Aptoria
Features
Product
Resources
Company
Tools
Log in
See the demo
REVIEWABLE WORKFLOW ACTIVITY
Trust isn't a feeling. It's a record.
Aptoria is designed to make supported agent-assisted work inspectable: what the workflow prepared or handled, what it asked you to approve, and what remains reserved for a human. Each supported action records the available source facts, policy, approval state, and outcome so your team can review it after the fact.
Reviewable action record
Hash-chained
Exportable review record
Source and policy context
Supported decisions land in one lane
What it handled. What it routed. What remains human.
Supported agent workflows resolve into one of these lanes when the policy engine has enough context, and each supported lane records available policy and review state.
THE AGENT ACTED
Handled on its own
For a supported, configured workflow inside your thresholds, the agent can handle a routine task such as preparing a tenant reply or coordinating maintenance. Provider-backed payments and dispatches have their own setup and approval conditions. The action record captures the policy, inputs, and review state so the work can be inspected later.
THE AGENT ASKED
Surfaced to you
When a decision falls outside your thresholds — a fee above your cap, a maintenance bid over your approval-rule limit, a message it isn't sure about — it stops and routes the item to your queue with the full context. Your approve/deny is recorded next to what the agent proposed.
HUMAN-REQUIRED FLOOR
Held for an accountable person
Eight defined action classes — including eviction filings, applicant denials, lease terminations, deposit deductions, rent-increase notices, and cross-landlord pricing signals — are not autonomously executed through supported Aptoria workflows. The product can prepare context while the authorized person retains the decision and any required external process. This product boundary is not legal advice or a compliance certification.
THE RECORD
One reviewable record per supported action.
Supported agent actions record the decision, policy context, lane, available source facts, approval state, and content hashes. For workflows that currently write hash-chain metadata, exports can be checked against available recorded hashes. Records are operational review artifacts, not complete legal records.
Supported decision records preserve available policy, source, and review context
Policy changes can snapshot BEFORE/AFTER where that workflow records them
"What was the configured dispatch cap on March 5?" has a reviewable answer where available fields record it
agent_decisions · row #4,812
Proposed step
Illustrative routine action · $50
Policy context
Configured rule · source review required
Lane
Eligible agent-assisted workflow
Soft-commit
Illustrative eligible window · not cancelled
Prev-row hash
3f9a…c17b
Row hash
a20e…88d4
Illustrative chain check - available hash fields match
Illustrative row. Field names map to real columns; values are an example.
How we measure
Numbers with a published method.
We measure supported metrics from available account snapshots and publish aggregate data only after the k-anonymity floor is met. The aggregate published here is weighted by units and gated by a 3-landlord floor so no single account can be reverse-engineered from it.
Methodology
At signup, each account's prior-12-month metrics are frozen as a "baseline" snapshot. Every Sunday, cron-cohort-outcomes recomputes the trailing-90-day metrics and the delta vs baseline. Aggregates here are weighted by units (a 50-unit landlord isn't averaged against a 1-unit landlord). The 3-landlord k-anonymity floor protects single-account reverse-engineering. Source data is the landlord's own payments / leases / work-orders — no synthetic or projected numbers.
Your record starts the day you sign up.
Availability depends on the workflow, provider setup, and your configured policy. Review product availability and test a bounded workflow before relying on it in production.
Start free trial
See the safety architecture