If you already pay for ChatGPT or Claude, the obvious question is why you would need anything else to check your landlord's bill. Sometimes you do not. This article is about telling those cases apart.
We use general assistants every day, and we wrote a full guide on auditing CAM charges with ChatGPT or Claude. We also built Tenantback, because the same assistants kept running out at the same five points. Here is where those points are.
- One lease, a few hours, and you will send the letter yourself: a general assistant with good prompts is enough.
- Several locations, a close deadline, or money you cannot afford to get wrong: use a dedicated agent.
- You do not have to pick one. Tenantback works inside ChatGPT and Claude as a connector.
What ChatGPT and Claude do well
Credit where it is due. A modern assistant reads a sixty page lease faster than any person, finds the exclusions article, explains a gross-up clause in plain English, and spots a line like "parking lot resurfacing" that does not belong in operating expenses. For understanding a lease, it is excellent.
Where a general chatbot runs out
- Arithmetic. Overcharges are chains of percentages on specific bases. Assistants pick the right numbers but slip on the calculation often enough that you have to redo it.
- Exact quotes. A paraphrase that sounds like your lease is not your lease. In a dispute, a clause that is not on the page costs you credibility on every other line.
- Memory and deadlines. The objection window is the whole game, and a chat does not remind you on day 150 of 180.
- Scale. One lease per chat is manageable. Forty leases, forty landlords and forty windows a year is a job, not a conversation.
- The landlord. The money comes back after the letter is sent, the pushback is answered and the credit is confirmed. An assistant can draft; it does not follow up.
Side by side
| Task | ChatGPT or Claude alone | Tenantback |
|---|---|---|
| Read the lease | Very good | Very good; amendments override the terms they change |
| Quote the clause | Must be checked by hand | Verified against the source page; unverifiable quotes are never sent |
| Compute the overcharge | Must be redone in a spreadsheet | Computed by code as a waterfall, so findings never double count |
| Track the deadline | You do it | Dated task, reminders and a calendar feed |
| Write the letter | Good draft | Letter and schedule drafted for your approval |
| Deal with the landlord | You do it | Reads replies, answers pushback, escalates anything unusual |
| Many locations | One chat at a time | Every statement, every location, as it arrives |
| Cost | Your subscription and your time | Free audit; 25% of what is recovered, nothing otherwise |
You do not have to choose
Tenantback is also a connector for the assistant you already use. In ChatGPT or Claude, add a custom connector (MCP) with this address, no key needed:
https://tenantback.com/mcp
Then attach the lease and the statement and ask, "Is my landlord overcharging me on CAM?" Your assistant does the conversation; Tenantback does the audit, with verified quotes and computed numbers, and returns the findings into the chat. If you would rather not set anything up, forward the two PDFs to audit@tenantback.com and the report comes back by email.
Questions
Can ChatGPT audit a CAM reconciliation on its own?
It can read the lease and flag likely issues well. Check every quote against the PDF and redo the arithmetic in a spreadsheet before you send anything, and track the deadline yourself.
Does Tenantback use ChatGPT or Claude?
Tenantback uses language models to read documents, but the dollar amounts are computed by code and every lease quote is checked against the source page. That combination is what makes the findings safe to send.
How do I connect Tenantback to ChatGPT or Claude?
Add a custom connector (MCP) with the address https://tenantback.com/mcp. No key is needed for audits. Then attach the lease and the reconciliation and ask your assistant to audit them.
