Short answer
Levy Street built an internal AI agent for APL Property, a New Zealand property services firm with offices in Rotorua, Wellington, Blenheim and Queenstown. The agent handles routine administration: triaging inbound email, drafting categorised responses, updating records and moving orders through a processing pipeline. Property managers review and approve its work, so judgement and client relationships stay with people.
- Built for a firm serving government, councils, commercial owners and private clients.
- Handles email triage, record updates and order processing.
- Staff approve outbound communication; the agent drafts and prepares.
- Client
- APL Property, a New Zealand regional property services firm
- Offices
- Rotorua, Wellington, Blenheim and Queenstown
- System
- Internal AI agent for administrative workflows
- Human role
- Staff review and approve outbound communication and decisions
About APL Property
APL Property provides property services to government entities, local and territorial authorities, commercial organisations and private owners, with offices in Rotorua, Wellington, Blenheim and Queenstown. Its work spans commercial property management, body corporate management, school property management, project management and valuation. That breadth creates a large volume of repetitive administration spread across many workflows and inboxes.
What we built
Levy Street built an internal AI agent that sits inside those workflows. It reads inbound messages, works out what each one is about and how urgent it is, drafts a response or prepares the next step, and records what happened. Work moves through a visible pipeline, so staff always know what is queued, in progress or done.
- Inputs
- Inbound email, attachments, property and client records, order details.
- Agent work
- Categorise and prioritise, extract details, draft responses, update records, create and progress orders.
- Human review
- Staff approve outbound communication and any decision with client or financial consequences.
- Outputs
- Categorised inbox, drafted replies, updated records, an auditable order timeline.
Design principles
The agent follows the rules we apply to every system that touches clients and money:
- Draft, do not send. The agent prepares; people approve anything that leaves the business.
- Narrow permissions. It can read and update what its workflows need and nothing else. See agent permissions and approvals.
- Visible state. Every item has a status and a history, so nothing disappears into a model.
- Escalate the unusual. Messages the agent is unsure about go to a person rather than getting a confident guess.
Lessons for property managers
Property management is a strong fit for AI administration because the work is high-volume, text-heavy and rule-bound, while the judgement calls (disputes, tenancy decisions, client relationships) are easy to keep with people. Our guide to AI for New Zealand property management maps the workflows, including maintenance request triage and inspection report drafting, and the tenancy rules that apply.
If your team spends hours a week on inbox triage and record keeping, a workflow automation assessment is the place to start.
Written by the Levy Street team in Wellington. Last updated 19 October 2026.