# AI in Commercial Leasing: What It Can and Can't Replace Blog | LeasePilot [Blog](/blog)Technology # AI in Commercial Leasing: What It Can and Can't Replace A balanced, credible take on where AI adds value in lease drafting and where human expertise remains essential. ![David Saltman](/_next/image?url=%2Fleadership%2Fdavid-saltman.jpg&w=3840&q=75&dpl=dpl_ByYGQUFkKUCm18Txx9yVAZK96Ktr) David Saltman CEO, Former CRE Attorney November 20, 20248 min readCopy link TL;DR AI excels at extraction and pattern recognition. But negotiation strategy, clause selection, and risk assessment require judgment that can't be delegated. Here's where the line is. § 01 ## [The Hype vs. The Reality](#the-hype-vs-the-reality) Every week brings new headlines about AI transforming legal practice. ChatGPT drafting contracts. Claude analyzing documents. The implication: attorneys are about to be replaced. The reality is more nuanced, and more useful. § 02 ## [Where AI Adds Clear Value](#where-ai-adds-clear-value) ### Document Extraction and Abstraction AI excels at reading existing documents and extracting structured data: - Identifying key terms across hundreds of leases - Flagging deviations from standard language - Building rent rolls from executed documents This is pattern matching at scale, exactly what AI does well. ### Clause Comparison and Analysis Given a corpus of your own executed leases, AI can: - Identify which clauses deviate from your current standards - Surface historical precedent for specific language - Compare negotiated positions across deals ### First-Draft Generation. With Caveats AI can produce initial draft language based on prompts. But there's a fundamental limitation: AI is probabilistic. It generates the most likely next word. Lease drafting requires deterministic outputs, rent calculations, escalation schedules, option dates, and cross-references that must be exactly right, every time. Additional concerns: - The output requires verification by someone who knows what "correct" looks like - Hallucinations (confidently wrong content) are common in legal text - The training data may include outdated or incorrect precedent A first draft that's "mostly right" still requires an attorney to find everything that's wrong, and finding errors in someone else's work is often harder than drafting from scratch. § 03 ## [Where Human Expertise Remains Essential](#where-human-expertise-remains-essential) ### Negotiation Strategy "Should we push back on this tenant's request for contraction rights?" That decision depends on: - Market conditions - Tenant creditworthiness - Portfolio strategy - Relationship dynamics - The landlord's risk tolerance No AI model has access to these factors. And even if it did, the judgment is inherently human. ### Risk Assessment "What's the exposure if this co-tenancy clause triggers?" Understanding the cascading effects of lease provisions requires: - Experience with how similar provisions played out in disputes - Knowledge of the specific property's tenant mix - Business context the AI doesn't have ### Clause Selection for Complex Deals "Which rent escalation structure is appropriate here?" The right answer depends on: - Tenant negotiating leverage - Landlord cash flow preferences - Market comparables - Tax and accounting implications This isn't pattern matching, it's professional judgment informed by years of practice. § 04 ## [The Augmentation Model](#the-augmentation-model) The productive frame isn't "AI vs. attorneys." It's "AI augmenting attorneys." **The attorney retains:** - Final decision authority - Accountability for the output - The relationship with the client - The strategic judgment **Where systems add value:** - Deterministic calculations that must be right every time (rent schedules, escalations, pro-rata shares) - Enforcing your standards and fallback positions consistently across every lease - Propagating deal terms so data entry happens once, not dozens of times - Quality control that catches cross-reference errors and inconsistencies before they reach execution The distinction matters: general-purpose AI tools require you to build the CRE intelligence yourself. A system built around your lease forms and your deal logic encodes that intelligence once, then applies it consistently. § 05 ## [A Note on Trust](#a-note-on-trust) Senior legal leaders are right to be skeptical of AI hype. The stakes in commercial leasing are too high for "usually accurate." Lease calculations, rent escalations, TI amortization, operating expense reconciliation, must be deterministic. They must produce the same correct answer every time, not a probabilistically likely answer. This is the line between AI-generated drafts that require extensive verification and system-generated outputs built from your own validated logic. The responsible approach: lease-specific systems for the deterministic work, attorney judgment for everything that requires it, and healthy skepticism toward any tool that asks you to trust probabilistic output on provisions worth millions over the life of a lease. * * * The future of commercial leasing isn't AI replacing attorneys. It's attorneys freed from mechanical work to do the strategic, judgment-intensive work that justifies their expertise, with systems handling the calculations and consistency that should never depend on human attention to detail. § Adjacent reading ## More from the ledger [§ 01JUN 06, 2024 Technology ### Why We Chose Automation Over AI Generation, 10 Years Ago, and Still Today Lior Kedmi7 MIN READ Read →](/blog/why-we-chose-automation-over-ai-generation) [§ 02MAR 13, 2026 Technology ### Automation vs. AI for Lease Drafting: A Practitioner's Guide LeasePilot Team8 MIN READ Read →](/blog/automation-vs-ai-lease-drafting-guide) [§ 03FEB 06, 2025 Technology ### Why Generic Document Automation Failed CRE Legal Teams David Saltman7 MIN READ Read →](/blog/why-generic-document-automation-failed-cre) § See it in practice ## Reading about it is one thing. Watching it happen is another. See LeasePilot draft a lease in your team’s own templates, with your clauses and your defaults. [Schedule a Demo](/demo)