Check your plan's data settings and your data vendors' terms before uploading confidential files or exports to an outside AI tool. Is AI safe for deal documents
- Treating it as a one-time course. The tools change every few months. What lasts is the habit: one new task a month, checked the same way.
Where to learn it, free to guided
- Free: our CRE guides and newsletter. The plain-English primer, the Claude playbook for CRE and a weekly newsletter with one workflow you can run that day. The AI for CRE Operators newsletter
- Free: Claude Academy from Anthropic. Courses on Claude itself. They aren't about real estate, so pair them with the steps above.
- Guided: the CRE AI Institute (ours). 38 lessons in 3 courses, worked on Desert Grove, a practice deal with a full set of documents and answer keys, plus weekly live office hours. The CRE AI Institute is $69 a month for Standard and $99 a month for Premium, after a 7-day free trial. See what the course covers
- For a whole team: training inside your firm (ours). We train your team on your own deals, documents and rules, and can build the workflows with you while we're there. AI training for CRE teams
- Comparing programs. Our side-by-side of the CRE AI courses, with dates and sources. The best AI course for CRE, compared
Common questions
- How do I learn AI for commercial real estate?
- Pick one assistant, use it on the one task that costs you the most time, and check its work on a deal where you already know the answers. Save what worked as a reusable instruction, add the workflows for your role, then connect it to your own files.
- Do I need to know how to code?
- No. The core skill is describing the task, the inputs and the output you want in plain language, then checking what comes back. Your real estate judgment matters far more than any technical background. Coding only comes in later, if you want to build your own tools.
- Should I learn ChatGPT or Claude for CRE?
- Either works for learning. Pick one and get good at it. We teach on Claude because it handles long deal documents and reusable Skills well, and our comparison page covers the differences for CRE work.
- How long does it take to learn AI for CRE?
- It depends on how often you use it on real work. Using it on one live task a week builds the habit faster than any course on its own. The CRE AI Institute has 38 lessons you can take at your own pace.
- Is there a free way to learn AI for commercial real estate?
- Yes. Our CRE guides and newsletter are free, and Anthropic's Claude Academy is free, though it isn't about real estate. The CRE AI Institute starts with a 7-day free trial.
- What should a CRE broker learn first?
- Prospecting: researching owners and writing a first touch that's specific to the building. It's repetitive, it's measurable in replies, and a person still approves every message before it goes out.






