How to run AI campaigns from CSV lists
How to turn a CSV of LinkedIn profile URLs into an outreach campaign: cleaning the list, writing the first messages and keeping the account safe.
Get startedShort answer
Upload a CSV of LinkedIn profile URLs, remove duplicates and broken links, then point a campaign at that list with a short invite, a first message and a follow-up. The URL is what identifies each person, so the other columns only need to be good enough to write a message from.
Preparing your CSV list
A CSV import is a file of LinkedIn profile URLs plus whatever context you want the first message to use. The URL column is the only one that has to be right, because it's how each person gets identified. Names, companies and notes are there to keep the message from sounding like a template.
Three things break most imports: the file is still .xlsx, the URL column contains something that isn't a profile link, or the same person appears twice under two different addresses.
Keep the first file small enough to read yourself. A hundred rows you have checked by eye tells you more about your message than five thousand you haven't.
Paste this job
profile_url,first_name,last_name,company,note https://www.linkedin.com/in/sarah-kim-northwind,Sarah,Kim,Northwind,Commented on the packaging post https://www.linkedin.com/in/marcus-hale-acme,Marcus,Hale,Acme,Spoke at the Midwest logistics meetup https://www.linkedin.com/in/priya-raman-fieldwork,Priya,Raman,Fieldwork,Second-degree through Dana
Cleaning the data before you upload
Duplicates cost you twice. The second invite fails, and the failed request sits in your pending pool, which makes the account look worse than it is.
Public profile links are the only kind that work. A link that opens a company page, a post or an internal directory record fails on import, and you usually find out because that row quietly disappears.
- 1
Remove duplicates
Sort the URL column in Sheets or Excel and delete repeats. Sort on the URL rather than the name, because two different people can share a name.
- 2
Check the URL shape
Filter the column so every row starts with https://www.linkedin.com/in/. Delete anything that doesn't.
- 3
Fill or drop missing names
A blank first name produces a message that opens with a comma. Find the name or delete the row.
- 4
Export as a UTF-8 CSV
Save as .csv rather than .xlsx. If the list has accented characters, export as UTF-8 so names don't arrive as symbols.
Writing the first two messages
A list of URLs is not a campaign. The campaign is the invite, the message after acceptance, and the follow-up you send when nobody answers.
The invite has one job, which is getting accepted. A line that shows you read the profile does that better than a line about your product. Two sentences is plenty.
The message after acceptance starts the conversation. Ask something the person can answer in one line, about their work rather than your offer.
Follow-ups should add something. A second message that says 'just bumping this' tells the person you had nothing else to say. A second message with a real question earns a reply.
Too aggressive
- Hi John, I see you work at Acme Corp. We have a tool that increases revenue by 20%. Do you have 15 minutes on Tuesday for a demo?
- Hello! I'm looking for partners in the logistics space. Check out my website and let me know if you want to collaborate.
Low pressure
- Hi John, I saw your post about supply chain shifts last week. I'm building a small network of logistics leaders and would like to connect.
- Hi John, I've been following Acme's growth in the Midwest. I'd like to keep up with your updates here.
Example
- You
Hi Sarah, I noticed your focus on sustainable packaging at Northwind. I'd like to connect and follow what you're working on.
- Prospect
Thanks for reaching out. Happy to connect.
- You
Thanks for accepting, Sarah. Are you still using plastic fillers for shipping, or have you moved to a compostable option?
Keeping the account out of trouble
LinkedIn doesn't publish exact limits on invites, messages or profile views, so anyone quoting a number is guessing. What you can control is whether the account behaves like the person who owns it.
The pattern that gets accounts restricted is a spike: nothing for three weeks, then a burst on a Monday morning. Steady volume inside working hours looks like someone with a job.
Find your own ceiling by starting below what you think you can send, adding a few actions each week while acceptance holds, and stopping when it drops. That ceiling belongs to your account and your list, and it moves.
- 1
Set daily caps
Pick a number you can hold every weekday. Hitting a bigger number once doesn't help you.
- 2
Set a send window
Send inside the lead's working hours. A message at 2am reads as a script even when the words are good.
- 3
Space the actions
Leave minutes between actions rather than seconds, since batch-sending is the tell.
- 4
Watch the pending pool
If invites pile up unaccepted, slow down and withdraw the old ones before you add more.
Withdraw invites that have sat unaccepted for weeks. A large pending pool tells LinkedIn your outreach is unwanted, and it makes your acceptance rate meaningless as a signal.
What to do with the replies
The list gets people into the sequence. Replies are where the list stops mattering and the conversation starts.
Sort by what the person said rather than which campaign they came from. Four kinds cover most of it, and each one needs a different next step.
The mistake is answering every reply with the same pitch. Someone who says 'not this quarter' and someone who asks about pricing need different answers, and treating them the same wastes the warmer one.
Interested
- Reply type
- Asks about price, timing or how it works
- What it sounds like
- Answer yourself within the hour if you can
Not now
- Reply type
- Says to come back next quarter
- What it sounds like
- Note the date and set a reminder instead of re-sequencing them
Wrong person
- Reply type
- Points you to a colleague
- What it sounds like
- Ask for the intro while the thread is still open
Objection
- Reply type
- Says they already have a vendor
- What it sounds like
- Answer it once, then stop
Sorting the list before you message anyone
A CSV will hold five thousand rows of people who will never buy. Reading the list against your buyer before you send costs an hour and saves a month of messages to the wrong title.
Write down the three or four things that make someone a real buyer for you: the title, the company size, the trigger that made them interesting this month. Then cut rows that miss two or more.
Cutting is the part people skip, because it feels like throwing away work. A short list where every row matches produces more conversations per hour than a long one where half the rows are wrong.
Title matches the person who signs off
- Signal
- Keep
- Keep or drop
- They can say yes without asking anyone
Title is a recruiter, student or competitor
- Signal
- Drop
- Keep or drop
- The message will never be relevant to them
Company is already a customer
- Signal
- Drop
- Keep or drop
- Outreach to an existing account reads as a mistake
Profile has no activity for years
- Signal
- Drop
- Keep or drop
- The account is dormant and no reply is coming
Doing this with Kinlead
Kinlead imports a CSV of LinkedIn profile URLs into a lead group, and any campaign can point at any group. Nothing else in the file is required.
Messages and follow-ups are written by AI from the campaign goal, tone and an optional sales playbook, which is your own script. Queued messages can be read and edited on the Actions page before they send, and any campaign can be paused.
Outreach runs from your own LinkedIn account through Unipile, a third party provider, and Kinlead never stores your password. Each account sends at most one action every five minutes, inside the daily limits and send window you set.
- Connection notes open with why that person became a lead and stay under 200 characters.
- Stale pending invitations are withdrawn automatically, and Settings shows your pending count and acceptance rate with warnings.
- Replies from every campaign land in one inbox, classified by intent, with interested replies first.
Frequently asked questions
Related
Get started with Kinlead
Prefer markdown? How to run AI campaigns from CSV lists.md