Approve the policy, not every outbound message

The worry about automated outbound is simple. You hand the sending to software, and one morning it is emailing the wrong people, too often, in a tone you would never use.

Two controls answer that worry. One is a person approving individual messages. The other is a person approving the policy that governs every message. They are not the same thing, and they fail in different ways.

Two kinds of approval

Some tools put review at the message level. 11x says every message can be reviewed, edited or overridden before it goes out, or it runs on autopilot [[1]](#sources). Artisan's page describes an "Approval mode, or let Ava send on its own" [[2]](#sources). In both cases review is a setting that can be switched off.

Policy approval puts the check one level up. Outlevo sends nothing outside a policy that a person approved [[3]](#sources). A policy sets who can be contacted, on which channels, how often and with what budget [[3]](#sources). Every external action is checked against the approved policy before it runs, and it is recorded in a ledger [[3]](#sources). Kill switches pause a campaign, a channel or the whole workspace at once [[3]](#sources).

A message review tells you whether one email reads well. A policy review tells you whether the right people are being contacted at all, and how many of them.

This does not mean nobody ever reads a message. Take interested replies. In Outlevo's Calendly integration, on autopilot the answer to an interested reply goes at once. Otherwise it waits in Replies for you to send [[4]](#sources). The choice between those two modes belongs to your team. The policy is the outer boundary, and your team decides where inside it a human reads each message.

What changed

Receiving mail systems are tying volume to authentication. Microsoft says: "For domains sending over 5,000 emails per day, Outlook will soon require compliance with SPF, DKIM, DMARC." [[5]](#sources) That wording was checked on 2026-10-22, and "soon" was Microsoft's word on that day, so check the page for the current status. Volume and authentication now belong in the same review, and that review should happen before sending starts.

So the question to ask a vendor is not only "can I review every message?" It is also "what limits are set before anything sends?" If every action is checked against a policy a person approved, the limits are the contacts, channels, frequency and budget written into that policy [[3]](#sources).

That moves the work. Your reviewer spends less time proofreading and more time deciding limits. The decisions are fewer, and they matter more.

What a reviewer should check

Treat the policy review as a short checklist. Each item below ties to something the system shows you.

Who gets contacted

Targeting is the first thing to read. AI-proposed ICP segments are labelled experimental until a person approves them [[6]](#sources). That label is a checkpoint: an experimental segment should not drive a send until you have looked at it.

Check how much evidence stands behind a proposed segment. Win/loss analysis counts companies, not deals, and it labels small samples. Under 10 won companies is very low confidence [[6]](#sources). If a segment rests on a sample that small, treat it as a guess.

Countries are yours to name. Outlevo asks which countries a customer sells into and never proposes markets [[6]](#sources).

Then check the accounts themselves. Each qualified account gets a Readiness Card with fit, intent and data quality scores [[7]](#sources). A card needs at least three observations, each with a source and a date [[7]](#sources). Facts read from a website are kept only when their quote is found on the page, and guesses are labelled as hypotheses, never presented as findings [[7]](#sources). Open a few cards. If the observations are thin or the dates are old, the account should not be in the policy.

Which channels, and under whose rules

Each channel has rules that sit outside any tool. LinkedIn's User Agreement tells members not to "Use bots or other unauthorized automated methods to access the Services, add or download contacts, send or redirect messages" [[8]](#sources). Read that before you turn on automation for a LinkedIn account, and read our page on LinkedIn automation for outbound, safely.

Here is how LinkedIn sending is set up. Outlevo sends the invitations and follow-ups of the campaigns where you turn it on, within daily limits [[9]](#sources). It sends a connection invitation with a note of at most 300 characters, then one message after the person accepts, and a reply stops that lead's emails [[9]](#sources). Automatic sending is included on Growth and Managed. On other plans the invitations and messages are drafted for you to send by hand [[9]](#sources).

Email has its own law. The FTC says of CAN-SPAM: "The law makes no exception for business-to-business email." [[10]](#sources) Your reviewer should not assume that B2B outreach sits outside the rules.

How often, and how much

Frequency is one of the four things a policy sets [[3]](#sources), and your reviewer can set it with a number. Pick the number yourself and write it into the policy. A policy that says "how often" without a figure is not a limit.

Some brakes are built in, and a reviewer should know them. Before a Smartlead campaign sends, every mailbox must be connected and warmed up, and its domain must have SPF, DKIM and DMARC checked in the last 7 days [[11]](#sources). A new managed mailbox sends nothing until it is warm: at least 14 days, usually about 3 weeks [[12]](#sources).

What it can spend and write

Budgets and writes need the same treatment. In the ad integrations, nothing spends until an owner or marketing approver sets the caps, and a kill switch stops all ads at once. Reddit Ads is one example [[13]](#sources). The HubSpot integration works the same way for data: nothing is written until an owner or admin approves the field mapping, and it writes only to Outlevo-owned properties [[14]](#sources).

Where tone fits

Be clear about what a policy does not do. The approved policy sets who, which channel, how often and what budget [[3]](#sources). It is not a proofreading step.

Tone stays under your control through where you place the human send. Replies can wait in Replies for you to send, instead of going out on autopilot [[4]](#sources). On plans without automatic LinkedIn sending, messages are drafted for you to send by hand [[9]](#sources). The Gmail extension never sends from Gmail. It drafts a reply for you to send, and a recipient guard warns before you email a contact Outlevo should not reach [[15]](#sources).

So the split is plain. The policy fixes the limits. You choose which messages a person reads before they go.

What to do

Write the limits down before anything sends. For each campaign, state who can be contacted, on which channels, how often and with what budget [[3]](#sources). Use numbers where the policy asks for them.
Read the targeting evidence. Leave experimental segments out until a person has approved them [[6]](#sources). Open several Readiness Cards and check that each has its sources and dates [[7]](#sources).
Check each channel's rules. Read LinkedIn's terms [[8]](#sources) and remember that CAN-SPAM covers B2B email [[10]](#sources).
Confirm the sending setup. On Smartlead, check mailbox warm-up and the authentication records before the first send [[11]](#sources). On managed mailboxes, remember that a new mailbox sends nothing for at least 14 days while it warms up [[12]](#sources).
Decide which messages a person reads. Choose between autopilot and review for replies [[4]](#sources).
Test the kill switch. Know whether you can pause a campaign, a channel or the whole workspace, and who is allowed to do it [[3]](#sources).
Plan to audit afterward. Every external action is recorded in a ledger [[3]](#sources). The public API exposes resources including policies, ledger and suppressions, so the approved policies and the record of actions can be reached from code [[16]](#sources).

If you want to see how the targeting side works before you write a policy, run the free ICP assessment. Our how it works page covers the approval model in more detail, and the sequence glossary entry defines the term.

Automated outbound is not safe or unsafe by itself. It depends on the limits a person set before it started.