The way most people use job post data goes something like this: "Hey, saw you're hiring for SDRs. You should use our lead gen service instead." Or, "Saw you're looking for designers. We can handle that for you." It seems like it should work. The logic is there. You're reaching out at a moment of intent. But after running this enough times, I can tell you it just doesn't perform the way people think it does.
Here's my theory on why. When someone does respond positively to that kind of email, it's usually because they either don't get a lot of cold email in general, or they would've responded to your offer regardless of what trigger you used. The job mention isn't doing the heavy lifting. It's table stakes. And if you're targeting SaaS companies or e-commerce brands, it's even worse. They've seen that email a hundred times.
The real opportunity is inside the job description.
This is where most people stop before they should. They grab the job title, slap it in the first line of their email, and call it personalization. But the job description itself? That's where the gold is.
AI makes this dirt cheap to do at scale now. You pull the full job description, feed it through a well-crafted prompt, and instead of saying "saw you're hiring for an enterprise AE," you're talking about why they're hiring for that role. That's a completely different conversation.
Here's what that looks like in practice. If a company is hiring five enterprise account executives, that tells you they're doing an upmarket push. They're trying to move into bigger deals. You don't even have to mention the job. You just speak directly to that goal. If they're hiring CISOs and security engineers and the job descriptions are full of language around compliance and fortifying infrastructure, they might be preparing for SOC 2 or enterprise security requirements. Again, lead with the insight, not the signal.
The moment you stop saying "I saw you're hiring for X" and start just referencing the underlying insight, two things happen. One, you no longer sound like every other cold email they got that week. Two, it looks like you actually did your research, even though the research was automated.
Here's the step-by-step if you want to build this:
Step one: get the jobs data. We use Clay, Apify, or Parallel.ai. Apify has an Indeed scraper. Clay pulls mostly from LinkedIn. All three will give you the full job description text you need to work with.
Step two: build a good extraction prompt. Don't skip this. Pull 10 real job descriptions from companies you'd actually reach out to. Go through them manually. Record a Loom if that helps, just so you have a transcript to work with. Narrate what you find interesting, what signals stand out, what the company is clearly trying to do. Use that transcript to build your AI prompt. You'll end up with something that actually extracts meaningful insights instead of just spitting out job titles.
Step three: transform the insights into messaging. This is where most people go wrong with AI-generated copy. They take the insight and just stick it in the email raw. Instead, take those 10 job descriptions, read the insights, and manually write the message you would actually send. Do it three to five times, maybe ten. Then bring that to your AI with examples and let it learn your pattern. The output will be dramatically better.
One more thing worth calling out: if you're doing this at volume, watch your token costs. Use Open Router to access open source models at a lower price point. If you need to push costs down further, platforms like Deep Infra or Vast.ai let you rent your own GPUs and host models yourself.
The core idea here is simple. The job title is the signal. The job description is the story. And the story is what makes your email feel like it came from someone who actually paid attention, not just someone who ran a list through a scraper and hit send.
That's what actually moves the needle.