The short answer: yes. alfred_ writes the reply in the client’s email thread when you ask and shows it to you. It sends only when you approve that email, with Send in the app or “send” by text. A draft you don’t approve isn’t sent. The only email it sends without asking each time comes from a rule or routine you set up yourself to send, and you don’t have to set up any.
Why “it might send” is a dealbreaker
You’re an associate at a seven-lawyer litigation firm. You spend an hour a day on client email that is mostly the same shape: confirming a deposition date, explaining what a motion means in plain English, telling a client the other side hasn’t responded yet. An assistant that drafts those would save real time.
But an email to a client is advice, and an email that goes to the wrong person, says the wrong thing or goes out before you’ve read it is a problem you can’t take back. “The AI usually asks first” isn’t good enough. You need to know exactly when something leaves your outbox.
What “nothing sends without your OK” means here
You ask, it writes. Ask alfred_ for a reply (“reply to Ms. Chen confirming the deposition is moved to the 18th, and that she should bring the bank statements”). It writes the email in her thread, from your address.
You see it before it goes. In the app it appears as a draft with Send. By text, alfred_ sends you the full reply and asks “Send it?”
Only an approval sends it. Send in the app. By text: “send”, “yes” or “do it”. “No” or “cancel” stops it. Anything else, like “say the 19th” or “drop the second paragraph”, changes the email and shows it to you again. The short replies are in Confirming by text.
Unapproved means unsent. If you don’t answer, nothing goes out. A draft you walk away from doesn’t send itself later.
One email per approval. If one text carries two drafts, a thumbs-up can’t say which one you mean, so alfred_ waits for you to say which.
The one exception, which you control
alfred_ has rules: standing instructions you write, such as “forward invoices from our court reporter to billing@yourfirm.com” or “text me when opposing counsel emails.” A rule you create acts on each matching email without asking each time. That’s what a rule is for. Routines work the same way: a routine asks before anything goes out unless you tell it to send without asking. If you don’t want anything going out automatically, don’t set up a rule that forwards email or a routine that sends on its own. Rules are listed in Settings, Rules and routines on the Routines page, where you can change or turn off each one.
Where each step happens
- At your desk: ask in the app; the draft opens with Send. Edit it there or ask for changes.
- Between hearings: by text. alfred_ texts you the reply and waits for “send”.
- On iPhone: the same as the desk, in the app.
What it doesn’t do
- It doesn’t decide what to tell a client. You say what the reply should say; it writes it properly.
- It doesn’t send on a guess, or because an earlier draft looked fine.
- It doesn’t connect to case management software. It works from the email, calendar and files you connect.
The honest caveat
Approval is only as good as the reading behind it. alfred_ shows you the whole email, recipients included, every time. Read it like you would read an associate’s draft before it goes out. Whether using an AI assistant on client email meets your jurisdiction’s ethics rules is your firm’s call, and the facts for that call are in the privacy policy and on the security page. alfred_ is a paid product; plans are on the pricing page.
