What is an AI email assistant that drafts follow-ups but never sends without my approval?

What the job actually is

An AI email assistant that drafts follow-ups but never sends without your approval is a tool that writes the chase on a silent thread, then holds it until a person taps send. The send gate is structural: the outbound text is frozen, and Send, Edit, or Skip is a person's tap. A prompt that says draft only is not that gate.

Founders and operators live in Gmail for one-to-one threads with a warm intro, a prospect, or an investor. They wait on a reply, or they typed I'll send that Friday, and the thread dies. The buying prompt collapses four jobs: an on-demand draft in compose, a follow-up draft when Sent stays silent, a promise kept until Friday, and a sequence that sends on its own. Only the last one is a different product.

The leak they name first is an AI that emails someone under their name. Worst case with a draft is you delete it. Worst case with auto-send is you explaining the email on Monday. Promises you typed are a neighboring job, on how to stop forgetting emails you promised. Silence after send is on how to keep track of emails you're waiting on. The Slack-shaped send gate is on the approval-before-send page.

Which tool wins which job

Fund the row that matches the leak. A compose-box draft, a mailbox follow-up, and a send gate across Gmail and Slack are neighboring jobs.

A prompt is not a send gate

Draft, don't send only holds if the send path cannot run until a person records a decision on the frozen text. A system prompt that says draft only can drift in a long session; the model still has the send tool. People have watched a connected assistant email the team after they told it not to.

Fyxer's pricing FAQ is exact: it does not send emails on your behalf; you decide what gets sent. Superhuman Auto Drafts appear as ordinary drafts. If a placeholder is still empty when you hit send, Mail alerts you before the email goes out. Gmail Help me write stays in the compose box until you send.

Lindy's pricing FAQ says anything with outside impact waits for approval: sending an email, updating a ticket, posting to another channel, publishing a doc. Read-only lookups do not. Each integration then has Always allow, Require approval, or Don't offer. Guardrails apply in shared Slack threads only, not in the web app, Slack DMs, iMessage, or SMS. Always allow turns the gate off for later writes.

A follow-up draft is not a sequence

Most inbox tools watch silence after you hit send, then draft the bump. That is emails I'm waiting on a reply to. Superhuman Auto Reminders fire when the latest outgoing message has had no reply, and Auto Drafts write the chase about an hour before the reminder returns. Fyxer keeps those threads in Sent until your wait period passes, then labels them and drafts. Lindy's follow-up bumps routine does the same after 1, 2, 3, 5, or 7 days.

None of those, by itself, is a cadence. A sequence that sends without a tap is a different product, built for lists. One-to-one threads with a warm intro, a prospect, or an investor are not a list. Lindy can Always allow. Its meeting path is a different job too: after a recorded meeting, a follow-up email goes out to attendees.

Work across apps with a send gate is on Lindy vs an AI employee in Slack.

The caveat that should decide the setup

The failure operators name first is an AI that sends the follow-up on its own. Approve the exact text that will go out. Approving a summary of the action is a different control from approving the payload. A confidence score, Always allow, or a prompt that says draft only is a different setup.

Fyxer never sends drafts automatically. Superhuman Auto Drafts wait for you to hit send. Lindy's email-drafting and follow-up-bump routines land in your drafts folder, and Always allow exists. Its meeting follow-ups can send. Put the gate in writing before you connect Gmail to anything that can send.

Concrete is an AI employee that works in existing Slack channels and reads Gmail alongside them. It keeps every promise made in a thread, and every thread waiting on a reply, as a follow-up. Each morning it nudges the owner. It drafts the follow-up as a Send / Edit / Skip card in Slack. A Gmail send needs the same tap. There is no auto-send mode. Follow-ups survive reboots. It is in design-partner preview; the longer thesis is on the vision page. It does not run sequences or cadences.

A practical setup for a 2-to-50 person shop

This is the boring version that holds when the founder still owns the one-to-one threads. Pin it before the next silent intro.

• Name the gate. Send is a tap on the exact text. Edit rewrites it. Skip drops it. A prompt, a confidence score, or Always allow is a different setup.
• Keep client-facing send on the gate. Lookups, labels, and internal notes can run. Anything a prospect, investor, or customer will see waits. If the vendor offers Auto or Always allow, leave it off on send.
• Separate the chase from a sequence. A follow-up draft on a warm intro is one-to-one. A cadence that sends without a tap is a different product. Draft the chase. A person still sends.
• Put the morning list where you will see it. A follow-up that only lives in Drafts is easy to skip. Gmail Nudges and Superhuman returned reminders live in the mailbox. Concrete posts the list in Slack.
• Keep Gmail and Slack on the same list. The same wait is often an email and a Slack ping. Two tools, two lists, and the intro still goes cold. One follow-up that survives both is the job.