Under the hood
AI reads the file. Code runs the math.
After upload, AI reads the files and pulls out lease terms and billed lines. Fixed code runs the math across 20 checks. Each finding points back to the source.
The pipeline, stage by stage
01
AI reads the files
AI reads each lease, statement, and ledger. It pulls out billed lines and lease terms. It then labels each charge. AI does not run the math.
02
Code runs the math
Every cap test, pro-rata check, and gross-up formula runs as fixed code. Same numbers in, same answer out, every time.
03
20 checks run in full
The engine runs all 20 rules: math checks, sorting checks, ledger checks, and advisory checks. Each one either finds an issue or it does not.
04
Findings cite the clause and the line
Each finding carries the lease clause, the statement line, and the math. It is ready for your review.
The hard line
Where AI stops and code takes over
AI is good at reading messy documents. It is not something a partner should trust to do arithmetic a client could dispute in court. So CAMAudit draws a hard line between the two.
What AI does
- Reads uploaded lease and statement files.
- Pulls out billed line items and lease clauses.
- Labels each charge: excluded, capital cost, insurance type, and so on.
- Ties each labeled charge to the clause that governs it.
What AI never does
- Calculate a fee cap, a pro-rata fraction, or a gross-up adjustment.
- Decide a dollar amount for any finding.
- Produce a number that cannot be re-run and checked.
- Replace your review before a report goes out.
Why it matters
Deterministic code, not a model guess
Every management fee cap, pro-rata check, gross-up adjustment, and cap violation runs through the same TypeScript code path every time. Nothing about the dollar figure on a finding comes from a language model. That is what lets you put your name on the number.
Reproducible
Run the same lease and statement twice. The math checks give the same result both times.
Checkable
The inputs, the formula, and the output are all visible. You can check a finding by hand.
Defensible
A number a model guessed at is hard to stand behind. A number code calculated from lease terms is not.
Common questions
Does AI ever calculate the overcharge amount?
No. AI reads and sorts the documents. Fixed code runs a set formula on the lease terms and statement figures to get the dollar figure.
What happens if the AI misreads a clause?
Every finding shows the clause it used. Check it against the source lease page before you sign off. Extraction sets up the audit. It does not replace your review.
Why not let AI do the math too?
A model can guess at a number. It cannot promise the same answer twice, and it cannot show its work the way code can. A landlord or a client will ask how you got the number. Fixed code answers that every time.
Does CAMAudit find every overcharge automatically?
No tool catches everything. CAMAudit runs the same 20 checks on every file and flags what it finds. Your team still reviews each finding before it reaches a client.
Go deeper
The 20 detection rules
What each rule checks, math or classification, and where it cites the lease.
See the rulesClause-to-finding traceability
See what a finished finding looks like: the clause, the line, and the math, side by side.
See traceability