Eleven questions for evaluating AP automation in property management

We sell one of these, so treat the framing with the suspicion it deserves. The questions are still the useful part, and they are deliberately ones we have to answer too.

Feature checklists do not separate vendors in this category. Everyone has OCR, an approvals module, and an ERP connector. The differences show up on your invoice and at month-end.

The four that decide your actual cost

1. What exactly is billable? Not the rate. What counts. Per invoice posted, per document received, per page, per user, per transaction with reminders and rejections included? On a realistic multifamily document mix, 2,665 posted invoices produced 5,495 billable pages, and about 1,035 of those pages were documents that were never payable. Get the unit in writing.

2. What happens when a vendor sends a statement, a W-9, or nine invoices in one PDF? This is the follow-up that produces the interesting silence. Compliance paperwork you required your own vendors to send should not be billable.

3. What is the annual cost including implementation? A savings figure that omits the software price is a gross number, not a return. Ours is $99 a month plus $1 per invoice uploaded, and our ROI calculator divides by an annual cost of $38,000 rather than hiding it.

4. What is the implementation timeline, and what does it require from my PMS vendor? A project that needs work from ResMan or Yardi has a schedule you do not control. A Yardi integration is commonly a six to twelve month investment.

The three that decide whether it works on your documents

5. Send them a batch PDF with several invoices across several properties, before signing. Ask what comes back. Then ask whether the split is automatic or human-approved. Ours is human-approved deliberately: nine invoices assigned to the wrong nine properties is nine miscoded transactions, and finding that at close costs more than twenty seconds of review.

6. Send them a handwritten invoice photographed on a dashboard. This is a real and recurring document type in multifamily. Most tools route it to a manual queue and the giveaway is a "scanned documents" bucket in the interface.

7. What share of invoices post with no human touch, on a portfolio like mine, and what is the denominator? Ask for the denominator, not just the percentage. Ours is 98% on 2,665 invoices, with 197 reaching a person.

The two that decide whether close gets faster

8. What is the PMS sync interval in minutes, and is it one-way or two-way? A batch sync on a several-hour cycle means an invoice can validate against a supplier record that has since changed. Ours is real time, and ResMan is our only integration.

9. Can I see approval aging bucketed by property and by approver? Approval latency, not approval time, is what delays close. An average is useless; you need the distribution. Ours buckets under one day, one to three days, and over three days.

The two nobody asks and everybody should

10. What does your error queue tell me, beyond that something failed? A queue that says "error" is a list of work. A queue that says supplier code mismatch, property code mismatch, price mismatch against the PO, missing invoice number, or failed extraction tells you which of four owners has a systemic gap.

11. What happens to my coding if two of my properties code the same expense differently? The honest answer is that the model learns both and applies whichever it saw more of. Any vendor claiming their AI resolves your chart of accounts inconsistency for you is overselling. Cleaning the GL is upstream work that pays off more than automating it.

Where we are the wrong answer

Two places where we are the wrong answer, stated so you do not waste a demo finding out.

If you are on Yardi, RealPage, AppFolio, MRI, or Entrata, we do not integrate and this is a short evaluation. Several incumbents do.

If you need a large supplier payment network and a meaningful share of your vendors already transact on one, an incumbent's network makes onboarding easier than ours.

If you are buying at enterprise scale and need the procurement posture that comes with it, vendor questionnaires, formal SLAs, a named implementation team, an incumbent meets all of that without a conversation.

If your monthly volume is under a couple of hundred invoices, the arithmetic probably does not clear the software cost, and you should keep doing it by hand.

How to actually run the evaluation

Do not run a feature comparison. Run a document test.

Pull thirty real invoices from last month, deliberately including the two worst batch PDFs, one handwritten pad invoice, a multi-page statement, and a W-9 a vendor attached. Send the same thirty to every vendor.

Compare what comes back coded, what lands in the error queue, and what the error says. That takes an afternoon and it tells you more than a quarter of demos.

Which of the eleven does your current tool answer badly?