Getting Lease Data Out of the Contract and Into Compliance

ASC 842 and IFRS 16 turned lease accounting into a data problem. Suddenly you needed structured terms — commencement date, payment schedule, discount rate inputs, renewal and termination options, whether that option is reasonably certain to be exercised — pulled out of contracts that were written by lawyers, not accountants. And leases hide everywhere: real estate, vehicles, forklifts, copiers, embedded leases inside service agreements.

The original compliance scramble was mostly humans reading PDFs and typing terms into a spreadsheet. Anyone who lived through adoption remembers it. What almost nobody built was a repeatable way to do it — so every new lease, every modification, every acquisition drags a person back into the same manual read.

What extraction actually does

IDP reads the document; NLP understands the legal language around the terms. The system finds the commencement date even when it's phrased three different ways, pulls the payment schedule including escalations and free-rent periods, identifies renewal and purchase options, and extracts the language you need to assess lease classification. It hands you a structured record instead of a 40-page PDF.

The value is partly time and partly coverage. Extracting terms from a lease portfolio manually is slow enough that things get missed — and a missed embedded lease or an overlooked renewal option is exactly the kind of thing that shows up as an audit adjustment. Consistent extraction across the whole portfolio closes those gaps. It also makes modifications and reassessments far less painful, because the baseline data already lives in structured form.

Where judgment stays human

Here's the important boundary. Extraction pulls what the contract says. It does not make the accounting judgment. Whether a renewal option is "reasonably certain" to be exercised is a call that depends on economics and business intent, not just contract language — the model can surface the option and flag it for assessment, but a person decides. Same with the discount rate: the model can extract any stated rate, but determining the incremental borrowing rate is an accounting judgment.

Legal language is genuinely hard, and lease contracts are inconsistent. A term buried in an amendment, defined by reference to a schedule in an appendix, phrased in a way the drafter thought was clever — these trip up extraction. So we keep a review step, especially for high-value leases and anything with unusual terms, and we treat the extracted data as a strong draft to be confirmed, not gospel.

Modifications are where I'd focus attention. A lease that's amended mid-term needs its terms re-extracted and the accounting reassessed, and it's easy for a modification to slip through if there's no process to catch it. Wire the extraction into your lease workflow so new documents and amendments flow through it automatically rather than waiting for someone to remember.

Done right, this turns lease accounting from a periodic fire drill into something closer to a maintained dataset — which is what the standards quietly assumed you'd have all along.