Most “AI personalisation” in LinkedIn outreach writes the first message, and that’s it. Once someone replies, you’re back to a person staring at an inbox.
That’s the wrong half to automate. The opener is the cheap part. The replies are where the time goes: reading what came in, deciding whether each person is a real opportunity, drafting an answer that fits, and getting a time on the table. It’s also where deals are won and lost.
In LinkedReach, the agent that warms people up also handles their replies. Here’s how it works.
What it does, end to end
When someone replies, to any of your LinkedIn accounts on any campaign, it takes four steps before the reply reaches your inbox:
- It reads the reply. In context: who sent the first message and what it said, the person’s profile, and what the campaign is for. Then it labels the reply: interested, a question, an objection, not now, not interested, out of office, or pointing you to someone else.
- It judges how real it is. Is this a genuine, interested reply from someone who fits what you sell, or does it only look like one?
- It drafts an answer, in your voice. Fitted to what they said. Someone who asks to talk gets a short answer with times in it. An objection gets an answer to that objection. Someone who isn’t the right person gets a polite ask for who is.
- It offers a time. Where it fits, it looks at your calendar and puts two or three times in the draft, only times you’re free.
You see all of it in your inbox: the label, the draft and the times. Approve the draft as it is, change it and send, or write your own. Or give it more room: on Autopilot it answers on its own, and pricing, contracts, complaints and opt-outs still come to you.
An example.
Dana Reyes · 23:12
Dana: Thanks for the note on handovers. Is this something you could show me properly?
Your answer, drafted: Happy to. I have Thursday at 9:30 or 14:00, fifteen minutes.
Ready at 23:14, in your voice. Waiting for your yes.
When it acts on its own
This is the part that took us longest to get right, because your name is on every message. So you decide how much it does on its own, and it starts with very little.
- Every new campaign starts in Control: every comment, invitation and message it drafts, replies included, waits for your yes.
- In Assisted, it sends the routine steps itself and asks before the ones that matter.
- On Autopilot, it acts on its own: it answers, and it books the meeting. It still stops for pricing, contracts, legal questions, complaints and anyone who asks to be removed. Those don’t get an automatic answer. They come back to you.
We suggest Autopilot once you’ve stopped changing its drafts, and not before.
How much it does on its ownExample
Control, chosenIt writes, you press send.
AssistedSends the straightforward things itself.
AutopilotIt acts on its own, and still stops for legal, pricing and opt-out.
Times, not a booking link
Most scheduling stops at a booking link: here it is, pick a slot. That isn’t enough. The friction is that people don’t want to click a link, face a wall of slots and choose one. They want to be asked, “How about Tuesday at 11 or Wednesday at 3?”, and say yes.
So it reads your calendar and puts two or three real times in the reply itself. It only offers times you’re free, and holds them so nobody else is offered the same slot. If they pick one, it books the meeting and creates the Google Meet link. If they suggest another time, it checks your calendar before it answers.
This sounds small. It isn’t. People say yes to a specific question. They don’t always click links.
An example.
Dana Reyes
Dana: Thursday at 9:30 works.
Meeting bookedThursday 9:30–9:45 · Google Meet
Booked with one tap at 08:05.
What it’s not good at
An honest list. It drafts well for the common replies: interest, timing, fit, referrals and the wrong person. Pricing doesn’t get an automatic answer. It comes to you. And there are replies it doesn’t handle well:
- New, tangled objections that combine several concerns: “we already use X, and the budget is committed for Q3, but come back in October if your roadmap covers Y”.
- Replies in a language your campaign doesn’t write in.
- Highly technical questions that need deep knowledge of your product.
- Anything that leans on history it can’t see: a long relationship, or a thread that refers to a conversation it wasn’t part of.
For all of those, it leaves the reply to you. It knows what it doesn’t know, and that matters more than anything else it does.
What it’s worth to you
We won’t quote you reply rates from someone else’s market. Yours is the one that counts. So every week a plain-English briefing tells you what happened: the replies with interest, the meetings booked, and what they could be worth at the figure you set for a customer. Each of your posts shows the replies and meetings it led to, and you can share a live report that opens without a login.
This week’s briefingExample
The best thing that happened: Dana Reyes booked a call for Thursday.
Replies with interest6
Meetings booked2
What they could be worthup to £10,000
At about £5,000 a customer, the figure you set. Your post on shift handovers led to 3 replies and 1 meeting.
Where this goes
The next layer is memory across your senders, the LinkedIn accounts you send from. Say someone connects with one of them and never replies, and two months later their colleague answers a different one. The agent should see that both work at the same company, bring in what happened before, and answer with it in mind.
That’s a different kind of intelligence: it understands relationships, not just conversations. It’s the work ahead of us.
Back to 23:12. This time, when they write back, the answer is ready by 23:14, in your voice, with two times you’re actually free. You read it over coffee, change nothing and tap yes. By 08:05 the meeting is booked. Nobody was chased, and nothing went out that you wouldn’t have sent yourself.