Skip to main content

CRE acquisitions

The AI Acquisition Infrastructure

One lead agent and five specialists that carry a parcel from the entity on the deed to a buy-box verdict, a first-pass underwrite, an IC memo and the letter. It names the owner where you supply the registry filings, and hands you the exact plan to name one where you do not.

Get the free packBook a call

The same pipeline we build for acquisition teams, written out so you can install it yourself.

See open times
Get the free packBook a call

Five stages, one system

Most acquisition stacks are five disconnected tools and an analyst copying between them. This is the same work as one pipeline.

  1. 01

    Resolve the owner

    The parcel arrives as an entity on a deed. Supply the registry filings and it works the chain until two records agree on one person. Supply the CSV alone and it says so plainly: zero humans resolved, and here is the registry, the filing and the field pair that would name each one.

  2. 02

    Screen against the buy box

    Every parcel gets a verdict: KILL, WATCH, PURSUE or PRIORITY, with a fit score and the reason written on the row. Nothing leaves the pipeline without a counted reason.

  3. 03

    Underwrite the first pass

    A first-pass model shell where every line is either sourced or tagged as an input to collect, never guessed. It also ranks the documents you could hand it by how much each one would fill.

  4. 04

    Write the IC memo

    Eight sections where every fact carries the record it came from or its collect tag, plus a sources block, an open-items list, and an honest-ceiling section that states what the memo does not cover.

  5. 05

    Draft the outreach

    A letter, an email and a call script that could only be about that parcel, built from what the pipeline established and nothing else. No dollar figure in a first touch.

One lead agent runs all five and owns the handoffs, so the screen, the underwrite and the memo read the same buy box and the same row. Two gates stay human on purpose: one before the memo goes to IC, one before anything leaves your mailbox.

What is inside

One lead agent, five specialists, and everything they read.

Install guide

File tree plus install paths for a Claude Project, Claude Code, ChatGPT and Codex. Ten minutes from download to a running pipeline.

Implementation guide

The architecture end to end: what each stage reads, what it writes, and where the two human gates sit.

The buy box

One JSON file: asset type, size, market, thesis, distress signals. Written once, read by every agent downstream.

Stage contracts

Input and output columns per stage, plus the exclusion count each stage owes you. This is what keeps rows from vanishing between steps.

A fifty-row sample and a memo

A real public-county run at the property level, one IC memo generated from the top row, and the provenance note naming the portal and the pull date.

Outreach templates

Owner letter, email and call script, with the property-specific slots the pipeline fills.

The two human gates

What a person checks before the memo goes to IC, and before anything is sent. Twenty minutes that keep a wrong-owner letter from going out with your name on it.

The lead agent and five specialists

Owner resolver, deal screener, underwriter, memo writer, outreach writer, and the acquisition lead that dispatches them. Paste-and-run, each one.

Worked example

The whole pipeline on the sample county, with the counts at every stage. The number nobody publishes is how many rows survive each filter.

The sample run

Fifty rows you can run today

The pack ships with a real run so you can watch the pipeline work before you have your own data. Here is exactly what those rows are, and what they are not.

What the rows are

  • One public county: Mecklenburg County, North Carolina.
  • Property level: parcel id, street name, asset type, size, year built, assessed value, last sale year.
  • The owner entity exactly as recorded on the deed, plus distress signals, buy-box fit and rank.
  • Every row badged SAMPLE, so no row can be mistaken for live deal flow.

What the rows are not

  • No owner contact data of any kind.
  • No resolved human names. The entity string on the deed is where the free sample stops.
  • No phone numbers, no email addresses, no mailing addresses.
  • Not a client run. Public county records only, never rows from work we did for someone else.

Owner contacts are not in this download. What ships is the method and the public property-level trail. Resolving an entity to a reachable human is a separate engagement with its own verification work, which is why a free zip does not hand you the people.

Want it built and run on your pipeline?

We build it on your data, in your markets, and hand back the working system.

