Can and should are not the same question.

Just think Jurassic Park. Jeff Goldblum.

“You were so preoccupied with whether or not they could, that they didn't stop to think if they should."

This is exactly how we should be considering AI and Agentic enablement in your GTM efforts.

Why? Well, open LinkedIn on any given day and you can watch it happening in real time.

A post goes up, and within minutes it's picked up comments from accounts that have never engaged with anything else you've written.

Vaguely on-topic. Weirdly generic. None of them respond to what you actually said, because none of them read it.

A connection request follows a few hours later, mentioning your industry once and your name never. Somewhere behind it, an agent matched a few keywords and pulled the trigger.

This is what a lot of go-to-market automation looks like now

Tools that scan LinkedIn for content matching a handful of keywords, generate a comment, and fire off a connection request, without a person ever seeing the post.

Is it clever? Totally. It’s really clever.

Does it help your efforts? No.

It's sold as efficiency. Engagement at scale, with none of the daily fifteen minutes it used to cost you.

Agentic tools can support a sales engine. We build engines for a living, and the right tooling makes an engine run better, with more scale and efficiency.

But there's a difference between a tool that supports how you engage and a tool that replaces the decision about whether you should engage at all. That second thing is what's flooding LinkedIn right now. It's why so much of what passes for engagement reads like nobody was actually there.

The keyword is doing the thinking, not you

Keyword matching finds content. It doesn't find context

It can tell you a post mentions "pipeline" or "growth" or whatever term you've told it to hunt for.

It can’t tell you whether the person who wrote it is someone worth talking to, whether the post was a genuine question or a rant, or whether this is the fortieth message they've had this week that opens the same way.

That judgement was always the actual work. The keyword match was never a shortcut to it. It was a way of skipping it.

A lot of this comes from the same place bad outreach always comes from: the belief that more volume beats better judgement. It doesn't. Chasing every match on a keyword list is cold outreach wearing an agent's badge.

It stands out for exactly the wrong reasons

We've always told people that short, generic comments read as bot behaviour on LinkedIn. That was true before agents existed. Now the agents are the reason it's true at scale. Five words of vague agreement, dropped under a post nobody on the other end actually read.

Prospects have seen enough of this now to spot the pattern by the first sentence. A comment that doesn't respond to what you said doesn't feel like interest. It feels like being processed. A connection request that could have been sent to anyone with your job title doesn't feel like an invitation. It feels like being added to a list.

That's the part that gets missed in most of the advice about scaling engagement. It was never really about the writing quality. Some of these comments are perfectly grammatical. It's that they arrive with no evidence anyone considered what the person receiving them would feel.

Can and should are not the same question

Every one of these tools exists because it's technically possible. An agent can be built to scan a feed, match keywords, draft a comment, and send it, all without anyone approving the individual message. The question that gets skipped is whether it should.

You can automate the search. You can automate the drafting. You cannot automate the decision about whether this is a person worth your time, or what they'll make of hearing from you.

That decision has always been the human part of the job. It's the part that makes the difference between engagement that builds a relationship and engagement that just adds a number to a dashboard.

Where the line actually sits

None of this means stepping back from agentic tools. It means being precise about which part of the job you hand over.

Let’s wrap this up

Let AI do the reading you don't have time to do.

Have it flag who from your target list has gone quiet after months of activity, or who's suddenly engaging with a competitor's content. That's pattern recognition across more accounts than you could scan yourself, and it's genuinely useful.

What it shouldn't do is decide, unsupervised, that a match on paper is reason enough to comment or connect. A person still needs to read the post, decide if they've actually got something to say, and write the reply themselves.

If there's nothing to say, the answer is to say nothing. That's not the tool failing. That's the tool working exactly as it should.

Build the engine so it can scale. Just don't hand it the one decision that was never about scale in the first place. Whether to engage with someone was always a judgement about a person, not a match on a keyword, and no amount of automation changes what the person on the other end feels when they realise which one they got.