How do approvals work in Lindy—can I review drafts before it sends emails or texts?

You stay in control the whole time. Lindy can fully run your inbox and texts, but every email or message it sends can go through you first—down to the exact draft, tone, and timing.

Here’s how approvals work, and how to review drafts before anything goes out.

The short version: Yes, you can approve every draft

If you want Lindy to “ask before sending,” you’ll use Training Mode and approval rules:

Lindy writes the email or text in your voice.
It sends the draft to you (usually via iMessage/SMS, or the channel you’re using).
You review, edit, or decline.
Only after you approve does Lindy send the message to the recipient.

No silent sends. No surprises.

Once you’re comfortable, you can gradually relax approvals for low-risk messages while keeping tight control on sensitive ones.

Approval modes: How much control you want

Lindy supports three main operating patterns for approvals:

1. Full Training Mode (Approve everything)

Best when you’re just getting started or testing Lindy with important relationships.

In this mode:

Lindy always waits for your approval before sending any reply or outreach.
You see the full draft (subject, body, and recipients for email; full text content for SMS/iMessage).
You can:
- Approve as-is.
- Edit and approve.
- Reject and write your own.

This is how you “train” Lindy:

Every edit is feedback.
Lindy learns your tone, phrasing, and preferences.
Over time, drafts start coming out exactly how you’d write them, but faster.

Use this if:

You’re nervous about AI sending anything without you.
You’re onboarding Lindy to a busy inbox or a high-stakes role (sales, partnerships, investors).
You want to refine its style before giving it more autonomy.

2. Rule-Based Approvals (Approve some, auto-send the rest)

Once you trust Lindy with the basics, you can set Triage Rules in natural language so it knows when to wait for you and when to just handle it.

You can say things like:

“Always ask for approval before emailing new contacts or VIPs.”
“Auto-send simple confirmations and scheduling replies.”
“If a message involves pricing, contracts, or legal, wait for my approval.”
“For warm leads, you can send follow-ups without asking unless they mention budget or timeline.”

Under the hood, Lindy:

Scores and classifies messages (e.g., lead, support, info request, negotiation).
Matches them to your rules.
Decides whether to:
- Draft and send for approval, or
- Draft and send automatically.

This gives you:

Tight control over high-risk or high-value communication.
Zero-friction handling of routine back-and-forth (confirmations, “sounds good, thanks,” scheduling nudges).

3. Auto-Mode with Safety Nets (Approve exceptions)

For people who’ve fully replaced their inbox triage with Lindy, approvals are reserved for edge cases.

Typical setup:

Routine threads, reminders, and calendar follow-ups → Auto-send.
Important or ambiguous threads → Ask for approval.
Anything that hits your “do not mess this up” criteria (e.g., key account, investor, boss) → Always approved manually.

You can define these safety nets using natural language:

“Anything involving [big client] must be approved.”
“If an email looks sensitive or emotional, don’t send without checking with me.”
“If you’re not at least 80% confident, draft and ask me first.”

You get the speed of a fully autonomous assistant with the control of a human review where it matters.

What the approval flow actually looks like

Here’s a concrete example of reviewing drafts before Lindy sends them.

Example: Email reply approval

Someone emails you:
- “Hey, can we schedule a call next week to discuss a partnership?”
Lindy reads the email and your calendar.
It drafts a response in your style:
- Proposes two time slots.
- Adds your usual sign-off.
- Keeps the tone consistent with your past replies.
You get a text from Lindy:
- “Drafted this reply to [Name]. Want me to send?”
- It includes the full email draft.
You respond:
- “Looks good, send” → Email is sent immediately.
- Or you tweak the text and say “Use this instead” → Lindy updates and sends your edited version.

Every approval and edit becomes training data for your voice and preferences (not for generic model training; Lindy doesn’t sell your data or use it to train base models).

Example: Text message approval

For SMS/iMessage style outreach or responses:

Lindy drafts the reply in the same thread where you’re talking to it.
You see the exact message that will be sent to the other person.
You reply with one tap:
- “Yes” / “Send it” → Lindy sends.
- “Change ‘Let’s hop on a call’ to ‘Want to do a quick Zoom?’ and send” → Lindy edits and then sends.

Same pattern: you review before anything goes out.

Controlling tone while you approve

Approvals aren’t just about “yes” or “no.” They’re where you shape Lindy to sound like you.

You can:

Paste example emails and say:
- “Reply in this tone.”
- “Keep everything more casual/concise going forward.”
Correct specific things:
- “Don’t use ‘Best,’ use ‘Cheers.’”
- “Stop saying ‘circle back,’ just say ‘follow up.’”
Set persistent style rules:
- “Be friendly but direct.”
- “Never over-promise; be clear about timelines.”
- “If someone is upset, be extra empathetic and solution-focused.”

With Training Mode + feedback, Lindy rapidly stops feeling like “an AI” and starts feeling like the version of you that got a good night’s sleep and has zero inbox anxiety.

Where approvals apply: Emails, texts, and beyond

You can choose approvals for:

Email replies and outreach
- Inbox triage responses.
- Lead follow-ups and nurture sequences.
- Partnership, sales, and negotiation threads.
Texts & iMessage-style communication
- Client updates.
- Quick confirmations.
- Follow-up nudges.
Multi-step actions that touch other tools
- “Email Jane, then add a note in the CRM.”
- “Text the team, then move the meeting on my calendar.”
- “Reply to this support email and then tag the ticket as ‘Resolved.’”

For these workflows, approvals can gate:

The message itself.
The downstream action (e.g., updating a CRM, marking something closed).

You can decide:

“Approve the email, but auto-handle the CRM updates.”
Or “Ask me before both sending and updating any external system.”

Keeping humans in the loop for safety & compliance

For teams, approvals double as compliance and quality control.

You can:

Put specific agents in permanent training mode so a manager approves everything they send.
Set team-wide rules:
- “All outbound sales emails must be approved by a human.”
- “Any email involving PHI or sensitive data requires manual review.”
Use enterprise controls (SSO, SCIM, audit logs) to see who approved what and when.

Behind the scenes, Lindy:

Uses encryption by default.
Keeps a clear separation between your data and the underlying models.
Does not sell your data or use customer data to train base models.

Approvals are built into this model: you decide what Lindy is allowed to say and do on your behalf.

How to decide your approval strategy

Use this as a quick decision framework:

If you’re new to Lindy:
Turn on Training Mode → Approve everything → Give lots of feedback for the first 1–2 weeks.
If your inbox is overwhelming but high-stakes:
- Approve:
  - New contacts
  - Deals, investors, VIPs
  - Anything with pricing or legal implications
- Auto-send:
  - Confirmations
  - Simple scheduling
  - Routine status updates
If you’re ready for “set it and forget it”:
- Auto-send for:
  - Leads below a certain priority.
  - Standard follow-ups and reminders.
- Require approvals only for:
  - Sensitive accounts.
  - Leadership and board.
  - Messages that look emotional or ambiguous.

You can adjust these rules anytime, in plain language. No config hell. Just tell Lindy what you want.

Final verdict

Yes—you can absolutely review drafts before Lindy sends emails or texts. Approvals in Lindy are flexible:

Training Mode if you want to see everything.
Rule-based approvals if you want to mix auto-send with human review.
Safety nets so important relationships always get your eyes first.

You get the time savings of an AI assistant that actually does stuff, with the control of “nothing leaves my name without my say-so.”

Next Step

Get Started