ARIA Intelligence v1.1.2.From detecting conflictsto finding opportunities.
Reasoning about when guests book, and on what terms.
ARIA's seventh domain agent analyzes the booking window. It surfaces when a cancellation policy is costing you — too loose for the lead time, or too strict and suppressing demand — and recommends holding or releasing last-minute inventory based on the demand signal.
Policy changes flow into the recommendation pipeline so the rest of ARIA always reasons with your current terms.

ARIA looks for upside, not just conflicts.
A new pricing-gap tension fires when your rate trails the market opportunity, with a tunable confidence threshold so you control how aggressively low-confidence calls are suppressed. A new minimum-stay tension catches length-of-stay mismatches.
A semantic 'Layer B' pass detects opportunity relationships in the graph — with advisory evidence, staleness guards, and a false-overbooking guard — plus an LLM-assisted scanner that proposes opportunities you never queried.

It recommends the automations you're missing.
The Automation Gap Scanner proactively recommends automations you don't have yet — recurring guest messages, task creation on specific events — working hand-in-hand with the Arqdia Automation Marketplace.
ARIA generates automation proposals with full context-fetching and planning, returned as inspectable cards, and can create or update automations with reversibility built in.

Good outcomes compound — across your portfolio and the network.
A new rejection-feedback loop tightens the signals behind any recommendation you decline. The planner now retrieves and weights past scenarios by whether they were approved and whether the outcome was good, so ARIA leans on what actually worked.
For organizations that opt in, cross-org global learning lets an approved outcome on one portfolio improve recommendations across participating organizations — while non-participating orgs are fully excluded.

ARIA asks before it assumes.
When a recommendation hinges on context ARIA doesn't have, it publishes a short clarification question to the operator — answerable in ArqChat or the dashboard — instead of guessing.
The answer flows straight back into the recommendation, keeping it accurate from the very first card.

Every plan declares how to undo it.
ARIA can construct the inverse of an executed plan, with tool specifications that declare their effect and reversibility. When two plans touch the same booking, it resolves the overlap and redacts the conflicting steps.
Postgres advisory locks ensure a plan never executes concurrently across replicas — and new delivery lanes (Slack, Gmail digest, Calendar) join WhatsApp, Telegram, and Teams.

Trust Ladder
ARIA earns autonomy,
one decision at a time.
Recommend & wait
Every recommendation requires explicit approval. ARIA never executes anything without your tap.
Assisted
ARIA executes low-risk, fully-reversible decisions automatically. High-stakes calls still wait for you.
Autonomous
ARIA acts on high-confidence calls without waiting. Every action audit-logged and reversible from the dashboard.
You stay in control
What ARIA still requires you to approve.
Any action that contacts a guest directly — messages, refunds, compensation.
Any cancellation, regardless of cost.
Any pricing change above a threshold you define in your workspace strategy.
Any action on a property still in Recommend & wait mode (default for new properties).
ARIA Intelligence · v1.1.2 Beta

Booking Window Agent
Reasoning about when guests book, and on what terms.
See ARIA decide
on a real property.
Get in touch and we'll show ARIA generating a live decision card on one of your real units.