
Pre-Development Risk: The Checklist AI Can Run for Developers
The pre-development risk checklist AI can run for developers: entitlement, constructibility, market, cost and financing, site and environmental, and timeline carry, held all at once so no category is skipped. What actually kills deals before construction, why completeness is the feature, and how to turn the checklist into a go, kill, or renegotiate. Verified Atlanta Fed and NAHB figures, developer ICP.
Pre-Development Risk: The Checklist AI Can Run for Developers
The Pre-Development Risk Checklist, Answered First
Pre-development risk is everything that can kill a development deal between site control and construction start, and the checklist is the fixed set of checks that catches it while you can still act. AI runs the checklist by holding every risk category at once and flagging where a site fails or where data is missing, so no category gets skipped under a deadline. The categories are consistent across deals: entitlement and approval risk, constructibility, market, cost and financing, site and environmental, and timeline carry. What AI automates is the completeness, running every check every time and surfacing the gaps. What it does not do is clear a flag. A human decides whether each risk is fatal, workable, or priceable. The output is a ranked risk picture that turns a vague bad feeling into a specific, addressable list before you commit.
This checklist supports the whole feasibility decision the hub describes. The full pipeline is in AI feasibility analysis for real estate development.
What Actually Kills Deals Before Construction
The pre-construction phase is where development risk concentrates, and the data backs the intuition. A Federal Reserve Bank of Atlanta study of multifamily projects found 15.3 months on average from project announcement to construction start, more time than the construction itself (Atlanta Fed). That is over a year of exposure before a single unit generates revenue, and most of what goes wrong in a development goes wrong in that window. The cost of the approval process is part of why: regulation runs an average of 40.6% of multifamily development cost per a joint NAHB and NMHC study (NAHB and NMHC), and a separate NAHB study put it at 23.8% of a new single-family home's price (NAHB).
"The deals that die before construction almost never die of one big obvious thing. They die of the small thing nobody bothered to put on a list, on the one site where it mattered." Lucas Eschapasse, NextAutomation
The pattern in failed deals is not exotic risk, it is missed routine risk. A checklist exists because human attention is uneven under pressure, and the value of running it with a system is that the twentieth site gets the same complete pass as the first.
The Checklist, by Category
The list below is the spine. Each category is a cluster of specific checks; the table names the category and the question it answers, and each links to the deeper build where one exists.
| Risk category | The question it answers |
|---|---|
| Entitlement and approval | Will it get approved, and how long and how uncertain is the path |
| Constructibility | Does the envelope actually support the program you priced |
| Market | Will it lease or sell at the pace and price the model assumes |
| Cost and financing | Can it be built and funded at the assumed basis |
| Site and environmental | Can the physical site deliver, from access to contamination |
| Timeline and carry | What does a slip cost, and can you carry the tail |
Two categories carry the most weight and have their own deeper builds. Entitlement and approval is priced in entitlement risk modeling with AI, and the cost and timeline carry sensitivity lives in the AI development pro forma. The checklist is what ensures the other four categories get the same rigor, rather than being waved through because the first two looked fine.
Completeness Is the Feature
The point of running a risk checklist through a system is not that a machine is smarter about any single risk. It is that a machine does not get tired, does not run out of Friday, and does not decide that this site is probably fine because the last three were. Completeness is the feature. A flagged risk you decide to accept is a managed risk; a risk you never checked is the one that shows up as a surprise in month four, when your option period is gone and your capital is committed.
The forward-looking version of this discipline has a live proof surface. The architecture that watches approval status across a developer's jurisdictions and traces a change to a schedule impact is our permit tracking and entitlement monitoring system, which runs the same "catch the flag, route it to a human" pattern the checklist runs, on permits already in motion.
Turn the Checklist into a Go, Kill, or Renegotiate
A risk checklist is only useful if it drives a decision, and the three honest outcomes are go, kill, or renegotiate. A clean pass with priceable risks is a go. A fatal flag, an easement through the buildable area, an approval the jurisdiction will not grant, is a kill, and finding it early is the checklist paying for itself. The most valuable outcome is often the third: a real but addressable risk that becomes a price adjustment or a longer option period. A checklist that only ever says yes or no is missing the outcome that makes most deals better.
Run early, the checklist protects your capital and your calendar both, which in development are the same thing. How the whole pre-development stack gets built and deployed on your terms is in our developer implementation work.
Frequently Asked Questions
What is a pre-development risk checklist?
It is the fixed set of checks that catches everything that can kill a development deal between site control and construction start: entitlement and approval risk, constructibility, market, cost and financing, site and environmental, and timeline carry. Run through AI, every category gets flagged for failures or missing data so none is skipped under a deadline. A human still clears each flag; the output is a ranked, specific risk picture rather than a vague bad feeling.
What actually kills development deals before construction?
Missed routine risk, not exotic risk. Most of what goes wrong in a development goes wrong in the pre-construction window, which a Federal Reserve Bank of Atlanta study found averages 15.3 months from announcement to construction start, longer than the build itself. Deals rarely die of one big obvious thing; they die of the small check nobody put on a list, on the one site where it mattered.
Why run the checklist through a system instead of by hand?
Because completeness is the feature. A system does not get tired, does not run out of Friday, and does not assume this site is fine because the last three were, so the twentieth site gets the same complete pass as the first. A flagged risk you decide to accept is managed; a risk you never checked is the surprise that shows up in month four when your option period is gone and your capital is committed.
Does the checklist just give a yes or no?
It should drive three outcomes, not two: go, kill, or renegotiate. A clean pass with priceable risks is a go. A fatal flag is a kill, and finding it early is the checklist paying for itself. The most valuable outcome is often the third, a real but addressable risk that becomes a price adjustment or a longer option period. A checklist that only says yes or no misses the outcome that makes most deals better.
Which risk categories carry the most weight?
Entitlement and approval risk and the cost-and-timeline carry, because that is where development exposure concentrates. Regulation runs an average of 40.6% of multifamily development cost per a joint NAHB and NMHC study, and 23.8% of a new single-family home's price per a separate NAHB study, and the approval calendar drives the carry. Both have deeper builds, but the checklist exists to ensure the other four categories get the same rigor rather than being waved through.
Related Articles
AI Agents vs Chatbots for Real Estate: Why the Difference Decides Your Result
A chatbot answers questions from what you paste in; an AI agent reaches your CRM, data room, and reporting stack and takes multi-step action on your real deals. This guide draws the line plainly for real estate investors and developers, shows where a chatbot still wins, and explains why so many AI pilots stall: they were chatbots that never got connected to the firm's systems. Decide by naming your bottleneck.
Connecting an AI Agent to Your Real Estate CRM with MCP
To connect an AI agent to your real estate CRM, you put an MCP server on it, exposing read, list, and gated update actions the agent calls through the open Model Context Protocol. The agent then reaches the full deal history, pipeline stage, and contacts that a chatbot could only see if you pasted them in by hand. A practitioner guide to how the connection works, why read-only comes first, and why this is where stalled pilots turn into daily tools.
How to Connect AI Agents to Your Real Estate Data Securely
Connecting AI agents to your real estate data safely comes down to three deliberate choices: default to read-only, scope each MCP server to a single system, and run everything on your own infrastructure. A practitioner-level guide for principals and CTOs on read-versus-write access, per-server scope, the governance checkpoints that make writes safe, and why an agent that reads widely and writes nothing is the right place to start.
