Your AI Said “Buildable.” Show Me the Zoning.

October 9, 2026

By Anna with Oppy

A buyer asks whether the backyard can hold a rental unit. Your AI sees a large lot, finds “accessory structures permitted,” and produces a confident yes.

Lovely sentence. Wrong job.

The useful AI employee does not approve the project. It assembles the evidence that lets the right person answer. A newly reported zoning-data business offers a concrete reason to build that employee now.

The housing number hiding inside the zoning map

In its October 8 coverage of Land Use Labs, RISMedia reported six zoning-data products: parcel reports, a zoning API, district layers, combined base-and-overlay “slices,” summary tables and standardized rasters. The company’s own product page confirms those formats and describes coverage of more than 96 million parcels. That is a provider claim, not an independent coverage audit.

The more revealing number came from founder Sara Bronin. She told RISMedia that, within the company’s mapped residentially zoned land, 94% allows single-family housing by right, while 12% permits multifamily housing by right. She described coverage encompassing areas where 70% of Americans live. These are reported figures for the covered geography, not percentages of every American parcel. The categories can overlap; they are not pieces of a pie. 1

That gap is not a prediction of which backyard can become a duplex. It is a warning against treating “residential” as a complete answer.

Bronin also described why maintenance matters:

“We also realized that some of our earliest completed geographies had grown stale because zoning codes change.”

Sara Bronin, speaking to RISMedia about the creation of Land Use Labs.

The National Zoning Atlas, which Land Use Labs supports, describes a process of extracting upwards of 200 regulatory characteristics for each zoning district. A listing’s two-word description is doing rather less homework.

The opportunity: use agentic AI to prepare a property-specific zoning question, not to manufacture a property-specific permission.

Give one Oppy a smaller, better job

Call it Zoning Homework. Its deliverable is a short research packet for a broker, planning professional or other designated reviewer. Not a feasibility certificate. Not a permit opinion. Not a marketing claim.

Oppy’s setup guide documents goals, personality, a teach section, prompts and webpage references. Use the relevant instruction controls available in your account. This is a proposed configuration, not an announced Oppy zoning product or a claimed native connection to Land Use Labs.

Start manually. Supply information your team is authorized to use. If you later connect a data provider, verify the actual connection and permissions first. Oppy’s homepage names Zapier and Make as connection options; that does not establish a particular zoning API integration.

The first workflow needs only three inputs:

The customer’s exact question. “Can I advertise this property as having potential for an accessory dwelling unit?” is different from “Is a storage shed allowed?”
An identified property and source packet. Include the municipality, parcel identifier, zoning district, mapped overlays, source URL or document locator, retrieval date and any stated code or dataset date. Preserve blanks as blanks.
A named review owner. Decide who checks the interpretation and who can approve customer-facing language. An empty approval field is not a personality setting.

A provider’s parcel report or public map may help organize the research. The current local code, applicable amendments and any relevant official clarification still need examination. Record disagreements instead of asking the model to vote.

Paste this job instruction into your Oppy

Adapt the reviewer name and authorized sources. Do not paste sensitive client information into a system your organization has not approved.

NAME: Zoning Homework JOB: Prepare a source-linked zoning research packet for a human reviewer. Work only from the property identity, customer question and authorized source packet supplied for this case. Do not guess missing zoning rules from neighboring properties, general knowledge or an attractive site plan. For each material statement, record: the exact supporting excerpt; its source URL or document name and section/page; the relevant district and overlay, if supplied; retrieval date and stated effective/update date, if supplied. Never turn "accessory structure" into "accessory dwelling unit." Never turn "permitted use" into "permit issued" or "buildable." Do not infer dimensions, calculate a buildable envelope or resolve conflicting legal text. Describe what requires review instead. If a parcel identity, overlay, amendment, date or material rule is missing, mark it NOT SUPPLIED. Say whether the gap blocks the customer's question. Return only: QUESTION: the customer's exact question. WHAT THE SOURCES SAY: at most five excerpt-backed facts. WHAT THEY DO NOT ESTABLISH: the unresolved conditions. REVIEW QUESTIONS: specific questions for the designated reviewer. CLIENT DRAFT: at most 80 words, plainly labeled DRAFT FOR REVIEW.

Do not send messages, buy reports, submit applications, alter listings, or promise approval. Your output is research assistance, not a legal, engineering, lending, insurance or permitting determination.

The prompt defines a job. It does not enforce permissions. Configure actual tools and sending rights accordingly, and keep customer messages in a real review process.

A fictional test that catches the expensive mistake

Use a deliberately incomplete case before touching a live listing. Every property detail and rule below is invented for testing.

Property: 18 Cedar Court, Example Town Parcel: TEST-018 District: R-1 Overlay: NOT SUPPLIED Source: fictional training excerpt, section 4.2 Retrieval date: October 9, 2026 Effective date: NOT SUPPLIED Excerpt: "Detached accessory structures are permitted subject to applicable dimensional requirements." Customer question: "Can I market the backyard as buildable for a rental accessory dwelling unit?"

Reviewer: Brokerage compliance lead, who will obtain local planning review

A passing answer identifies the narrow supported fact: the excerpt discusses detached accessory structures. It then flags that the supplied material does not establish a dwelling use, rental permission, overlay restrictions or current applicability.

A useful client draft might read:

Draft for review: The excerpt provided addresses accessory structures, but does not establish permission for a rental dwelling. We need the current dwelling-use rules and parcel-specific restrictions reviewed before describing the backyard that way.

That is not timid. It is accurate.

Now add an undated amendment that conflicts with the excerpt. The Oppy should expose the conflict, not choose the friendlier rule. Remove the parcel identifier. It should flag the identity gap, not silently attach a nearby property’s zoning.

These tests measure whether the agent respects the question’s boundary. Beautiful prose is a separate test, and a cheaper one.

Make the output useful beyond the brokerage

A mortgage team could use the packet to identify a zoning question for its underwriter, without declaring proposed rental income eligible. A property manager could prepare questions about an owner’s intended use, without promising lawful occupancy. A title team could identify which zoning questions need routing, without treating zoning research as title clearance.

Those are proposed work assignments, not demonstrated Oppy integrations or promised industry outcomes. Each business retains its own professional review and authorization requirements.

There is a larger operating-model point here. In BCG’s 2026 research on agentic AI value, based on a survey of more than 1,300 executives and senior leaders across over 20 sectors, the firm emphasizes a few high-value workflows, defined outcomes and controls. That is enterprise context, not evidence that this zoning workflow saves a particular brokerage money.

For your pilot, measure the narrower result: how many submitted packets a reviewer can use without requesting missing sources or correcting unsupported claims. Record preparation time and reviewer rework. Do not count a confident answer as a completed job.

The future is not an AI permission slip

Better structured data can make zoning research more accessible. It cannot make two different land uses mean the same thing.

Build the Oppy that preserves the distinction. Let it collect the question, organize the evidence and prepare the handoff. Keep interpretation and commitments with the authorized people.

“Buildable” is a conclusion. Make your AI show its homework first.