Intent-based outreach means you only message people who just showed a buying signal, and your first line references that exact signal so it cannot read as generic. You wait for a comment, a like, a job post, or a complaint, then join a conversation the person already started in their own head. Same effort as cold, a completely different reply rate.
Cold outreach fails for one structural reason: it interrupts strangers who never asked. You buy a list, you spray it, and you get one or two replies in a hundred, most of them asking who you are.
Intent-based outreach flips the order. Instead of convincing a stranger to care, you show up for someone who already does. Here is exactly what that changes about the message itself:
Cold: "Hi John, I help B2B founders book more meetings with AI-powered outreach. Open to a quick 15 minute call this week?"
Signal-anchored: "Hey John, you said cold outreach is dead under that post and honestly, same. Curious what's actually working for you now, referrals, inbound, something else?"
Same person, same day. The second one references the exact thing they just did and asks one real question. That is the whole difference between the message that gets ignored and the one that gets a reply.
The rest of this is the system behind that second message on LinkedIn: what the signals are, how to write it so it sounds human, and how to run the whole thing by hand this week before you automate anything. For the version that runs across every channel, see Intent-Based Outreach.
What is intent-based outreach?
Intent-based outreach (also called signal-based outreach) is the practice of triggering a message off an observed action rather than a static list. The action is the intent: someone engaged with a post about the problem you solve, a company posted a role that implies your pain, or a person published something this week about the exact thing you fix. You reach out referencing that action, which means your first message is specific by construction. It cannot feel like a template because it is anchored to something only that person did.
The contrast with cold outreach is not tone, it is timing and targeting. Cold treats everyone the same and hopes volume finds the few who happen to care today. Intent-based waits for the person to raise their hand first, so you are early to a need that already exists instead of manufacturing one.
What counts as a buying signal?
The strongest signals are the ones where someone volunteered attention on the exact problem you solve. Ranked easiest to hardest to act on:
| Signal | Why it's intent | How to pull it |
|---|---|---|
| Post engagement | Anyone who liked or commented on a post about your problem (yours or a competitor's) just self-identified | Scrape the likers and commenters off the post, no login required |
| Hiring signal | A company hiring three SDRs cares about outbound right now; the job post is the intent | Watch job boards for roles that imply your pain |
| Recent-post signal | Someone who posted this week about the exact problem you fix is thinking about it today | Track keyword posts from your ICP |
Start with one source. Post engagement is the simplest: pick a post in your space with real comments, open the reactions, and there is your list of warm leads, already sorted by the fact that they engaged.
Target companies, not just people
A single person's like is a weak signal on its own. Three people from the same company engaging with posts about your problem is a strong one.
Roll individual signals up to the account level and you stop chasing one commenter and start reading a whole buying committee that is warming up at once.
How do you scrape LinkedIn signals without wrecking your account?
This is where most people quietly break their own LinkedIn. Tools that automate actions through your logged-in account (auto-connect, auto-DM bots that ride your session cookie) put your profile at real risk of restriction.
The safer pattern splits the work in two: pull the signal data with a no-cookie scraper that never touches your session, then send the actual messages by hand or through official channels. The data step and the send step never share a login.
Practically: an Apify actor can pull every commenter and every reaction off a public post without logging in as you. Your account stays clean, and you still get the exact list of people who raised their hand. The blueprint below names the specific actors and the Claude Code prompt that turns a post URL into a ready lead list.
How do you write the DM so it doesn't read like AI?
The reason most AI-written DMs read as AI is not the grammar. It is that they are generic enough to send to anyone. The fix is a single rule: anchor every message to the one specific signal, then say the least possible around it. Specific, plus short, plus one real question, equals human.
Here is the principle as a prompt you can run today:
Write a short, warm LinkedIn DM that reads like a real, curious person. Open by referencing the exact signal (their comment / job post / recent post). Mirror their own words if it's a quote. Ask exactly one genuine question. Pitch nothing, no calendar link. No em dashes, max four short lines.
Feed it a real signal and you get something a friend might send, not a rep: a line about the exact thing they said, one honest question, and nothing else. The pitch is not in the first message. It is not in the second one either.
The four-stage engine
Every intent-based system, manual or automated, runs the same four stages in order. The value compounds because each stage feeds the next with something specific.
- Signal. Find people already raising their hand. One source to start (post engagement).
- Draft. Write the DM anchored to that person's exact signal, with Claude doing the wording and you doing the judgment.
- Send. Lead with a real question about their situation. Give first, pitch never in message one.
- Follow up. Keep the loop open without pressure. One-word nudges, a wave GIF, the occasional meme. If a follow-up would annoy a friend, it will annoy a prospect.
That last stage is where cold-outreach habits do the most damage. A warm lead does not need a five-paragraph "circling back." They need a light touch that reads like a person who genuinely forgot to reply, because half the time that is exactly what happened.
No paragraphs. Each nudge is short enough that even a no comes back with a laugh:
Day 2: a wave GIF, or just "👋"
Day 5: one word, "worth it?"
Day 9: a light meme, "most people's outreach rn 😂 you doing better?"
Then: re-give something useful, then a friendly exit that leaves the door open.
If a follow-up would annoy a friend, it will annoy a prospect. That single test kills 90% of bad follow-ups before you send them.
Should you stack other channels on a LinkedIn signal?
For high-value accounts, yes. The signal is channel-agnostic: once you know a person engaged, the same signal can carry a short personalized email and, for the accounts worth it, an AI voice touch, all referencing the exact thing they did. That full multi-channel build is its own topic. This piece stays on LinkedIn, where most people should start. For the system that runs one signal across every channel, read Intent-Based Outreach.
How do you start this week without any tooling?
You do not need automation to get the first win. The manual version is the proof, and it takes an afternoon:
- Pick one post in your space with real engagement.
- Open the likes and comments, write down 20 people who fit who you help.
- For each, grab the signal (their exact comment, or the post they engaged with).
- Run the Claude prompt with that signal, send the message by hand, one real question, no pitch.
- Follow up light using the cadence above.
Twenty signal-anchored messages beat two hundred cold ones. Do it once by hand and the reply rate tells you everything. The automation only matters after the manual version already works for you, because automating a bad message just sends it faster.
FAQ
Is intent-based outreach the same as buying intent data?
No. Purchased intent data (third-party "these accounts are in-market" feeds) is a probabilistic guess. Intent-based outreach here means a first-party observed action: this specific person commented on this specific post. It is a signal you can see and reference directly, which is why the message lands warm instead of vaguely targeted.
Do I need a paid LinkedIn or Sales Navigator account?
No. You can run the entire manual system on a free LinkedIn account plus Claude. The no-cookie scraper reads public post engagement without your login, and the messages go out by hand. Paid tooling only changes the volume, not whether the method works.
Will scraping engagement get my LinkedIn restricted?
Not if you keep the scraping off your own session. The risk comes from bots that automate actions through your logged-in account. Pulling signal data with a no-cookie scraper and sending messages manually keeps your profile clean, which is the whole reason to separate the data step from the send step.
How is this different from just buying an AI sales agent?
An AI sales agent handles the outbound step. Intent-based outreach is the targeting logic that feeds it: without a real signal to anchor to, an agent just personalizes cold messages faster. The signal is what makes the personalization true instead of plausible.
The full playbook, the three DM scripts, the exact Claude prompt, the specific Apify actors, and the follow-up cadence live in the Intent-Based LinkedIn Outreach guide. Run it by hand this week. If you'd rather have the whole thing surfacing warm leads and drafting the messages for you every morning, book a 30 minute audit and we'll map it against your funnel.