Question 1

Will it send emails without me?

Accepted Answer

Only if you switch a context to Autopilot and the action clears your guardrails — a confidence floor, an allow-list of domains, and a send-delay you can undo. In Copilot, the default most people start with, nothing leaves without your one-click approval.

Question 2

How does it learn my voice?

Accepted Answer

It studies your own sent mail to match your tone, formality, and phrasing, and pulls real facts from your context engine rather than inventing them. You can fine-tune the model and creativity in Settings → AI & Voice.

Question 3

Can I undo what it does?

Accepted Answer

Yes. Every autonomous action is reversible within a window and written to an append-only audit log with its reasoning, so you can always see — and roll back — exactly what happened.

Question 4

Can AI really run my inbox?

Accepted Answer

More than you'd expect. The agent reads every new message at sync, scores its priority and intent, runs your rules, drafts replies in your voice, stages follow-ups, and — in Autopilot — sends within your bounds. The typing was never the real work; the deciding, remembering, and chasing were, and that's exactly what it takes off your plate. You stay at the helm: in Copilot it prepares everything and waits, in Autopilot it acts within guardrails, and either way every move is reversible and logged.

Question 5

What is an AI email agent?

Accepted Answer

An AI email agent is software that doesn't just draft on command — it acts on your inbox the way an assistant would. AI Emaily's agent triages mail as it lands, decides what matters, writes replies in your tone, schedules sends, and closes loops by following up when threads go cold. It runs server-side at sync, so your inbox is already organised before you open the app. Unlike a chatbot bolted onto email, it carries reasoning and a confidence score on every action, and stays bounded by the authority mode you choose.

Question 6

How is this different from a draft button?

Accepted Answer

A draft button stops the moment text exists — you still triage, decide, schedule, and chase every loop yourself. AI Emaily's agent crosses that line. It reads new mail on arrival, prioritises it, prepares the reply with real facts from your context engine, and stages or sends follow-ups on its own when you allow it. Drafting is one step of many here, not the whole product. The goal isn't to help you type faster; it's to hand you decisions instead of a queue.

Question 7

Does the agent work while I'm offline?

Accepted Answer

Yes. The agent runs server-side at sync, not in your browser, so it keeps triaging mail, scoring priority, and staging replies even when you're asleep, travelling, or heads-down in deep work. You come back to an inbox that's already sorted with replies waiting. In Copilot those drafts sit for your one-click approval; in bounded Autopilot routine actions can dispatch within your guardrails, each with a send-delay you can undo.

Question 8

How accurate is the AI triage?

Accepted Answer

Every message gets a priority and intent score, and the agent shows its reasoning and a confidence number so you can judge the call rather than trust it blindly. When confidence is low, it escalates to you instead of guessing. Accuracy climbs as your rules and context engine teach it your judgement — what counts as VIP, which senders matter, what an investor update looks like. And because nothing is silently deleted, a mis-triage is always one tap to correct.

Question 9

Is it safe to let AI handle my email?

Accepted Answer

AI Emaily is built so autonomy never means losing control. Email is treated as untrusted input with prompt-injection defences and an action allow-list, so a malicious message can't trick the agent into acting. Autopilot is bounded by a confidence floor, a domain allow-list, and a send-delay undo window, and Copilot requires your approval before any send. Every action is reversible and written to an append-only audit log. You can halt an in-flight run or drop the whole agent back to Manual at any time.

Question 10

Does AI Emaily train on my emails?

Accepted Answer

No. We never train models on your mail. The agent learns your voice and context only to serve you — your sent history and context engine inform your own drafts and nothing else. AI calls run with zero-retention from model providers, so prompts and content aren't stored or used to improve anyone's model. If you'd rather keep keys entirely your own, BYOK runs drafting on your provider account; sensitive keys are decrypted only in an isolated worker, never client-side or logged.

Question 11

Which email providers does the agent work with?

Accepted Answer

Every major one. The agent runs on top of your connected mailboxes — Gmail, Outlook and Microsoft 365, iCloud, and any standard IMAP account — and works across all of them in one unified inbox. You don't switch email providers to use AI Emaily; you connect the accounts you already have, and the agent triages, drafts, and follows up across every one of them at once with least-privilege access.

Question 12

How long does it take to set up?

Accepted Answer

Minutes. Connect a mailbox — an app password for Gmail or Google Workspace, one-click OAuth for Outlook — and the agent starts triaging on the next sync. It works out of the box in Copilot, and gets sharper as you add context — client profiles, variables, and a few rules. Voice-matching kicks in automatically as it studies your sent mail; there's no model to train by hand. When you're ready for more autonomy, switch contexts to Autopilot and set your guardrails. You can start cautious and graduate at your own pace.

Question 13

What does the AI Chief-of-Staff cost?

Accepted Answer

There's a free $0 tier to try the client and on-demand AI help. Pro at $17.99/month unlocks the full agent — voice-matched drafting, triage, context engine, and Copilot. Autopilot at $29.99/month adds bounded autonomous send, scheduling, and follow-up with the full guardrail and audit stack. BYOK lets you run drafting on your own provider key without metered caps. You can move up or down between tiers as your trust in the agent grows.

Question 14

Will the agent send something embarrassing?

Accepted Answer

It's designed so it can't surprise you. In Copilot, the default most people start in, nothing leaves without your one-click approval. In Autopilot, sends only fire when confidence clears your floor and the recipient is on your allow-list, and every send waits out a delay window where one tap pulls it back. A 'did-you-forget' check flags unaddressed commitments before anything goes out. And every action is logged with its reasoning, so there are no silent moves to discover later.

Question 15

Can I see why the agent did something?

Accepted Answer

Always. Every action — a triage decision, a draft, a send, a follow-up — carries its reasoning and a confidence score, and is written to an append-only audit log. You can open any item and see exactly what the agent concluded and why, then roll it back within the undo window if you disagree. Trust here is earned through transparency, not assumed; nothing the agent does is a black box.

An AI that actually runs your inbox.

Drafting was never the bottleneck.

From inbox to done, step by step.

Everything this does for you.

Live on your next sync.

The payoff.

Where it earns its keep.

Good to know.

Related documentation.

Meet the inbox that works while you don't.