The control stack
Five questions turn “it uses AI” into something inspectable.
The layers are ordered because later controls depend on earlier ones. A perfect approval screen cannot rescue the wrong source balance. A provider receipt cannot establish that the system had authority to send the request. A queue cannot repair duplicated actions if retries lack a stable identity.
01
Name the authoritative record
Which field wins when the lease, ledger, bank, and integration disagree?
Minimum evidence: Field owner, source identifier, effective time, and correction path
02
Reconcile money and obligations
Can a landlord trace each reported balance back to transactions and external evidence?
Minimum evidence: Opening balance, charges, receipts, reversals, adjustments, and cutoff
03
Bound the action authority
Is the system observing, preparing, requesting approval, executing, or prohibited?
Minimum evidence: Authority class, consequence, reversibility, policy version, and approver
04
Make delivery retry-safe
What prevents a timeout or duplicated event from becoming a duplicated real-world action?
Minimum evidence: Idempotency key, attempt history, provider identifier, and final outcome
05
Route uncertainty to an owner
When the workflow cannot prove the safe next step, who sees it and what must they decide?
Minimum evidence: Reason code, consequence lane, decision packet, owner, deadline, and closure proof
Choose the job in front of you
Five reading paths, not fifty keyword variations.
Each path owns a different operating decision. Definitions explain the record or control, guides explain implementation, articles examine the tradeoff, and tools let you test your own situation.
Choose a PMS
Test the records before the feature list
Use this path to evaluate exports, record authority, data ownership, and whether a future migration can be proved instead of merely attempted.
Add AI
Separate assistance from action
Use this path to decide what may be drafted, what requires approval, what can execute inside a narrow policy, and what stays human-only.
Connect systems
Design for late, duplicated, and missing events
Use this path when rent payments, messages, maintenance events, or accounting data cross system boundaries and the happy path is not enough.
Integration reliability checklist →Webhook delivery →Retry policy →Duplicate-charge protection →Dead-letter queue →Safe rent-payment integration controls →Integration retry runbook →Retry and duplicate-action simulator →Why an integration badge is not a control →Why AI property managers need provider receipts →
Close the month
Prove the report before acting on it
Use this path to trace rent-roll and owner-reporting numbers back to ledgers, cash movement, corrections, and the reporting cutoff.
Rent-roll reconciliation →Rental opening balance →Payment allocation →Ledger adjustment →Rental month-end close →Rent-roll reconciliation tool →Opening-balance acceptance guide →Unapplied-cash resolution guide →Payment-allocation review guide →Why unapplied cash is not extra rent →Month-end close is an exception review →
Run the week
Match review capacity to the work
Use this path to turn a small portfolio into a repeatable operating cadence and prevent an exception queue from becoming an unattended inbox.
Evidence test
Ask for the artifact, not the assurance.
“Logged,” “reviewed,” and “sent” sound reassuring but do not identify what actually happened. The right evidence depends on the claim. This comparison can be used during a vendor demo, migration rehearsal, incident review, or monthly close.
How should a small landlord use this library?
Start with one live workflow rather than trying to score an entire platform. For rent collection, identify the lease and ledger fields that establish the obligation, trace a payment through pending and settled states, simulate a timeout, inspect how a duplicate request is prevented, and find the final provider evidence. Then repeat the exercise for maintenance, notices, documents, and reporting.
A portfolio with three units does not need enterprise bureaucracy. It does need unambiguous records and a way to recover from mistakes. The practical standard is proportional: fewer roles and queues may be appropriate, but money, tenant-impacting actions, and hard-to-reverse decisions still deserve explicit authority and retrievable evidence.
What this library does not claim
These controls do not make every automated action safe, lawful, accurate, or appropriate. They do not replace jurisdiction-specific legal advice, accounting review, security assessment, vendor due diligence, or the landlord’s responsibility for consequential decisions. A diagram or checklist is useful only when the actual product behavior and records can be inspected.
Aptoria publishes a page only when it owns a separate reader job and adds an original operating artifact. Closely related wording is consolidated into one canonical page. See the editorial methodology for the review and anti-spam publication gates.