AI can do the heavy lifting on lease extraction, but the risk flags need a human eye.

The extraction task (rent escalations, renewal dates, TI allowances) is structurally consistent across leases, which favors automation. However, commercial leases vary significantly in structure, defined terms, and clause placement, so the agent must handle real document variance rather than a clean template.

Extracting specific dates and dollar figures has crisp success criteria. But 'flagging negotiation risks' is inherently subjective — what counts as a risk depends on market context, tenant creditworthiness, and portfolio strategy that aren't defined in the documents themselves.

The agent needs access to the actual lease PDFs, which must be provided by the user. Assuming document access, modern document-parsing and LLM tools can handle extraction well. Market benchmarking data for risk context is not automatically available and would require additional integration.

Misreading a renewal option deadline or rent escalation clause on a $4M–$12M lease could cause a missed negotiation window or incorrect cash flow projection with serious financial consequences. Errors here are not easily reversible if acted upon before review.

Pure extraction requires minimal human judgment. But assessing negotiation risk — whether a tenant is likely to renew, whether market rents have shifted, whether a TI package is competitive — requires relationship context and market intuition that AI cannot supply from the documents alone.