How the board supervises the AI workforce

AI HOA's core promise to boards is leverage without loss of control: AI staff do the drafting and legwork; humans approve anything that matters. Here's how that actually works.

The AI workforce

Under AI Workforce in the dashboard sidebar, your community has a team of specialist agents:

• Research Assistant — cited answers from governing documents and Georgia law (residents get this one too, as the portal's AI Assistant).
• Board Secretary — drafts agendas, minutes, and announcements.
• Treasurer's Assistant — financial questions and bookkeeping support.
• ACC Reviewer — pre-reviews architectural requests against your covenants.
• Delinquency Manager — aging, late-fee notices, collection follow-ups.
• Budget Drafter — full draft budgets from your actuals.

Review-before-execute

The default posture for anything consequential is review before execute. When an agent produces something with real-world effect — a violation notice, a collection letter, meeting minutes — the output lands as a draft in the Supervision Inbox (sidebar → AI Workforce → Supervision Inbox). A board member opens the draft and can:

• Approve — the action proceeds (the notice sends, the minutes record).
• Edit, then approve — your changes are what goes out.
• Reject — nothing happens, and the rejection is recorded.

The inbox badge shows how many drafts are waiting. The decision — who made it, when, and what changed — is written to the audit trail alongside the draft itself.

Tuning how much supervision each agent gets

Under Supervision Settings, boards set the mode per agent. Routine, low-stakes output can be set to auto-resolve; sensitive domains can require review on everything, or escalate to specific people. Most boards start with everything supervised and relax deliberately as trust builds — the platform is designed to make that a per-agent dial, not a leap of faith.

Why this design

Volunteer boards get sued over process, not typos: who approved this notice, was the fee authorized, when was that decided. Routing agent output through explicit human decisions means the answer always exists — and it's a person, with a timestamp. The AI never has authority; it has a very fast pen. Costs stay visible too: every agent invocation is logged, so the platform can account for what its AI staff did and spent.

As with everything the agents produce about Georgia association law: it is general information, not legal advice — consult an attorney for your specific situation.