Security

See what Executive Assistant can do before you let it act.

Signals stay visible, permissions stay explicit, and every recommendation stays attached to the office loop. The goal is dependable execution, not hidden automation.

Can it act without me?Drafts, prioritization, and recommendations can help early. Sensitive actions and outbound delivery stay reviewable so the office remains in control.

Does it imply broader access than it has?No. Each connection has a visible support contract so the product does not imply read, write, import, or sync behavior it cannot deliver.

Does context disappear?No. The workspace can keep people context, commitments, approvals, and recurring office patterns so the next loop starts from reality instead of from zero.

Trust model

What an office usually wants to confirm

Most offices want three answers quickly: what Executive Assistant can see, what it can prepare, and where human review still stays in the loop.

Tight account permissions

Workspace access uses secure email. Google is an optional integration later, not a hidden demand for broad mailbox permissions.

Visible review boundaries

The product should say clearly what it can read, draft, recommend, or send today, and where approval is still required.

Durable office context

People context, commitments, and evidence should stay attached to the office loop instead of drifting across isolated chats and prompts.

Where context is stored

EA Postgres is the durable source when durable storage is configured.
Operational projections are mirrors, not the only source of truth.
Large files and exports belong in object storage when the deployment supports it.

Saved review defaults

Retention: not set
Metadata-only channels: none
Drafts allowed: false
Action suggestions: false
Automatic digests: false