AI for Commercial Real Estate Due Diligence: Build the Exception Desk

Due diligence fails quietly when the team does not know what it has, what it is missing, which version is current, or which document controls. AI can create a register of every uploaded file, classify it by diligence workstream, extract its effective date, and identify likely duplicates or superseded versions.

The register should be compared against the deal-specific request list. A blank cell is not a negative finding; it is a missing record. That distinction must remain visible from first upload through final committee approval.

• Legal: leases, amendments, title, survey, zoning, contracts, and notices.
• Financial: rent roll, T-12, general ledger support, tax bills, insurance, utilities, and capital history.
• Physical and environmental: property-condition materials, Phase I records, permits, plans, and maintenance history.
• Market and tenant: comps, tenant financial information, public filings, and concentration analysis.

A useful diligence system does not produce one long summary. It creates issue records with a source, description, severity, financial or operating implication, responsible owner, next action, and status. The team can then sort unresolved issues by decision impact.

AI can propose the issue and the supporting excerpt. Counsel, environmental consultants, engineers, accountants, and deal team members remain responsible for professional conclusions within their disciplines.

The highest-value diligence work often sits between files: lease rent versus the rent roll, rent-roll area versus the survey or stacking plan, T-12 expenses versus invoices, tax assumptions versus current bills, and the OM's claims versus the underlying record.

A reconciliation should show both reported values, the calculated difference, and the source for each. The system can suggest a likely explanation, but it should label that explanation as a hypothesis until a responsible person confirms it.

Environmental due diligence and other specialist work have formal standards and professional requirements. AI can organize records and surface questions; it does not replace the qualified professionals or the required process.

The same principle applies to legal interpretation, engineering judgments, and tax conclusions. The diligence platform should make those handoffs explicit and keep the professional's final report attached to the issue it resolves.