Curriculum, not software

AI agents for commercial real estate, explained

An AI agent for commercial real estate is Claude given tools, your deal files, and standing instructions, so it can carry a defined workflow, screening a deal, abstracting a T-12, or drafting an IC summary, rather than answering one prompt at a time. A chatbot answers questions; an agent owns a step of your pipeline and returns output in a format your team reviews. Most products sold as CRE agents are rented software, so the buy-versus-build question is really rent-versus-own. PSV builds CRE AI on Claude by Anthropic and teaches operators to build agents of their own.

Agent vs chatbot

A chat answers. An agent owns a step.

A chat answers questions

You ask, it answers, and the context ends with the conversation. Useful, but every deal starts from zero.

An agent owns a pipeline step

It carries one step of your pipeline with standing rules, access to the real files, and an output format your team reviews.

Everyone selling AI agents for CRE is selling software. But an agent is not a product category, it is a capability, and one a CRE professional can learn. Where agents sit in the wider toolkit is covered in What is CRE AI.

The stack

Five layers of a working agent

01
The model
Claude by Anthropic, the reasoning underneath. Well-suited to long deal documents and detailed, multi-step instructions.
02
The instructions
Your rules, written down once: what to screen for, which assumptions to apply, what the output must look like.
03
The knowledge
Your deal files. The OM, T-12, and rent roll the agent works from, so answers come from your documents, not general knowledge.
04
The tools
What the agent may touch: files, spreadsheets, data connections. The scope you grant defines what it can and cannot do.
05
The review gate
A person signs off. The agent drafts, cites its sources, and stops. Your team approves what moves forward.

The model is the least of these decisions; the reasoning behind PSV's choice is in Claude vs ChatGPT for CRE. The layers that make or break the agent are yours: the rules, the files, and the gate.

This stack is exactly how the CRE AI Institute teaches members to build their first agent, on a real practice deal, a multifamily asset called Desert Grove.

Rent vs own

Own the agent, keep the trail

Most agent products rent you a black box: your files go into someone else's system and the reasoning stays hidden. An agent you own inverts that.

Runs on your files, in your environment
Every number traceable to the document it came from
A citation trail you keep, not one that expires with a subscription

That is the PSV thesis: cited truth as a system property, not a feature. Operators can build this themselves. Firms that want it built and deployed across the team work with PSV.

Frequently asked

AI agents for CRE, answered

What is an AI agent for commercial real estate?: An AI agent for commercial real estate is a model like Claude given standing instructions, access to your deal files, and defined tools, so it can carry a full workflow such as screening a deal or abstracting a T-12. Unlike a chat, it follows your rules every time and returns output in a format your team reviews.
How is an AI agent different from a chatbot?: A chatbot answers the question in front of it and forgets the context when the conversation ends. An agent owns a step of your pipeline: it starts from your rules, works from the real files, and produces the same reviewable output every run.
How do I build an AI agent for real estate?: Pick one defined workflow, write down the rules you already apply, give Claude those instructions plus the relevant deal files, and require a source citation behind every number. Keep a person signing off on the output. The CRE AI Institute teaches this build step by step on a real practice deal.
Do I need to code to build an AI agent?: No. The instructions that define an agent are written in plain English: what to check, which assumptions to apply, what the output must look like. Code becomes relevant when a firm wants agents wired into its systems, which is the kind of build PSV does for enterprise clients.
Should I buy an AI agent product or build my own?: Buying is really renting: your files flow through someone else's system and the reasoning stays hidden. Building means the agent runs on your files, in your environment, with a citation trail you keep. Start by learning how one works; the build is smaller than most operators expect.
What can an AI agent do for a CRE team?: Well-defined document workflows suit agents best: screening incoming deals against your criteria, abstracting a T-12 or rent roll into your template, and drafting an IC summary with sources attached. A person reviews every output before it moves forward.

Keep reading

Claude for real estate underwriting Claude MCP for CRE data The CRE AI Institute

Build your first agent

The CRE AI Institute, by PSV, walks you through every layer on a real practice deal, from standing instructions to the review gate. 7-day free trial, $0 today.

Explore the Institute Have it built for your firm

AI agents for commercial real estate, explained

A chat answers. An agent owns a step.

A chat answers questions

You ask, it answers, and the context ends with the conversation. Useful, but every deal starts from zero.

An agent owns a pipeline step

It carries one step of your pipeline with standing rules, access to the real files, and an output format your team reviews.

Five layers of a working agent

The model

Claude by Anthropic, the reasoning underneath. Well-suited to long deal documents and detailed, multi-step instructions.

The instructions

Your rules, written down once: what to screen for, which assumptions to apply, what the output must look like.

The knowledge

Your deal files. The OM, T-12, and rent roll the agent works from, so answers come from your documents, not general knowledge.

The tools

What the agent may touch: files, spreadsheets, data connections. The scope you grant defines what it can and cannot do.

The review gate

A person signs off. The agent drafts, cites its sources, and stops. Your team approves what moves forward.

The model is the least of these decisions; the reasoning behind PSV's choice is in Claude vs ChatGPT for CRE. The layers that make or break the agent are yours: the rules, the files, and the gate.

This stack is exactly how the CRE AI Institute teaches members to build their first agent, on a real practice deal, a multifamily asset called Desert Grove.

Own the agent, keep the trail

Most agent products rent you a black box: your files go into someone else's system and the reasoning stays hidden. An agent you own inverts that.

Runs on your files, in your environment

Every number traceable to the document it came from

A citation trail you keep, not one that expires with a subscription

That is the PSV thesis: cited truth as a system property, not a feature. Operators can build this themselves. Firms that want it built and deployed across the team work with PSV.

AI agents for CRE, answered

What is an AI agent for commercial real estate?

An AI agent for commercial real estate is a model like Claude given standing instructions, access to your deal files, and defined tools, so it can carry a full workflow such as screening a deal or abstracting a T-12. Unlike a chat, it follows your rules every time and returns output in a format your team reviews.

How is an AI agent different from a chatbot?

A chatbot answers the question in front of it and forgets the context when the conversation ends. An agent owns a step of your pipeline: it starts from your rules, works from the real files, and produces the same reviewable output every run.

How do I build an AI agent for real estate?

Pick one defined workflow, write down the rules you already apply, give Claude those instructions plus the relevant deal files, and require a source citation behind every number. Keep a person signing off on the output. The CRE AI Institute teaches this build step by step on a real practice deal.

Do I need to code to build an AI agent?

No. The instructions that define an agent are written in plain English: what to check, which assumptions to apply, what the output must look like. Code becomes relevant when a firm wants agents wired into its systems, which is the kind of build PSV does for enterprise clients.

Should I buy an AI agent product or build my own?

Buying is really renting: your files flow through someone else's system and the reasoning stays hidden. Building means the agent runs on your files, in your environment, with a citation trail you keep. Start by learning how one works; the build is smaller than most operators expect.

What can an AI agent do for a CRE team?

Well-defined document workflows suit agents best: screening incoming deals against your criteria, abstracting a T-12 or rent roll into your template, and drafting an IC summary with sources attached. A person reviews every output before it moves forward.

Keep reading