The LinkedIn outbound tool we ended up building ourselves

We didn't set out to build a SaaS. We set out to deliver outbound for our clients at a quality the off-the-shelf tools couldn't reach, and after a year of writing workarounds and wrappers around platforms that almost did what we needed, we looked up and realised we'd built our own.

The portal we now run our agency on started as a folder of scripts patching gaps in tools we were already using. It became a system. It became the thing clients now log into to see their pipeline, edit their messages, and approve what goes out in their name.

There's a lot to write about it, and I'll get to most of it over the coming weeks. Today I want to focus on one piece - because it's the piece that took the longest to get right, and the one where the rest of the market has, I think, quietly given up: automation that doesn't read as automation.

What every outbound tool actually does

If you've used LinkedIn outreach tools, you know the shape. You build a list. You write a sequence with a few variables in it - first name, company, maybe job title - and you load it. The tool fires the sequence on a schedule. The same template, lightly substituted, lands in every inbox.

It works mechanically. It doesn't work editorially.

Even with the variables filled in, the read is templated. Your buyer can feel it. The reply rates show it. The category collectively decided that genuine personalisation was too hard to scale, so it scaled volume instead - and the inboxes of every senior buyer in professional services have been paying the bill ever since.

Why we built our own

We tried to live inside those tools for a long time. We tried writing better templates. We tried more variables. We tried smarter list segmentation. The ceiling was always the same: the message sounded like a message that had been sent to a thousand other people, because mechanically that's what it was.

So the portal grew out of all the workarounds. It now sits at the centre of our delivery, and it does something the off-the-shelf tools don't: it treats the moment of sending as the part that matters most, not the part to automate away.

If you want to understand why the underlying strategy matters as much as the tooling, the article on why most LinkedIn users never find the right leads covers the lead-identification side that has to come first.

How the personalisation actually works

There are four steps.

1. We engage before we connect

Before a connection request goes anywhere, we engage with the lead's recent posts - if they have any. A genuine like, a thoughtful comment, a small signal that we've read their work. This does two things at once: it lifts acceptance because LinkedIn reads the relationship as real, and it tells the lead that the person about to introduce themselves has actually looked at their work.

2. The connection request lands warm

By the time the request arrives, the lead has seen the name once or twice on their own posts. Acceptance rates lift noticeably - nothing magical, just the basic courtesy of showing up before you ask for time.

3. The first DM is written from scratch

This is the part most tools skip. When the connection is accepted, our system drafts the first message by reading the lead's profile and recent activity, the client's persona and voice, the client's ICP and offer, and what makes the client credible to this specific person.

The output isn't a template with a first name dropped in. It's a message that reads like the client wrote it themselves - because they did. The system just spared them the hour at the keyboard.

4. The AscendAI team reviews every message - and the client can too

The drafts come out strong and genuinely personalised, but they don't go out unread. The AscendAI team reviews every message before it sends. For our more active clients, the portal also lets them step in and edit the draft themselves before it goes out under their name. There is a human in the loop on every outbound message, every time. That's the line we don't cross.

The follow-up question

A second DM goes out automatically if there's no reply. Just one - not two, not three.

We've now run this across our clients for long enough to have a clear answer to a question the rest of the industry seems to disagree on. In professional and financial services, a two-message sequence outperforms a three-message sequence. The market is saturated. Persistence past two touches reads as spam, and the people you actually want to reach treat it that way.

The second message itself stays deliberately simple. Something close to: "reaching out again in case the first message got buried - happy to chat, or point me to whoever's better placed." No pressure, no stacked questions, an offered exit. The senior people we want to talk to recognise the tone immediately, and the ones who were going to reply do.

For the content side of the equation - turning a post's reach into a conversation - the piece on why LinkedIn is still your number two channel covers how outbound and content work together in practice.

What this is really about

Volume is cheap, and it's getting cheaper. The tools that scale spam will keep getting better at scaling spam, and the inboxes will keep getting noisier.

The edge isn't more sends. It's making every send feel like the senior person actually sent it. Automation isn't the enemy of personalisation - the way most tools use it is.

You can see the outcome of that thinking in our Vuna case study, where personalised outbound combined with content drove a consistent stream of qualified meetings from a standing start.

The bottom line

Engage before you ask. Have the system draft from real context, not from a template. Keep a human in the loop on every send. Cap the sequence at two messages.

That's the loop we built the portal to run, and it's the reason our reply rates from senior buyers look the way they do.

Frequently asked questions

Why don't off-the-shelf LinkedIn outbound tools work for professional services?

Most outbound tools scale volume rather than quality. They fire the same template - lightly substituted with first name and company - to every person on a list. Senior buyers in professional services can feel the template, reply rates show it, and the category collectively decided genuine personalisation was too hard to scale. The result is that every inbox is noisier and the messages that do land unread.

How does a properly personalised LinkedIn DM actually get written?

The first message should be drafted from real context - the lead's profile and recent activity, the sender's voice and persona, the client's ICP and offer, and what makes the sender credible to this specific person. The output should read like the sender wrote it themselves. It should not be a template with a first name dropped in.

How many follow-up messages should a LinkedIn outbound sequence have?

In professional and financial services, a two-message sequence outperforms a three-message sequence. The market is saturated and persistence past two touches reads as spam to the senior buyers you actually want to reach. The second message should stay simple and offer an easy exit - no stacked questions, no pressure.