Is Cold Email Really Dead? Why AI Gatekeepers Won't Kill It
Greg Isenberg says AI inbox agents will kill cold email. We disagree. Gatekeepers will crush spray-and-pray and reward high-signal outreach. Here's why.
Greg Isenberg says AI inbox agents will kill cold email. We disagree. Gatekeepers will crush spray-and-pray and reward high-signal outreach. Here's why.
I've been thinking about Greg Isenberg's post on X the other day:
I think cold email is going to die.
— GREG ISENBERG (@gregisenberg) August 30, 2026
Every email inbox will have an agent gatekeeper soon, and the only way through will be a warm intro or being interesting enough that the agent decides you're worth its person's time.
Greg's claim is that inbox agents will become default gatekeepers, so unsolicited email stops working except via warm intros or some rare "interesting enough" exception.
I disagree. I think cold email survives by mutating into a higher-signal channel, and that agents may even improve the economics of good outbound.
Humans already treat most cold email as noise. They skim, delete, or never open it. An agent that can actually read the message, check the sender's company, look at prior contact, estimate relevance to the recipient's current work, and score whether similar people buy this kind of thing is not automatically more hostile than a tired human. It can be more discriminating in the useful direction.
That means the thing that dies is spray-and-pray: 100k barely personalized emails hoping for a 0.3% reply. What remains is outreach that can present evidence: this person, this problem, this timing, this proof. In that world, a tight, specific email can look better to an agent than it currently looks to a human drowning in similar-looking slop. The bar rises; the relative payoff for clearing the bar can rise with it.
Spam filters, Promotions tabs, domain reputation, and resume screens all looked like they would kill the previous channel. They killed the cheap version. Then people optimized for the new judge. An agent gatekeeper is still a filter with criteria. Those criteria become public-ish through trial and error within weeks. Senders will write for the agent the way they now write for Gmail's classifiers and for human pattern-matching.
The historical pattern is not "filter appears → channel dies." It is "filter appears → volume of garbage falls or gets rerouted → remaining traffic has to carry more signal → the channel professionalizes." Cold email has already been through several of those cycles and is still a core B2B motion.
Warm intros are great and they do not scale to the number of conversations most companies need. Most buyers and most sellers do not share a dense graph. New products, new categories, and companies without famous founders still have to create first contact somehow. If email is no longer that surface, something else has to be, and that something else will have its own gatekeepers (LinkedIn agents, calendar agents, community agents). The underlying job of initiating a relevant conversation with a stranger who has a problem you can solve does not disappear.
Email remains the closest thing to a universal, durable, searchable, permission-light protocol. Agents sit on top of that protocol. They do not retire the address.
Greg already concedes a path through: the agent decides you are worth the person's time. That is not a tiny loophole. That is the product spec for the next generation of outbound. Relevance, proof, timing, and reputation become the payload. Tools will exist to assemble that payload from public signals (posts, hiring, tech stack, recent comments, funding, job changes) and to present it in a form an agent can evaluate cheaply.
So the other side's prediction is not that 2023-style sequences keep working. It is:
Cold email as a tactic changes. Cold email as the act of reaching a stranger with something specific does not have an obvious replacement that scales better.
