How to use an AI SDR for LinkedIn outbound
How to set up an AI SDR for LinkedIn outbound: define a buyer profile, write a playbook, set safe limits, and hand off replies at the right moment.
Get startedShort answer
An AI SDR for LinkedIn finds prospects that match your buyer profile, sends connection requests and first messages from your own account, and handles the early conversation using a playbook you write. You decide where it stops and a human takes over.
Define your buyer profile
An AI SDR only works if your instructions are tight. Tell it to find founders and it will message people who will never buy, which burns invites and drags your acceptance rate down. The tool isn't the problem in that case. The brief is.
Describe the buyer the way you'd describe them to a new hire: role, company type, company size, and something happening right now that makes them care. That last part does most of the work. A job change, a funding round, a public complaint about the tool they use today, a post asking for a recommendation.
Write the profile as one sentence you could read out loud, then test it against ten people you already know are good customers. If the sentence leaves them out, it's wrong. If it also matches people you'd never sell to, it's still too loose.
Too vague
- Founders of SaaS companies in the US
- Marketing managers
- B2B software buyers
Specific enough to search
- CEOs of B2B fintech startups with 10 to 50 employees who posted about payment processing this quarter
- Head of demand generation at Series B healthcare companies who follow two named industry analysts
- VPs of sales at logistics firms with 100 to 500 employees still running a legacy ERP
Build your sales playbook
The AI needs a framework to talk to people without sounding like a bot. A sales playbook is that framework: the goal, the tone, the one problem you solve, a qualifying question, and what to do when someone says no.
Don't lead with a pitch. The first few messages should confirm the prospect has the problem you solve before you describe the product at all. Most conversations die because the first message asked for a meeting instead of asking a question.
Pick a tone you'd actually use on the phone. A founder selling to other founders usually gets further with casual and direct than with formal corporate language, and the AI copies whatever you give it.
- 1
Set a clear campaign goal
Decide what success looks like before you write anything. For most teams that's a 15 minute discovery call on a calendar, not a reply.
- 2
Define the tone and persona
Direct and professional, or casual and peer to peer. Position the sender as someone who has seen the problem before, not as a salesperson working a list.
- 3
Map one pain point
Write the script around a single problem the prospect has. Listing five product features in message two gives them five reasons to stop reading.
- 4
Add a qualifying question
Draft a question that makes the lead confirm the problem, like asking how their team handles list building today. A yes tells you where to take the conversation next.
Set safety limits to protect your account
Automating LinkedIn carries risk. LinkedIn doesn't publish exact limits on invites, messages or profile views, so anyone quoting a safe daily number is guessing. Work from what your own account has tolerated and raise it slowly.
Spread activity across a window that matches when your prospects are online. Forty invites in one hour looks nothing like a person, even if forty in a day is fine for your account.
Don't let pending invitations pile up. Hundreds of sent requests nobody accepted makes your account look low quality to LinkedIn, and the next batch gets less reach. Withdraw the stale ones on a schedule instead of letting them sit.
If your account does get pushed back on, stop sending and wait it out. Adding volume to a flagged account is how a short restriction turns into a long one.
LinkedIn does not publish exact daily limits. Set your own from your account history, keep activity inside a send window, and withdraw stale pending invitations so the pool stays healthy.
Write connection notes that don't read like a pitch
The connection note has one job: give the person a reason to accept that isn't "I want to sell you something." Keep it under 200 characters so it reads on a phone without a tap.
Skip "I came across your profile." Name the actual trigger: the post they wrote, the company they just joined, the tool they complained about. One specific detail beats three sentences of flattery, and it proves a person looked.
Don't ask for a meeting in the note. The pitch belongs in the first message after they accept, and even there it should be one sentence about the problem rather than a feature list.
Example
- Suggested reply
Hi Sarah, your post on SDR ramp time stuck with me, especially the month reps spend building lists before they talk to anyone. I work on that problem with B2B teams. Worth connecting?
- Prospect
Thanks for reaching out. We're looking at this now actually. How does it work?
- Suggested reply
It takes list building and first touch off the rep, so the first month is conversations instead of spreadsheets. Do you have 15 minutes Tuesday?
- Prospect
Tuesday works. One thing though, we need it to write into HubSpot. Is that supported?
- You
Good question, and I'd rather answer it properly than guess. Tuesday at 2pm, I'll bring our integration notes.
Paste this job
Hi {first name}, your post about {topic} stuck with me, especially {one specific detail from it}. I work with {type of company} on {problem}. Worth connecting?
Keep it under 200 characters. One detail is enough. If you can't name a specific detail, that person isn't ready for a note yet.Manage the conversation hand-off
The AI shouldn't handle every word. There's a point where a human needs to step in, and deciding that point before the campaign starts is what keeps the hand-off clean.
AI is good at the repetitive parts: greetings, basic product questions, role checks, and sending your scheduling link once someone shows interest. It's bad at reading a room, and it's bad at telling a real buying signal from a polite question.
Hand off as soon as the prospect asks something that needs a strategic answer: custom pricing, a technical integration, a security review, or anything that sounds like a complaint about a past purchase. Those are the moments where a wrong sentence costs the deal.
Initial outreach
- Task
- Sending connection requests and first messages
- AI handles
- Personalizing notes to your ten highest-value targets
Qualification
- Task
- Answering basic product questions and confirming role
- AI handles
- Handling objections and deep pain points
Booking
- Task
- Sending the scheduling link when interest shows
- AI handles
- Confirming the invite and prepping the call
Follow-up
- Task
- Gentle reminders to a first message that got no reply
- AI handles
- Re-engaging someone who went quiet after booking
Doing this with Kinlead
Kinlead combines lead discovery and AI outreach in one place. Search for people who match the profile you described, or point the Steal Customers agent at competitor company pages and let it turn commenters on their recent posts into leads. Each lead keeps the post URL, what the post was about and what they commented.
Outreach runs from your own LinkedIn account through Unipile, a third-party provider, so Kinlead never stores your password. You set daily invite and message limits and a send window in the lead's time zone, and the account sends at most one action every 5 minutes.
You write the goal and the playbook. Each campaign chooses how replies are handled: hand the conversation to you at the first reply, let the AI reply until there's real interest, or let the AI reply until a meeting is booked with your scheduling link. Replies from every campaign land in one inbox, interested ones first.
Lead search runs on LinkedIn's standard search, so a Sales Navigator seat isn't required. Settings shows pending invitations and acceptance rate with warnings, and stale pending invitations are withdrawn automatically. If LinkedIn pushes back on an account, invites pause on that account and you get an email.
- 1
Find and score leads
Use lead search or the Steal Customers agent. Kinlead scores each lead from 0 to 100 against your profile and drops anything below the fit line.
- 2
Configure safety and timing
Set daily invite and message limits and pick a send window. Pending invitations and acceptance rate are visible in Settings, with warnings when the pool gets unhealthy.
- 3
Launch the campaign
Enter the goal, tone and your own sales playbook. Connection notes come out under 200 characters and open with why that person became a lead.
- 4
Work the inbox
Replies are classified by intent with interested ones first, and you get an email when someone replies. Queued messages can be read and edited on the Actions page before they send, and any campaign can be paused.
Frequently asked questions
Related
Get started with Kinlead
Prefer markdown? How to use an AI SDR for LinkedIn outbound.md