Lease accounting is the use case nobody gets excited about and everybody benefits from. It's repetitive, it's policy-bound, it's audit-sensitive, and it eats a disproportionate amount of senior finance time at every period end.
In other words: it's perfect.
The shape of the problem
Under IFRS 16 and ASC 842, every lease has to be classified, measured, remeasured, and disclosed. The mechanics are well-defined. The pain is in the inputs: dozens of PDFs, scanned addenda, side letters, and an embedded option here or there that changes the entire schedule.
The work is structured enough that a junior accountant can be trained on it in a fortnight. It's also boring enough that the same junior accountant will leave within a year, taking the institutional memory with them.
Where agents fit
A lease agent has a narrow, useful job:
- Read new and amended lease documents
- Extract the data points that drive measurement — commencement date, term, options, payments, incentives, IBR inputs
- Propose a classification with the reasoning written out
- Generate the amortisation and ROU schedules
- Flag modifications that trigger remeasurement
Every output cites the clause it came from. A finance reviewer signs off in the lease system. The agent never posts.
Why this one deploys
Lease accounting has three properties that make it the easiest agentic AI win in the office of the CFO:
- Low-novelty inputs. Lease contracts are variations on a small number of templates.
- High-volume, low-individual-materiality decisions. Errors are caught at aggregate, not per-lease.
- An obvious eval set. Your existing lease population is the test harness — feed the agent last year's portfolio and see if the schedules match.
Most teams we work with see a lease accounting agent produce reviewable output within four weeks, and meaningful capacity within a quarter. That's not because the technology is magic. It's because the problem is unusually well-shaped for it.
If you're looking for the use case that proves the operating model without scaring anyone, this is it.