Workflow Guide
Query Your Lease Portfolio with ChatGPT, Claude, and NotebookLM
A 2-step workflow for analyzing a commercial lease portfolio with AI: extract structured data, then load it into a Project/Notebook and ask questions in plain English. Works with LeaseParse output or any other abstraction tool.
The pipeline
The 2-step workflow
Step 1
Extract structured lease data
Convert your lease PDFs into a structured Excel with one row per lease and one column per field (rent, escalator, renewal options, CAM structure, etc.).
Why first: A 100-page lease is roughly 50K–100K tokens. A portfolio of 20 leases exceeds every consumer AI tool's context window. Structured data compresses 100K tokens to ~2K tokens per lease, making portfolio analysis feasible.
Step 2
Load into a Project or Notebook
Upload the Excel (and optionally the source PDFs) into ChatGPT Projects, Claude Projects, or NotebookLM. Set project instructions framing the AI as a CRE analyst.
Why a Project, not a chat: A Project persists your portfolio across sessions, so every question has the same context. Chats reset.
What the extracted data looks like
One row per lease, one column per field. Every value carries a confidence score so you and the AI both know what to spot-check.
lease-portfolio-abstract.xlsx
preview · 4 of 100 rows| Tenant | Suite | Annual rent | Escalator | Renewal option | Confidence |
|---|---|---|---|---|---|
| Mariposa Group | 3A | $48,200 | CPI cap 4% | 2 × 5yr FMV | 0.94 |
| Northgate Partners | 12 | $112,800 | Fixed 3% | 1 × 5yr | 0.88 |
| Halcyon Retail Co. | R-2 | $26,400 | Porter wage | None | 0.72 |
| Ferrum Logistics | IND-7 | $310,000 | Fixed 2.5% | 2 × 10yr | 0.96 |
Illustrative preview. Real output includes 50+ columns and one row per lease.
Step 1: Pick an extraction tool
Any AI lease abstraction tool that outputs Excel will work for this workflow. LeaseParse is the cheapest ($7–$15 per lease, first lease free) and the only one that lets you evaluate output before paying. Other options:
- • LeaseParse — $7–$15/lease, first lease free, 50+ fields. See pricing.
- • Lextract — $12–$15/lease, 126 fields. Compare.
- • LeaseWizard — $20/contract, 100+ fields. Compare.
- • Prophia / MRI Contract Intelligence — enterprise tiers with portfolio analytics built in.
- • Manual — $100–$500/lease, 2–4 weeks turnaround.
See the full ranking or the cheapest options.
Step 2: Pick an AI tool to query the data
Each handles lease portfolios differently. Here is how they compare for this workflow specifically.
Claude Projects
- Context size
- 200K tokens (1M with extended context)
- File support
- Excel, PDF, CSV, DOCX, images, text
Best for: Portfolios up to a few hundred leases. Best at reading messy abstract Excels and writing structured summaries.
Setup: Create a Project in claude.ai, drop the Excel and source PDFs in Knowledge, ask questions in chat.
ChatGPT Projects
- Context size
- 128K–200K tokens (varies by model)
- File support
- Excel, PDF, CSV, DOCX, images
Best for: Same workflow as Claude. Strong for mixing lease analysis with code (running pandas on the Excel inline).
Setup: Create a Project in chatgpt.com, upload Excel and PDFs, set Project Instructions ("you are a CRE analyst…"), ask questions.
NotebookLM
- Context size
- Up to 50 sources, 500K words each
- File support
- PDF, DOCX, Google Docs, web URLs, audio (no Excel/CSV directly)
Best for: Large lease document portfolios. Best at synthesizing across many PDFs with source-grounded citations.
Setup: Create a notebook in notebooklm.google.com, add lease PDFs as sources, optionally add the abstract Excel as a CSV or copy-pasted text.
Click-by-click setup for each tool
Where to click, what to upload, where to paste instructions — for each tool.
Claude Projects — step by step
- 1
Open claude.ai and sign in
Pro plan or higher is recommended for the larger context window.
- 2
Click "Projects" in the left sidebar
Or go directly to claude.ai/projects.
- 3
Click "Create Project"
Top-right of the Projects page.
- 4
Name it
Something like "Q3 Portfolio Review" or "Acquisition X Lease Audit".
- 5
In the Project, find "Project knowledge" on the right side
This is where Claude will read from for every chat in the project.
- 6
Click "Add content" → drop the LeaseParse Excel
Optionally add the source lease PDFs too if you want Claude to cross-reference exact lease text.
- 7
Click "Set custom instructions" and paste the CRE analyst instructions
See the instructions block further down this page.
- 8
Start asking questions in the chat
Every reply uses your Excel as context — the Project persists across sessions.
Where you'll be working
Schematic — illustrative layout, not a screenshot.
ChatGPT Projects — step by step
- 1
Open chatgpt.com and sign in
Plus, Team, or Enterprise plan is needed for Projects.
- 2
Click "+ New project" in the left sidebar
It sits above your chat history.
- 3
Name the project
Same naming convention as Claude.
- 4
Click the project to open it
You'll land on the project home page.
- 5
Click "Add files"
Either at the top of the project or via the paperclip icon in the chat input.
- 6
Drop the LeaseParse Excel
ChatGPT will index it for retrieval across every chat in the project.
- 7
Click "Instructions" to add the CRE analyst framing
Project-level instructions apply to all chats inside.
- 8
Ask questions — ChatGPT can also run pandas on the Excel inline
For NOI roll-ups, escalation modeling, charts.
