How to Automate Outbound Without Spam
By Sammy from Cora Intelligence
A message lands in my inbox: "Hi Sammy, saw you're scaling Cora Intelligence and thought I'd reach out." I don't need to check the footer for a sending-tool logo. I already know.
Congratulations, outreach is now infinite. So is spam.
I wrote that line as a joke at the end of the last piece. Turns out it was a forecast.
For a decade, the only thing keeping outbound sane was that a human had to do it — look up the company, guess the contact, write the email, wait, follow up. Slow, but self-limiting. AI deleted the limit. Now anyone can send ten thousand personalized-looking emails before their coffee gets cold, and most of them read exactly as mediocre as the process behind them — just faster.
Buyers noticed first. They always do.
Inboxes are quietly growing bouncers, filters that read a message the way a tired VP would and decide, on someone's behalf, it isn't worth thirty seconds.
Give it a year and both sides of the wire have an agent deciding whether the humans attached to them should even talk.
Once that's normal, sending the most messages isn't an edge — it's the tell that nobody screened you in.
So what survives an inbox like that? Evidence, not volume. Write because something real happened — funding, a hire, a launch — not because a name matched a filter. And keep exactly one decision human: whether this specific person, right now, is worth a message at all. Everything upstream of that can be a machine. That one call shouldn't be.
That's the whole bet behind how we built Cora.
Not how many contacts it chews through an hour, but how good the thinking is behind each one — the leads worth the trouble, the stakeholders worth calling, the research underneath before a word gets typed. Sending was always going to get solved by software. Judgment was the hard part, and it's the only part still worth bragging about.
Trust doesn't scale the way messages do.
You can 10x your send volume and still be worth exactly as much as your worst email made you look.
Paul Graham has this line he gives founders in office hours: stop asking how to make more money, ask what would make you more powerful. His answer, borrowed from Tim O'Reilly, is annoyingly simple — be generous. Squeeze every last reply out of your list and you're capped at maybe double. Actually have something worth saying, and now you're playing for 10x, 100x, whatever — the kind of return you only get from being trusted, not just seen.
I still wince at the occasional message that slips through my own filters — once, memorably, from a customer who'd forgotten to turn a setting off.
Good reminder of exactly what we're building against.
Frequently asked questions
How do you automate outbound without becoming spam?
Automate the work, not the judgment. Research, list building, drafting, follow-up, CRM hygiene and booking can all run without a person touching them. The one decision that should stay human is whether this specific person, right now, is worth a message at all. The practical test is whether the message exists because something real happened at that company, a funding round, a hire, a launch, or only because a name matched a filter.
Does personalization still work if AI writes it?
Personalization was never the point, evidence was. A merge field wearing a first name is not evidence, and a model that writes a paragraph about a company from nothing is just a more fluent version of the same emptiness. What still earns thirty seconds is a specific, checkable reason this message was sent to this person now. If there is no such reason, the honest move is not to send.
What happens when buyers put AI filters in front of their inbox?
It is already starting. Filters read a message the way a tired VP would and decide, on someone's behalf, that it is not worth thirty seconds. Within a year both sides of the wire plausibly have an agent deciding whether the humans attached to them should even talk. Volume stops being an advantage at that point, because a high send count is exactly the signal a screening agent uses to reject you.
How many emails per day is safe to send?
Asking for a number is usually the wrong instinct, because the number that keeps you out of spam folders is not the number that keeps you worth reading. Deliverability has technical limits worth respecting, warmed domains, sane per-mailbox volume, real unsubscribe handling. But no send limit saves a message nobody had a reason to receive. Cap yourself on reasons, not on sends, and the safe volume falls out of that.