Agentic Operating Model
Overview
1 / 6
First map the Queensland residential property management operating model. Then assess where AI can carry the work and the assurance each workflow needs.
Define the work that must exist before designing AI capability, assurance checks and the right level of human involvement.
The scope is intentionally narrower than "the whole company" and broader than "what a property manager personally does". Read this page top to bottom — the order is deliberate.
Map the industry-standard operating model required to manage a residential rental property from owner acquisition/onboarding through the full property and tenancy lifecycle in Queensland.
This baseline focuses on mainstream Queensland general tenancies such as houses, units, apartments and townhouses. Rooming accommodation, moveable dwellings and specialist disability accommodation can be added as explicit extensions.
Not in the first universe: corporate HR, payroll, general IT, unrelated sales operations and broad corporate marketing.
Vanilla is compared against the benchmark later; current practice does not define the benchmark.
An L2 must be a core business process with a recognisable outcome. Forms, reports, individual decisions and atomic actions belong later under deliverables or tasks.
After the process universe is validated, we can assess AI capability and then design the checks, loops, golden examples and human involvement around each workflow.
The level of involvement scales with consequence: external or high-impact work receives stronger validation and approvals than low-risk internal work.
Avoid automating isolated tasks before we understand the operating system they belong to.
Important distinction: L1 and L2 are process levels. Deliverables and tasks appear only after an L2 core process is selected for decomposition in Phase 3.
Explore the normalised L1 domains and L2 core processes. Search any topic to test coverage, or select a domain to inspect its process set.
This is a structural check on the map. It compares the number of normalised Level 2 core processes inside each Level 1 domain; it does not measure workload, importance, complexity or AI potential.
Use the distribution to challenge domains that look unusually thin or unusually broad before we start decomposing them.
Select any domain row to reveal its Level 2 core processes. Select it again to collapse.
The map should not be perfectly even. We are looking for obvious structural outliers, not uniformity.
A very dense domain may contain multiple workflows or exception paths that should eventually be separated.
If most domains sit in a similar band, that is a useful sign that the first-pass architecture is reasonably balanced.
The Green / Orange / Red assessment is only a capability lens. It does not mean “Green = no human”. In the eventual agent file structure, human involvement is calibrated alongside golden examples, checks and operating loops.