Can AI help with underwriting at a real estate development firm?

It depends on which part of underwriting you mean. AI is already useful for the reading and organizing work: summarizing offering memorandums, pulling figures from rent rolls and leases, organizing comps, researching zoning and drafting investment memos. It is weak at the parts that decide the deal: setting assumptions, judging a site and a market, and building the pro forma itself. If your acquisitions team loses hours to documents and data gathering, AI can help quickly. If the bottleneck is judgment, it will not replace your team.

Who this is for

Heads of development, acquisitions leads, analysts and COOs at real estate development firms, and anyone underwriting deals who spends a lot of time reading, extracting and organizing information.

The short version

The best early uses are document-heavy: offering memos, leases, rent rolls, zoning codes and memo drafts.
Every number that goes into a model must be checked against its source.
The pro forma and the key assumptions stay with people.
Searchable past deals and memos are one of the most valuable things AI can give a development firm.

Where AI helps today

Offering memorandum review. An AI assistant can summarize a broker’s offering memo, pull out the stated figures and flag claims that need checking. Analysts start with a structured summary instead of a long PDF.

Rent rolls and leases. AI can extract tenants, rents, lease dates, renewal options and expense terms into a table. This is a good first task because the output is easy to check against the original documents.

Organizing comps. If your team already gathers sales and rent comps, AI can help clean, sort and summarize them, and point out which ones differ most from the subject property. The comps still come from sources your team trusts.

Zoning and entitlement research. AI can help read long zoning codes and planning documents and summarize allowed uses, height and density limits, parking rules and approval steps for a site. Treat this as a research starting point. Codes change and local interpretation matters, so confirm with the source documents and your land use counsel.

Investment memo drafts. From the team’s notes, model outputs and site research, AI can draft sections of an investment committee memo in your firm’s format, such as the site description, market overview and risks. The team rewrites the parts that carry the argument.

Searching past deals. Many firms have years of memos, models and lessons learned spread across drives. A well-organized search assistant lets an analyst ask how a similar site, product type or risk was handled before. This depends on organizing and connecting your documents, not just buying a chat tool.

Where AI is weaker

The pro forma. AI can help write formulas, check a model’s structure or explain outputs. The model and its numbers need to be built and verified by people.
Assumptions. Rent growth, exit cap rates, construction costs and timelines are judgment calls based on your market knowledge. AI can help you list what to consider, not decide.
Market and site judgment. Walking a site, reading a neighborhood and knowing local players remain human work.
Live market data. General AI assistants may not have current, reliable market data. Do not treat a figure from a chat answer as a comp.

Data and confidentiality

Offering memos and seller data often come with confidentiality agreements, and your own models and pipeline are sensitive. Before anyone uploads documents:

Use AI tools on business plans with clear data protections, not personal accounts.
Check whether your confidentiality agreements limit where material can be processed.
Agree what can go into which tool and write it down. See keeping client data safe with AI and how to write an AI acceptable use policy.

How to start

Pick one document-heavy task. Rent roll and lease extraction or offering memo summaries are good first choices.
Test it on a past deal. You already know the answers, so you can judge quality without deal pressure.
Train the acquisitions team on real material. Analysts need to learn to give AI the right documents, ask precise questions and check outputs. See how to train non-technical employees to use AI.
Measure one thing. For example, hours to turn an offering memo into a first underwriting summary, before and after. See how to run an AI pilot that doesn’t stall.
Then decide whether to build. If the same extraction and memo steps repeat on every deal, a custom agent connected to your files and templates may be worth it. See build vs buy AI tools.

Off-the-shelf tools or something custom?

A general AI assistant on a business plan, with good training, can handle offering memo summaries, lease extraction and memo drafts for most teams. Some real estate software providers are adding AI features, so check what your current tools already offer.

Custom work makes sense when you want AI to work across your own systems: searching every past deal, filling your underwriting template from a new offering memo, or applying your firm’s screening criteria the same way each time. See custom AI agents for your business.

Common mistakes

Copying AI-extracted numbers straight into the model. One wrong rent or date flows through every output.
Trusting AI zoning answers without checking the code. Treat them as a lead, not a conclusion.
Uploading seller material to personal accounts. It can breach confidentiality terms and your own policies.
Training only analysts. Senior people set the standard for what is acceptable. See how to get partners to use AI.

When you do not need us

If your firm has an approved AI tool, clear data rules and an analyst keen to build good prompts for underwriting work, you can get real value in-house with the steps above. Outside help is more useful for structured training across the team, or custom agents that work across your past deals and templates.

How Ruby helps

AI training: hands-on sessions for acquisitions and development teams using tasks from past deals.
AI strategy: choosing first uses, setting data rules and running a measured pilot.
AI engineering: building agents that search past deals, extract lease and rent roll data and draft memos in your format.

If underwriting work is stretching your team, book a call with Ruby.