Published:

Last updated:

AI-powered lease accounting software: what to require, and what to ignore

Where AI belongs in lease accounting tooling and where it does not, plus the control evidence your auditor will expect from packaged software.

Emma Sukenik9 min read
Lease accounting — AI-powered lease accounting software: what to require, and what to ignore

"AI-powered lease accounting software" generally means one of three things: software that reads lease documents and extracts terms, software that calculates the ASC 842 or IFRS 16 numbers, or software that watches the resulting data and flags what looks wrong. Those are different jobs with different tolerances for uncertainty, and lumping them under one label is what makes the category hard to evaluate.

The three layers, and what to require at each

LayerWhat machine reading plausibly doesWhat to require as evidence
Getting terms out of documentsExtracts dates, options, payments, and escalationsPage-level provenance, confidence indicators, a review workflow for low-confidence output
The measurement engineNothing probabilistic. This is defined arithmeticReproducibility: identical inputs always produce identical output
Monitoring and anomaly detectionFlags a lease whose terms suggest a missed triggerFlags presented as suggestions requiring human disposition, never automatic postings
Drafting disclosuresAssembles tables and drafts narrative from dataEvery figure tied back to source data, human review before filing

Why reproducibility is the requirement, not accuracy

An auditor asking why the right-of-use asset moved needs a chain: this lease, this modification, this discount rate, this date, therefore this number. The practical test at demo stage is to run it, change one input, run it again, and ask the system to show its working, then ask what happens when the same lease is reprocessed after a software update.

What your auditor will actually ask about

PCAOB Auditing Standard 2201 lists among the risk factors for a control "whether the control relies on performance by an individual or is automated (i.e., an automated control would generally be expected to be lower risk if relevant information technology general controls are effective)." Automation earns the credit conditionally. The amended standard, effective 15 December 2026, also notes that auditor testing "might focus on the application controls built into" packaged software and the IT general controls around them, which means the vendor’s controls become part of your control environment.

The evaluation sequence

  1. 01Which layer is the AI in?
  2. 02Is the measurement deterministic, and can you prove it?
  3. 03What provenance comes with extracted fields?
  4. 04What is the review workflow for low-confidence output?
  5. 05What control documentation exists, and is there a current SOC 1 report?
  6. 06What happens on a software update? Can previously calculated periods change?

Test all of it on your genuinely difficult leases, not a clean sample. See how our lease accounting platform handles reproducibility and provenance together.

Frequently asked questions

Can AI do ASC 842 calculations?

The calculations are defined arithmetic under the standard and should be performed deterministically. Machine reading is well suited to extracting the inputs from lease documents, and it is the wrong tool for producing the measurement itself.

What is the difference between AI lease abstraction and AI lease accounting software?

Abstraction extracts terms from lease documents into structured data. Lease accounting software takes that structured data and produces the measurement, journal entries, and disclosures. Most "AI-powered" marketing describes the abstraction step.

Will auditors accept output from AI-based lease software?

Auditors assess controls rather than technologies. PCAOB AS 2201 treats an automated control as lower risk where relevant IT general controls are effective, so the question is whether the control environment supports reliance.

What is a SOC 1 report and why does it matter here?

It is an independent report on a service organization’s controls relevant to a user’s financial reporting. When a vendor’s software computes numbers that reach your statements, its controls form part of your control environment.

Emma Sukenik

Emma Sukenik is REAL’s Director of Business Development, focused on how enterprises recover spend across tax, lease obligations, and CAM reconciliations.

Director of Business Development, REAL

See REAL run end to end.

Watch a demo

Related posts

Book a Demo