The loop
An agent drafts from live data. You review on the record or in chat. Approve, edit, or dismiss; it is logged either way.
Maya AI
Maya AI proposes.
You approve.
Twelve agents read your live data and draft the next move. Every draft waits on a person, autonomy is a dial you raise per agent, and every outcome is logged.
Illustrative product preview. All records and counts are sample data.
What it does
The fleet
Twelve agents, each a card with a toggle per organization. Six you will feel in week one.
Governance
Four autonomy levels per agent, starting off. Anything that moves money waits for a person at every level.
How it works
- 01
An agent runs
Nightly sweeps, hourly checks, or event triggers. What it produces is a draft.
- 02
You review
On the record or in chat with Maya. Approve, edit, or dismiss.
- 03
It is logged
Every decision lands in the audit trail. Automatic actions are recorded as system actions, never as a fake user.
- 04
Autonomy earns its way up
Watch only, propose, propose and notify, or auto-approve safe actions. Health is computed from seven days of your approvals, dismissals, and reversals.
Six you will feel in week oneSample data
SLA Breach PredictorHourly1 proposal
Fraud & Abuse DetectionOn eventHealthy
Photo InspectionOn eventHealthy
Return ForecastDailyHealthy
Low-Stock Reorder AlerterNightlyHealthy
Vendor News WatchNightlyHealthy
Off by default.
On when you trust it.
Raise an agent's autonomy and it applies safe, non-monetary actions itself. Money-moving actions always wait for a human, however confident the agent is.
About Maya AIDo agents act on their own?
Not out of the box. Every draft waits on a person until you raise that agent's autonomy. At its highest level an agent may apply safe, non-monetary actions itself. Anything that moves money waits for a human at every level.
How many agents are there?
Twelve, each a card with a toggle, on or off per organization. Six you will feel in week one: SLA breach predictor, fraud and abuse detection, photo inspection, return forecast, low-stock reorder alerter and vendor news watch.
What are the autonomy levels?
Four, set per agent: watch only, propose, propose and notify, or auto-approve safe actions. Autonomy starts off.
Is there a record of what an agent did?
Yes. Approve, edit or dismiss, the outcome is logged either way, and automatic actions are recorded as system actions, never as a fake user.
Does Maya AI work on the recovery side?
Yes. Agents explain why an invoice did or did not qualify, run a pre-flight check over a packet before it is committed, and draft the cover memo for a claim.
How do we switch agents on?
Agents ship with the platform and switch on per organization, each a card with a toggle.
Put the fleet to work.
Arthiva is invite-only while we build with early customers. Agents ship with the platform and switch on per organization.