Book a 20-min call

Get the pack

Free, and it installs in the tool you already use. It is a first-pass system, not straight-through processing, and the IC call stays human.

Download the pack

Want the system, not just the files?

Get the full pack once, then use the walkthrough below to see exactly what is inside and how to run it.

Get The AI Acquisition Infrastructure Pack

Enter your details and the full pack downloads right away. No cost.

We will email you occasionally about real estate AI. Unsubscribe anytime.

Have us build it for you

Twenty minutes. Bring your buy box and we will walk the pipeline on your own markets.

Loading calendar

Questions real estate teams ask

What is the AI Acquisition Infrastructure?

It is a free pack that installs one acquisition lead agent and five specialists into Claude. The lead runs the pipeline: the owner resolver works the entity on the deed toward a named human through the public trail and says plainly when the records you supplied cannot get there yet, the screener scores each parcel against your buy box, the underwriter builds a first-pass model shell, the memo writer produces an investment committee memo where every fact carries the record it came from, and the outreach writer drafts the letter. Two human gates sit in the flow, one before the memo goes to IC and one before anything leaves your mailbox.

How do I install and run it?

The README gives install paths for a Claude Project, a Claude Code skill, a ChatGPT project and Codex. Fill in 01-buy-box.json once, paste the acquisition lead, and hand it the sample CSV. Every agent restates your mandate back to you and waits for a confirmation before it touches a row.

What does this pack not do?

It does not close deals and it is not straight-through processing. It is a first-pass system: the underwrite is a first pass on public records, sharper when you supply documents, and from public records alone it returns a shell of named inputs to collect rather than finished numbers. The memo is a draft whose honest-ceiling section states what it does not cover, and the IC call stays human. It also ships no data pipeline. You bring the parcel rows, whether from your own export or from a county portal.

Does the pack include owner contact data?

No. The pack teaches the resolution and gives you the agents that run it, and the sample stops at the property. What ships is the method and public property-level records. Resolving owners to reachable people is a separate engagement with its own verification work, which is why no contact data is in the download.

What is in the fifty-row sample?

Fifty ranked rows from one public county, Mecklenburg County in North Carolina, at the property level: parcel id, street name, asset type, size, year built, assessed value, last sale year, the owner entity exactly as recorded on the deed, the distress signals, the buy-box fit and the rank. Every row carries a sample note. There are no resolved human names, no phone numbers, no email addresses and no mailing addresses in it.

Which markets does it work in?

The three records the method runs on, the county assessor, the state business registry and the county recorder, exist in all fifty states. What varies is depth and digitisation county by county. Some recorders index older deeds without searchable images, and land-use code legends differ per county, so the worked example shows what a clean county looks like and where a thin one costs you time.

How is this different from the Off-Market Sourcing System?

The Off-Market Sourcing System is the front of the funnel: it screens a county and traces owners. This pack is the whole desk. It takes the same public-record trace and carries it through screening, underwriting, the IC memo and the outreach, with typed contracts between the stages so nothing is lost in the handoffs. Start with either one. They install the same way.

Where AI changes the answer

The bottleneck in acquisitions was never the analysis, it was the handoffs. A stage contract between agents means the screen, the underwrite and the memo read the same buy box and the same row, so nothing gets retyped and nothing gets silently dropped.

See how we build this →

Related resources

Want the system, not just the files?

Two ways to go further with NextAutomation.

Work with us

Have us build the sourcing-to-reporting pipeline these files only hint at. Custom AI systems for your firm.

Apply to work with us

Join the community

Live builds, the template library, and a room of operators using AI on real deals. from $49/mo.

Join the community

From free pack to full system

Solution

AI Deal Sourcing for Off-Market Pipeline

How the front of this pipeline runs in production: finding owners, scoring motivation, and routing targets before deals reach the open market.

Learn More
Solution

The AI Underwriting Copilot

Want the underwriting and memo half run on your own pipeline? We build it on your documents and your underwriting standards.

Learn More