AI outbound tool that drafts personalised emails but waits for my approval before sending

Several tools do this, with different defaults. Unify drafts from research and queues every draft for a rep to approve. Artisan lets you require approval before sending, though its AI BDR is autonomous by default. Apollo lets AI act as a co-pilot or run on autopilot. Nagent's NORA researches each account, drafts from that research, and on a new workspace holds every send, follow-ups included, for a person. If approval is the point, choose a tool where it is the default.

Which AI outbound tools draft emails and wait for approval?

Every vendor fact below comes from the vendor's own site, checked 24 October 2026. We make Nagent; NORA is listed last and described against the same headings.

Unify

• What it is: AI-assisted outbound with signals, research, sequences and a dialer, built for reps rather than replacing them.

• Approval: its research-to-outreach page says drafts queue for approval before anything sends, and every step is a draft until you approve it and start the sequence.

• Pricing: Free for up to 3 seats; Base $20 and Pro $60 a seat a month; Business custom (pricing).

• Limitation: approval happens per rep, so throughput depends on how many reps review.

Artisan

• What it is: Ava, an AI BDR that finds leads, writes, handles replies and books meetings.

• Approval: its pricing FAQ says you can approve before sending, lock tone and calls to action, set banned phrases and coach Ava's writing, and describes Ava as "Autonomous by default, adjustable by design".

• Pricing: quote only; Team is sized at about 2,500 leads contacted a month, Scale about 6,000.

• Limitation: approval is something you switch on, so confirm it is set before the first campaign.

11x

• What it is: AI agents for outbound and inbound across email, voice, chat, social and SMS.

• Approval: its Deep Research page says you can check sources, change the angle or override the recommendation before outreach is sent. Its outbound pages describe a largely hands-off motion, so ask how per-email approval is configured.

• Pricing: no public price; the pricing URL returned a 404.

• Limitation: built for scale across channels, which can sit uneasily with reviewing every message.

Apollo

• What it is: a contact database with sequences, AI research and AI email writing.

• Approval: its AI page offers AI as a co-pilot or, once you are ready, on autopilot, without detailing the review flow.

• Pricing: Free, Basic $49 and Professional $99 a seat a month in the pricing page's structured data; Custom by quote.

• Limitation: drafting quality depends on the data and prompts you set; review is your process, not a gate.

NORA (Nagent)

• What it is: Nagent's AI Outbound Research Lead. She finds accounts that fit your profile, researches each into a sourced dossier, and VANI (Beta) drafts the email from it. See NORA.

• Approval: sending is gated separately from her autonomy level. On a new workspace every send waits for a person, and running the follow-up cadence is held too. Approval is checked again immediately before the email leaves.

• Pricing: no per-agent price; she works on the Sales team, metered in credits on Nagent's plans.

• Limitation: email only, and her pipeline does not sync with an outside CRM.

What should "waits for my approval" actually mean?

Many tools say they keep a human in the loop. Test what that means with these questions:

• Is approval on by default? A setting that starts off gets forgotten in the first busy week.

• Is it per email, or per campaign? Approving a template and then letting 500 personalised variants go out is not the same as approving each email.

• Do follow-ups wait too? Many tools gate the first touch and let follow-ups run on a timer.

• Is the exact text approved? If the AI rewrites after approval, you approved a different email.

• Is approval checked at the moment of sending? A draft approved last week for an account that has since unsubscribed should not go.

• Is every decision logged, with who approved what and when?

• Can you edit before approving, and does the tool learn from the edit or the rejection reason?

What should you check before approving each email?

A good review takes about a minute when the research is visible next to the draft:

1. Is the reason for writing true? Check the one specific fact against its source.

2. Is it still current? A hiring signal from three months ago is not a reason to write today.

3. Does it claim anything you cannot back? Customer names, results, figures.

4. Is this the right person at the account, and is anyone else on your team already talking to them?

5. Is there one clear ask, and is it easy to say yes to?

6. Does it sound like you? Tone, length, sign-off.

If you find yourself rewriting most drafts, stop approving and fix the inputs: the customer profile, the product description and the list of things the AI must never claim.

How do you keep approval from becoming a bottleneck?

Approval slows outbound only when drafts are poor or the queue is unmanaged.

• Fix research before volume. Good drafts take seconds to approve; bad ones take minutes to rewrite.

• Review in batches at a fixed time each day, rather than reacting to every new draft.

• Release approved emails inside set sending hours, so approval and sending are decoupled.

• Route by account value. Strategic accounts get a careful read; your process decides whether lower tiers ever run with less review.

• Widen autonomy only on the record. Once a type of email is approved almost unchanged for weeks, it is a candidate for less review. A person should make that call, not the tool.

How should approval relax over time?

The safest model treats autonomy like a promotion. Nagent's ladder has five levels: L1 Observe, L2 Assist, L3 Operate, L4 Represent and L5 Lead. Agents arrive at L2 Assist. The system can nominate an agent on its record, a person decides, and levels rise one step at a time. The autonomy and governance docs set out each level, and our explainer on earned autonomy for AI agents covers the idea.

For outbound specifically, NORA's sends stay held on a new workspace regardless of level, because a sent email is the one step that cannot be undone.

How does NORA's approval flow work, step by step?

This is where Nagent fits, so here is the flow in detail.

1. Brief her. Address her in the Sales room, which SERA, the AI Sales Chief of Staff, leads, or use Ask on her Discover screen. Describe who you want in a sentence.

2. She finds and scores. Each company gets a 0 to 100 fit score against your profile; contacts are revealed only at 70 or more.

3. She researches. A dossier per company: overview, why it fits, buying signals and pitch hooks, each with a source and date.

4. VANI drafts. One email per call: a subject of 60 characters or fewer, a preview line, three to five short paragraphs and one ask, plus a list of every claim made about the prospect. A claim not found in the research fails a check, and the draft is refused. Read more about VANI.

5. You review. Drafts wait on her Messages screen, filtered by Draft, Approved, Rejected and Sent. Read, edit, approve or reject.

6. Approval does not send. Approving moves the company to Ready to send. Sending is a separate step that waits for a person on a new workspace.

7. Release on your schedule. Broadcast sends only already-approved drafts inside the sending hours you set. Each send is checked against your suppression list at the moment it leaves.

8. Everything waiting is in one list. Her Pending screen gathers lead reviews, cost-cap exceptions and approvals, each linked to where it is decided.

A send is never offered from a room thread; it starts from her Messages screen or a workflow.

What NORA does not do: she does not call, send LinkedIn or SMS messages, close deals or sync her pipeline with an outside CRM. If you need multi-channel sequences with approval, Unify or Artisan may fit better.

Frequently Asked Questions

What is an AI outbound research agent?

An AI agent that finds accounts matching your customer profile, researches each into evidence such as signals and reasons to reach out, and drafts outreach from that research. The research step is what makes approval quick, because the approver can check each claim against its source.

Do follow-up emails also wait for approval?

It depends on the tool. Some gate only the first touch. In NORA's case, running the follow-up cadence is held for a person, just as the first send is.

Does requiring approval hurt reply rates?

Not in itself. Approval adds review time, not delay for the prospect, especially if approved emails are released inside set sending hours. Catching a wrong fact or a badly timed email protects both replies and your sending domain.

Can an AI outbound tool learn from my edits?

Some do. In Nagent, a person's Good answer or Not what I wanted in the room feeds the loop that moves NORA's trust. VANI keeps no memory between drafts and writes from what NORA hands it each time, so improve drafts by improving the research and knowledge she draws on.

Watch NORA draft and wait for your approval on your own accounts: book a demo, or read NORA's page.