AI Outbound Playbooks: Sequences Worth Copying
Last updated on .
The play, step by step
6 moves, in order. Tick each one as you do it; the line under a move is the detail, and “Done when” is how you know it is done.
Choose the event that makes a message timely, such as a new hire in the buying team, a funding announcement or a job post, and write the rule that decides who qualifies.
Done whenYou can say in a sentence why this contact gets a message now.
Keep the list to accounts that fit your ideal customer and show the signal. AI can sort and flag rows, but a person approves the list before anything is drafted.
Done whenEvery row carries a reason you can read aloud.
Pass the model the event, the date it happened and who the contact is, and ask for a brief opener that starts from that event. Leave the wording of the offer and the call to action to a person.
Done whenThe opener names the event and its date, and nothing the signal did not say.
Read each draft, or a sample of each batch once the openers hold up, and check every fact it states against the source of the signal. Delete anything the model guessed.
Done whenEvery claim in a sent message traces to the signal or to your own offer.
Add follow-ups that each bring something new: a relevant example, a short question or a note on another channel. Stop the sequence for anyone who replies and route that reply to a person.
Done whenA reply stops the sequence and lands with a named person.
File each reply under the signal and the follow-up that drew it, then alter one element from one sequence to the next so the result has a single explanation.
Done whenEach reply in your log is filed under the signal that triggered the message.
Tick each step as you do it.
Keep going: the steps left are the ones people skip.
Every step ticked. Now pick your route below.
Which situation is yours?
3 situations. Pick the line that describes you: the page brings forward the route that fits, with what to do and what changes for you.
Small team, hand-sent
What to doStep 4 becomes a read of every draft, and step 5 is spaced by hand. Skip automating the send until the openers hold up under your own review.
Signal-driven at scale
What to doStep 2 needs strict rules for who qualifies, and step 4 moves from reading every draft to sampling each batch and holding the batch when a sample fails.
Few large accounts
What to doStep 1 names an account event instead of a person event, step 3 yields research notes rather than a finished opener, and step 5 leans on calls and social notes.
Every route is written on this page; picking one only brings it forward.
Where it goes wrong
4 moves that look right and cost the most. Each card names one, what it costs, and what to do instead.
Sending AI drafts without reading them
What it costsA model can fill gaps with plausible detail, such as a funding round that never happened, and one wrong line can lead the prospect to discard the sequence.
Do insteadGive the send step an owner: a named person marks each batch as read, and nothing is scheduled before that mark.
Personalizing only the opening line
What it costsWhen only the first line is specific, the paragraph after it can give the game away, and the prospect may read every follow-up as part of the same mass mailing.
Do insteadTie the offer to the same event as the opener, and cut any paragraph that would read the same without that event.
Letting the sequence run after a reply
What it costsA follow-up that lands after the prospect already answered suggests nobody is reading replies, and it can turn a warm conversation into an annoyance.
Do insteadCheck the inbox for replies before each scheduled send, and hand the thread to its owner the same day.
Changing several things between sequences
What it costsWhen the signal, the opener and the channel all change together, a better or worse result cannot be traced back to any of them.
Do insteadKeep a dated note for each sequence with its signal, its opener and its channel, so a later result points to one difference.
Related questions
A timely signal, a target list that matches it, an opener the AI drafts from that signal, a person who reviews it, follow-ups that each add something new, and a way to read replies and adjust.
AI suits drafting, sorting rows and summarizing research. A person keeps the choice of who to contact, the offer, the review before sending and the handling of every reply, because those carry the judgment.
The event decides who enters the sequence and when, not a calendar date, and each follow-up refers back to it. A list-based sequence sends to everyone on a schedule, whatever happened at the account.
It depends on how fast the signal goes stale, how much the account is worth and how the buyer reads each channel. Add a touch only when it brings something new, and stop when it repeats the last.
No fixed figure fits every list, because the rate depends on the audience, the offer and the channel. Set your own reference from past sequences, then compare each AI-assisted sequence against it.
Make every message answer why now: the event that triggered it comes first, each claim after it is one you can source, and a person words the offer. A message with no event to cite is a template.
This page gives a sequence; it does not score tools.