Here's the trap. You get quoted $45,000 a year for an AI SDR and you hold it against the $150,000 a human SDR costs you fully loaded. It looks like the easy call. Then the thing underperforms, you work that out somewhere around month four, and you keep paying for it until month twelve.
At a coin-flip chance of cancelling, budget double the sticker. That $45,000 contract costs $3,750 for every month it works, and $7,500 for every month it works once you price in the months it doesn't. Signed deals in the AI SDR category run from $18,000 to $65,000 a year, so the premium you are absorbing sits between $1,500 and $5,500 a month.
The mechanism is that you are quoted for a year and you commit to a year, but you find out in a quarter. It works like a twelve-month season ticket bought for a job you might leave in April. The ticket price assumes you travel all year, the railway keeps the rest when you stop, and every journey you did take ends up costing three times what the brochure said.
You cannot price whether it will work. You can price what happens if it doesn't. Here is the arithmetic, what it does to the comparison against your own team, and the single clause that removes the whole premium.
Nobody can tell you how often these get cancelled
Start with the number everyone quotes and nobody sources. This category is said to churn 50 to 70 percent a year, attributed to UserGems and repeated across most of the 2026 coverage. We went looking for the primary and could not reach one.
The claim also changes shape between the places that carry it. Some render it as 50 to 70 percent annual tool churn. Others render it as 50 to 70 percent of projects dying within 90 days. Those are different statements about different things, and nothing we found explains the gap. Treat the range as circulating rather than measured.
The one figure with actual reporting behind it comes from TechCrunch's March 2025 investigation into 11x. An employee told the publication the company was "losing 70-80% of customers that came through the door." 11x answered on the record that its "retention rate is currently 79%." Both appear in the same article. They are close to mirror images of each other, and the company has not reconciled them publicly since.
Gartner separately predicts that over 40 percent of agentic AI projects will be cancelled by the end of 2027. That is a forecast rather than a measurement, and it covers agentic AI broadly rather than this category specifically, so it sets a floor for expectations rather than answering the question.
What the arithmetic does to the price
The months you keep paying for after the tool has stopped earning its place deserve a name, because they are the entire cost of being wrong. Call them stranded months.
Assume failure lands at month four. That is ramp plus one full quarter of real pipeline, and it is about the earliest point at which a reasonable buyer would stop blaming onboarding. On a twelve-month term with no way out, you get four useful months and pay for twelve.
So the price per useful month is the annual figure divided by twelve if it works, and divided by four if it doesn't. Weight those two outcomes by how likely each is and you get what the contract actually costs you.
| Annual contract | Sticker, per useful month | At 50% churn | At 70% churn | With a 3-month break |
|---|---|---|---|---|
| $18,000 (category low) | $1,500 | $3,000 | $3,600 | $1,500 |
| $21,000 (Artisan median) | $1,750 | $3,500 | $4,200 | $1,750 |
| $40,125 (11x, Vendr median) | $3,344 | $6,688 | $8,025 | $3,344 |
| $45,000 (category median) | $3,750 | $7,500 | $9,000 | $3,750 |
| $65,640 (11x, observed high) | $5,470 | $10,940 | $13,128 | $5,470 |
The multiplier does not depend on the contract size at all. At a 50 percent chance of cancelling it is exactly 2.00 times the sticker. At 40 percent it is 1.80, at 70 percent it is 2.40, at 80 percent it is 2.60. Contract values for Artisan and 11x differ by a factor of three and the premium multiplies both identically.
What moves the answer far more than the churn rate is when you find out. Hold churn at 50 percent on a $45,000 deal and vary the month it fails: month two costs $13,125 per useful month, month three $9,375, month four $7,500, month six $5,625, month nine $4,375. A vendor who gets you to real pipeline quickly is worth more than a vendor who is cheaper, because early failure is the expensive kind.
The clause that deletes the premium
Look at the last column of that table again. It is identical to the sticker column, at every contract size, to the dollar.
A break clause does not reduce the churn premium. It removes it. If you can leave at month three or four, the bad outcome costs you four months for four months of use, which is the same rate as the good outcome. The uncertainty stops being something you pay for.
These clauses exist. TechCrunch reported that 11x contracts carried a break clause "typically at three months." The same reporting carries the company's qualification that this is not universal and that the majority of its mid-market customers qualify for other terms, which is worth reading as written: the term you get depends on which segment the vendor puts you in, and you will not be told which one that is.
Is a break clause worth more than a discount?
This is the part that changes what you do on Monday, because buyers spend their negotiating capital on price and hand back the exit terms for free.
Twenty percent off a $45,000 contract saves $9,000, guaranteed. A three-month break clause on the same contract avoids $30,000 of stranded months, but only in the case where you cancel, so its value is $30,000 multiplied by the chance you do.
Those two are equal at exactly 30 percent. Below that, take the discount. Above it, take the clause, and the gap widens fast: at 50 percent the clause is worth $15,000 against the discount's $9,000, and at 70 percent it is worth $21,000.
One reading breaks this, and it should be said plainly rather than buried. If you accept 11x's own figure of 79 percent retention, the implied churn is 21 percent, which sits below the 30 percent breakeven and makes the discount the better trade. That number comes from the vendor, it measures annual account retention rather than whether a project survived, and it sits in the same article as an employee saying the opposite. It is still the only company-published figure in the category, and if you believe it, the arithmetic says negotiate on price.
How does this compare to keeping the SDR?
Our own cost model puts an in-house SDR at $134,000 to $154,000 a year fully loaded, producing around 140 held meetings, which works out at $960 to $1,100 per held meeting. That is the number the AI contract has to beat, and the churn-adjusted figure is what should be doing the beating.
At $7,500 per useful month, the AI SDR costs $1,875 per meeting if it books four a month, $1,250 at six, $938 at eight, and $750 at ten. Break-even against a human sits at roughly eight held meetings a month, every month, throughout the contract.
Eight held meetings a month from one automated system is a real target rather than an impossible one, but it is the number to put in front of the vendor, and it is much higher than the number you get if you compare sticker to salary and stop there.
What to ask for before you sign
Four things, all of which are cheaper to get in writing now than to argue about in month five.
- A break clause at three or four months, in the contract rather than in an email. Ask what it costs to add. If the answer is a higher rate, run the comparison above, because paying 10 percent more for an exit is usually the better deal than 20 percent off without one.
- A written definition of what month four success looks like, in meetings held, not meetings booked and not replies. Eight a month is the benchmark this arithmetic produces. Agree the number before it becomes contested.
- Your data and your domains on exit. Ask what you leave with and in what format. Our walk-through of a ZoomInfo contract exit shows what this looks like when it was not agreed up front.
- A monthly option, priced. Several tools in this category sell without an annual commitment at all, including AiSDR at $900 a month and Clay from $185. We keep a census of which tools skip the annual contract. Paying monthly is the same thing as a break clause with a one-month notice period.
The answer to the question in the title is twice the sticker price, at the churn rate the category is most often credited with. The answer you should be aiming for is the sticker price, and the difference between the two is one clause rather than a better vendor.