The idea
I designed and built an interactive case management prototype for property management teams to explore how AI can become part of operational workflows, rather than sitting beside them as an assistant.
Property managers receive a constant stream of messages about repairs, payments, documents and ongoing issues.
The challenge is not writing replies. It is understanding what a message belongs to, what has already happened and what should happen next.
I used this as a playground for designing AI around triage, context and workflow automation.
AI Triage
AI does the first pass.
Humans handle uncertainty.
Incoming messages land in a triage inbox. The system interprets the message, identifies its intent and connects it to the relevant property, unit, people and existing case.
When the context is clear, it can propose the next step or continue the process. When information is ambiguous, it surfaces the uncertainty instead of pretending to know.
I designed the interaction around that boundary: automate what is clear, make uncertainty visible and give humans an obvious point of intervention.
Agentic Workflows
Making AI behavior explicit
I designed a node-based workflow builder to make automated processes visible and configurable.
Workflows can combine triggers, conditions, AI decisions, system actions and human checkpoints. Instead of hiding automation behind a prompt, the workflow shows when AI acts, what context it uses and where humans remain in control.
The entire flow is implemented as an interactive prototype with simulated end-to-end logic inside the application.
