Part 1 of a series on GTM in the AI era.
Average cold email reply rates have fallen from roughly 8.5% in 2019 to about 3.43% today. That’s not a soft decline. That’s most of your pipeline evaporating while your CRM still says “outbound” at the top of the funnel.
The easy story is that AI ruined it. Everyone’s using the same tools, generating the same sequences, and buyers have gotten faster at spotting a machine than we’ve gotten at disguising one. There’s truth in that. But it’s not the useful part of the story, and founders who stop there end up solving the wrong problem, usually by turning the volume down instead of turning the effort up.
I’ve watched this play out with more than one client who was certain their outbound problem was a tooling problem. It almost never is.
The tell was never AI. It’s whether anyone did the work
Here’s the detail that gets buried under the “AI killed cold email” headline: the reply-rate gap between fully AI-generated sends and human-written ones, on comparable lists, is now down to about a single percentage point. That’s not nothing, but it’s not the story either. The real gap is between generic and signal-based outreach, and that gap is enormous. Emails that reference an actual trigger, a funding round, a leadership change, a specific and current problem, are landing 15 to 25% reply rates. Five times the baseline. Same inbox, same buyer, same year.
So AI isn’t the variable that predicts whether a message gets ignored. Effort is. AI just made it possible to produce a thousand messages that look personalized without anyone having actually looked at the prospect. That used to take real time, which meant it only happened for accounts worth the time. Now it’s free, which means it happens for everyone, which means the buyer has learned to assume nobody did the work until proven otherwise.
The tell was never AI vs. human. It’s whether anyone did the work.
What the winners are doing differently
The teams still getting real reply rates aren’t the ones who went back to writing every email by hand, and they’re not the ones who out-automated everyone else either. The pattern that keeps showing up: AI handles the research and the first draft, and a human closes the gap between “technically personalized” and “actually relevant” before it goes out. AI finds the signal. A person decides whether the signal is worth acting on and says something true about it. There’s also real evidence for this in the conversion data, not just replies: AI-run sequences convert meetings to opportunities at roughly 15%, versus 25% for human-run ones. AI wins on volume and persistence. It loses on the part of the job that actually closes.
Why presence is winning back ground
The other shift worth naming, because founders underrate it, is that buyers now expect to move across something like ten different channels over the course of a purchase, and the accounts that switch suppliers cite inconsistent information and a lack of knowledgeable support more often than price. Meeting the client where they physically are, a conference, their office, their vertical’s trade show floor, isn’t nostalgia. It’s now one of the only moves left that a generic AI sequence structurally cannot fake. Everything that can be automated has been. Presence, specific understanding of the prospect’s actual problem, and a track record of being right about their industry are what’s left to differentiate on, which is exactly why they’re worth more now than they were five years ago, not less.
That’s the reframe. AI didn’t make outbound harder. It made cheap outbound worthless, and expensive outbound, the kind built on real research, a real point of view, and a human who shows up, worth more than it’s been in a decade.
The mistake isn’t using AI in your outreach. It’s using it to avoid the parts of selling that were never supposed to scale in the first place.
Next in this series: why your first ten deals should be run as research, not sales – and why hiring an AE to “own the pipeline” before deal ten is the most expensive sequencing mistake at pre-seed.