← All writing
Sales Jun 18, 2026 5 min read

The AI SDR boom is real. Retention is the test it has to pass.

Christopher Dorsey

Christopher Dorsey

AI & MadTech Advisor · Enterprise Sales Leader

TL;DR

AI SDR is a real, fast-growing category, but several vendors are wrestling with high first-year churn — largely because the tools get demoed on activity and judged on outcomes. When prospecting gets over-automated, reply rates can drop and raw volume masks it. The takeaway cuts both ways: if you buy one, write pilot criteria around converted pipeline, not meetings booked; if you sell one, anchor success on outcomes, not activity — that's what earns the renewal.

The pitch for an AI SDR is the cleanest in software. Fire the expensive humans who keep missing quota, hire an agent that prospects around the clock for a fraction of the price. It worked. The category is worth billions. It also loses customers faster than almost anything else you can buy: annual churn runs 50 to 70%, and some tools are down most of their buyers inside three months. A market that grows that fast while leaking that hard is running on a treadmill. Much of the “growth” is the same seat sold to the next believer after the last one quit.

The category keeps one number off its landing pages. When teams handed prospecting to the machines, reply rates didn't hold steady at a lower cost. They dropped, from about 2.1% with human reps to 1.3%. Every individual touch got worse. Volume buried the evidence, because ten thousand mediocre emails still book a handful of meetings, and a dashboard full of activity feels like progress right up until the quarter closes and the pipeline isn't there.

Churn that was built in at the sale

The same trap waits for anyone selling AI, not just outbound tools. These products get demoed on activity: emails sent, meetings booked, pipeline “sourced.” Activity closes the deal, because activity is visible in week two. Nobody renews on activity. They renew on whether qualified pipeline turned into revenue, and that verdict doesn't land until two quarters later, long after the contract was signed on a number that never predicted it. The vendor tuned the demo to the metric that wins the sale and ignored the one that wins the renewal. The customer churns out feeling burned and can't quite say why.

I learned this the expensive way. At Zeta I sold an AI customer-acquisition product into brands that didn't have a budget line for it yet, and the lesson that stuck was that the renewal is the sale that counts, and it's mostly won or lost in the first 90 days of production. Spend that quarter celebrating activity instead of standing up the one outcome the buyer's CFO actually counts, and the renewal is already gone. You just don't find out for nine months.

Make the vendor live on the metric that renews

Two things can be true: AI belongs in outbound, and most of the way it's sold today is a churn machine wearing a growth chart. The teams getting real value run a hybrid, where the agent carries volume and a human owns judgment, qualification, and the relationship, and those teams report more pipeline, not less. If you're buying, write your pilot's success criteria around converted, qualified pipeline, not meetings booked, and make the vendor live or die on it. If you're selling, refuse activity-based success criteria even when the buyer offers them, because the easy yes in month one is the expensive goodbye in month twelve. Win the metric that renews, and let the demo be the smallest promise you make.

Share this post

About the author

Christopher Dorsey

Christopher Dorsey

Enterprise Sales Leader · AI Go-To-Market · Startup Advisor · Denver, CO

Fifteen years selling technology to Fortune 500 brands across AI, advertising, and data infrastructure — most recently at Zeta Global, Oracle, and Fastly. Currently advising founders and sales leaders on AI go-to-market and Generative Engine Optimization.

Questions, pushback, or just want to compare notes?