Field notes
The autonomous AI SDR is this year's spray-and-pray
The volume got cheaper and the market did not get warmer. That is not a new motion. It is the old mistake, automated.
An autonomous AI SDR that blasts a cold list is not a go-to-market motion - it is spray-and-pray with the spraying automated. Cheaper volume does not warm a market, and evidence and fit still decide whether outreach works.
On this page
An AI SDR that blasts a cold list is not a go-to-market motion. It is spray-and-pray with the spraying automated. The volume got cheaper. The market did not get warmer. Your domain reputation got more expensive.
The pitch is good, I will give it that. An agent that sends five hundred personalised emails a day while you sleep, follows up on its own, books the meeting, and never asks for a raise. If outreach were a volume problem, this would be the answer.
Outreach was never a volume problem.
Volume was never the constraint
You could always send more email. A junior rep with a list and a caffeine problem could send hundreds a day, a decade ago. Nobody was ever short of send capacity.
Restraint was the point. The reason you did not blast the whole list was that it worked once and then never worked again, because the market remembers.
An AI SDR removes the one constraint that was doing useful work. It does not add a strategy. It just takes the floor off the volume.
And a market is not a spreadsheet you get to append to without consequence. Every send spends something - a stranger's attention, a little goodwill, the small credit a name extends you once. Automate the spending and you do not end up with more market. You end up with less of it, faster.
The personalisation tell
The demos always lead with personalisation, because it is the objection they know is coming. Watch what it actually personalises. The first name. The company. A line scraped off the About page or a recent post, reworded into a compliment.
Every other agent is reading the same About page. The prospect gets six versions of the same synthetic warmth in a week and learns the pattern fast. Personalisation that everyone can generate is just spam with its homework showing.
Real personalisation is knowing something the prospect has not published. A model reading a public profile does not have that. It has what every other model has, rephrased to sound like it noticed you specifically.
A cold list is a room full of people already sick of the pitch
A rich, clustered segment - the kind that looks like an obvious target - is usually a segment already worked to exhaustion. I have written about this as pay to play: the cluster of prospects everyone wants is a cluster of people who have already learned to hang up.
Now point an autonomous agent at that same list. Same tired buyers. Same recycled angle. Ten times the volume, arriving at machine speed, from a domain you actually need for real email.
The market has antibodies. You are about to give it a very large, very fast dose of the exact thing it built them for.
Spam gets cheaper to send and more expensive to receive
Here is the asymmetry the pitch leaves out. The send cost falls toward zero. The receive cost does not.
Every ignored blast trains a spam filter. Every annoyed recipient marks a domain. Deliverability is a reputation, and a reputation is slow to build and fast to burn. You can torch the deliverability of a domain you will still need in two years, to send email that was never going to land. [STAT: reply-rate collapse on high-volume cold automated outreach - Alex to add a vetted figure]
The invoice for the sending looks small. The invoice for the reputation shows up later, and it is not itemised.
Evidence and fit still decide it
Outreach works when you have earned the right to send it. A segment your own data points to. A pattern you can see repeating in your closed-won. A market you warmed for months before you asked for the meeting.
An AI SDR checks none of that. It does not know whether you can win the game. It knows how to send. Handed a list with no evidence behind it, it does the thing junior reps get scolded for, just with a bigger invoice and a better caterer.
The receipts still matter. The AI cannot produce them for you, and it cannot skip them.
What an AI SDR is actually good at
This is not an argument against the tool. It is an argument against pointing it at a cold list and calling it a motion.
The same agent is genuinely useful the moment there is a warm signal to work. Following up the fifth time on a lead that went quiet - most deals need five or more touches, and that is exactly the patient, unglamorous work people drop. Routing the inbound to the right rep in seconds. Reading a half-filled form and deciding it is worth a call. Enriching a record before a human ever sees it.
That is outreach connected to revenue instead of detached from it. The difference is not the technology. It is whether there was a market on the other end with any reason to hear from you.
The rep who sprayed a cold list got a talking-to. The agent that does it a thousand times faster gets a budget line. Same email. Nobody checked whether it should have been sent.
Common questions
Do AI SDRs actually work?
They work when pointed at a warmed market with evidence behind it - follow-up, routing, and enrichment on real signals. Pointed at a cold list, an AI SDR just automates spray-and-pray and burns your domain reputation doing it.
Is an AI SDR a go-to-market strategy?
No. Sending capacity is not a motion. A strategy is knowing which segment your own data says you can win and warming it first; the AI only changes how fast you send, not whether anyone wants to hear from you.
Will AI outreach hurt my email deliverability?
High-volume cold automated sending can. The send cost falls toward zero but the reputation cost does not - ignored blasts train spam filters and mark your domain, and that is slow to rebuild.
A pipeline you can trust and a forecast that holds. That is the work we do.
Get the next one
New field notes twice a month. We don’t do newsletters, follow RevOps XL on LinkedIn instead.
Follow on LinkedIn ↗