Where you'll be working
Schematic — illustrative layout, not a screenshot.
NotebookLM — step by step
- 1
Open notebooklm.google.com and sign in with Google
Free tier works; paid Plus unlocks higher source counts and longer outputs.
- 2
Click "Create new notebook"
Top-left "+" button.
- 3
Upload sources
Drop your lease PDFs into the Sources panel. Up to 50 sources, 500K words each.
- 4
For the Excel: paste the contents as a text source
NotebookLM does not consume Excel directly. Open the Excel, copy the table, and paste into "Paste text" source.
- 5
Wait for sources to index
NotebookLM extracts and embeds — usually 10–60 seconds per source.
- 6
Ask questions in the chat
Every answer is grounded with citations back to the exact source and page.
- 7
Optional: generate Audio Overview
For commute-friendly portfolio briefings.
Where you'll be working
Schematic — illustrative layout, not a screenshot.
What a query looks like in practice
A typical prompt-and-response inside a Claude Project with the lease Excel loaded as Knowledge.
Claude Project · Q3 Portfolio Review
Knowledge: lease-portfolio-abstract.xlsxFound 12 leases expiring within 18 months across the portfolio. Three have notice deadlines in the next 90 days.
| Tenant | Expires | Notice by | Renewal |
|---|---|---|---|
| Halcyon Retail Co. | 2026-09-30 | 2026-06-30 | None |
| Mariposa Group | 2026-12-31 | 2026-09-30 | 2 × 5yr FMV |
| Northgate Partners | 2027-06-30 | 2027-03-31 | 1 × 5yr |
+ 9 more rows. Halcyon Retail flagged [unverified] — escalator field confidence 0.72.
Illustrative — what the workflow produces inside a Project.
Example prompts that actually work
Paste any of these into a Project or Notebook with your lease Excel loaded. They are designed to use the structured fields (not the raw PDFs) so they run reliably.
Renewal exposure
List every lease expiring in the next 18 months. For each, give me: tenant, suite, base rent, escalator clause, renewal option terms (FMV vs fixed), and notice deadline. Format as a table sorted by expiration.
Escalation patterns
Group the leases by escalation type (fixed step, CPI-linked, porter wage, pass-through). For each group, show the count, average annual increase, and weighted-average remaining term. Flag any leases where the escalator is unusual for the asset class.
CAM / OpEx risk
For office leases only: identify base-year clauses without gross-up language and pass-through structures without an audit-rights clause. Cite the exact lease section for each finding.
Co-tenancy exposure (retail)
For retail leases, summarize every co-tenancy clause. For each, tell me which named anchor or category triggers it, what the remedy is (rent reduction, termination, fixed rent), and the sales-drop threshold required. Flag the riskiest five.
Tenant credit and concentration
Calculate concentration metrics from the rent roll: top-5 tenants by annual base rent, top-5 by total occupied SF, and any tenant where annual rent exceeds 10% of portfolio NOI. Flag any month where lease expiration concentration exceeds 15% of portfolio rent.
Acquisition due diligence
Compare the abstract to the operating statement: do contractual rents match the rent roll? Are there concession amounts in the leases not reflected in the seller's NOI? List every discrepancy with the lease section and rent roll line item.
Audit triggers
Identify leases where CAM charges, OpEx pass-throughs, or property tax allocations look anomalous vs the cohort (asset class + market). For each anomaly, suggest the specific audit question to ask the landlord.
Standard-form benchmarking
For the office leases, compare each one against a typical institutional Class A form: TI allowance, free rent, base year structure, OpEx caps, assignment/subletting consent standard, holdover penalty. Show me where each lease is more or less favorable than market.
Project instructions that improve every answer
Set these once as Project Instructions (Claude/ChatGPT) or as the first message in NotebookLM. They constrain the AI to behave like a CRE analyst rather than a generalist.
1. Always cite the source row (tenant + suite, or lease ID).
2. If a field has a confidence score below 0.8, prefix the finding with "[unverified]".
3. Show calculations explicitly when they involve money or dates.
4. If a question cannot be answered from the abstract alone, say so and tell me which document to consult.
5. Use industry-standard CRE terminology (NNN, OpEx, TI, FMV, CPI, gross-up, etc.).
FAQ
Can ChatGPT or Claude analyze my lease portfolio?
Yes — through Projects. The pattern is extract once into Excel, upload to a Project (Claude or ChatGPT) or NotebookLM, then ask portfolio questions. Direct PDF upload doesn't scale past a handful of leases.
Why extract data first instead of uploading PDFs?
A 100-page lease is roughly 50K–100K tokens. A 20-lease portfolio exceeds every consumer AI tool's context window. Extraction compresses that to ~2K tokens per lease and gives the AI a uniform schema to query against.
Which AI tool is best for this?
Claude Projects or ChatGPT Projects for portfolio analysis on the Excel. NotebookLM for synthesizing across many source PDFs with grounded citations. Many analysts use both.
How accurate is this approach?
Output quality is bounded by extraction accuracy (90–97% on standard fields). Always include confidence scores in your Excel and instruct the AI to flag low-confidence findings.
How much does the whole workflow cost?
Extraction: $7–$25 per lease ($7–$15 with LeaseParse, first one free). AI tool: $20/month for Claude or ChatGPT Plus, free for NotebookLM. 100 leases costs $700–$2,500 plus $20/month for the AI tool.
Start with one free lease
Extract one lease free, drop the Excel into a Claude or ChatGPT Project, and see if AI-driven portfolio analysis works for your team before scaling up.
Upload a lease free