A vague job title is never treated as proof of fit. Unclear means skip, and skip is final.
· Updated · 2 min read
Most automation tools contact anyone who matches a keyword search. Ours does the opposite by default: it looks for a reason to skip before it looks for a reason to send. The rule that does it is the role fence. It runs before anyone is contacted, on every channel, in whatever language the profile is written in.
The fence works from definitions you give it once. You define your ICP and, for each product, the buyer groups that actually buy it (operations, production, quality, IT, finance, C-level), together with the industries and countries you sell to. The AI recognises the role behind a job title, in any language, and matches it to the right group before anyone is contacted.
Titles like “Manager” or “Director” are common, vague and present on thousands of unrelated profiles, so they are never treated as sufficient evidence of fit on their own. The AI reads the job title together with the company context before deciding, because a title only means something next to the company it belongs to.
When the fit genuinely can't be determined, the system skips the profile. The skip is final: there is no second pass later, hoping for a different answer from the same evidence. A smaller, more relevant list consistently outperforms a bigger, noisier one, and it protects your sender reputation and the LinkedIn seat doing the work.
Passing the fence is not enough on its own. For every prospect the model must return a structured decision (which product, which persona, which angle) before the message itself. Bare prose without that decision fails closed: no decision, no send. That is what stops the AI from improvising its way around the rule.
On LinkedIn the fence has a second job. Before the AI messages someone who opened your email, it checks on Sales Navigator that it is really the same person: same name, same company, the role you target. An uncertain match is never messaged: writing to the wrong person with the right name is worse than not writing at all.
Fabricated meeting commitments, unresolved placeholders and prohibited links are stripped from every draft before it leaves. If both the primary and fallback model fail to produce a clean draft, the message is held. A generic template is never sent as a substitute.
Fewer messages, by design. A package's target companies (500, 1,500 or 3,000 a month in Starter, Pro and Premium) are not guaranteed messages. How many of them can actually be reached depends on market size, how strict your ICP is, data availability and LinkedIn limits. A tighter fence is a decision to spend those companies on people who fit.
More about this product: LinkedIn Sales Navigator AI
← Back to the blogWe're happy to show you exactly how many profiles get skipped versus contacted, and why.
Book my demoMost "AI sales tools" are a mail-merge with a chatbot logo glued on. Ours is a real decision engine.
1 min readNot marketing promises: hard-coded floors the bot enforces on every action, every account, every day.
2 min readA reply gets read for what it actually says, in the language it was written in, before the AI answers it.
3 min readEvery tool promising LinkedIn automation is working around the same fact. The honest question is how.
2 min readTell us about your sales team and target markets. We'll reply within one business day, usually much faster.
📧 info@metasmartfactory.com
☎️ +386 41 441 757
🏢 Meta Smart Factory d.o.o., Dunajska cesta 106, 1000 Ljubljana, Slovenia
🏛️ Technical Office · Vienna, Austria
🌐 www.metasmartfactory.com
💼 LinkedIn · YouTube
🌍 Partner network in Turkey, Saudi Arabia, Pakistan, Brazil, USA, Canada, UK, France, Croatia, Serbia, South Korea, Indonesia, Mexico & more
💬 Or just ask the chat assistant in the corner →