Startups & Founders
Enterprise AI Deals Reopen on the Same Clock It Took to Close Them
Published: 2026-09-09
In short
Madrona's August 13 report found 77% of enterprises re-examine their AI vendors at least every six months. In the same survey, 52% of deals close in under six.
Mr. Latte's take
Annual recurring revenue in this market is not a twelve-month promise. What the Madrona numbers describe is not a churn rate but a re-evaluation cadence, and that cadence runs the same length as the sales cycle, so the hand that wins new contracts and the hand that defends existing ones are the same hand. Before reading the 74% who plan to grow AI budgets as a growth signal, read the condition attached to it: that budget gets reallocated twice a year.
52% and 77% Point at the Same Six Months
Two numbers sit next to each other in the enterprise AI report Madrona published on August 13. Fifty-two percent of deals close in under six months. And 77% of enterprises re-evaluate their AI vendors at least every six months, with 29% of them doing it on a rolling basis.
Put the two together and you get the contract lifespan of this market. A deal that took half a year to win comes back before the same review board half a year later. The survey covers 150 senior enterprise decision-makers, alongside a practitioner survey of engineering leaders and five years of IA40 list data.
Budgets themselves are growing. Seventy-four percent of respondents plan to expand AI spending over the next twelve months. Nothing in the survey says the expanded budget returns to the same vendor.
Counting Pilots and Counting Companies Are Different Claims
The conversion numbers in the same report read like this. Eighty-three percent of enterprises moved fewer than half of their AI pilots into production over the last twelve months. More than a third converted fewer than one in four. One percent got past three quarters.
This is the passage that drifts when it gets restated. “Fewer than half of AI pilots reach production” and “83% of enterprises converted fewer than half of their pilots” count different things. The first counts pilots, the second counts companies. For a vendor the second is the worse news, because it means your pilot can be going well and still get cut alongside the other pilots inside that account.
The report also names what tips a scale decision. Sixty-five percent of buyers ranked strong end-user adoption and feedback in their top three factors. As for how enterprises find AI tools at all, 41% said internal tech-team research, the largest single channel.
Churn Sorts by Price Band, Not by Product Category
Knowing that re-evaluation is frequent does not tell you where the damage lands. The dataset that fills that in is the retention report Kyle Poyar produced with ChartMogul. They scraped the websites of 3,500 software companies, sorted them into roughly 2,700 B2B SaaS, 600 B2C SaaS and 200 AI-native businesses, and compared annualized retention through 2025 among companies past $250k ARR.
The medians split cleanly. B2B SaaS net revenue retention was 82%, with an upper quartile of 97%. B2C came in at 49%. AI-native companies posted 40% gross revenue retention and 48% net, worse on the gross figure than B2C.
Slice by price and the picture changes shape. AI-native products selling above $250 per month showed 70% GRR and 85% NRR, effectively the same as B2B SaaS. The $50 to $249 band ran 45% and 61%. Below $50 per month it fell to 23% and 32%. Same category of product, and the price tag alone spreads retention by roughly three times.
There is a recovery signal in the same report. Median gross revenue retention for AI-native companies rose from 27% in January 2025 to 40% in September. The reading offered is that the early tourists left and the ones who stayed were more committed.
What You Can Put on the Table at Renewal
For a seed or Series A team, the trap these two datasets expose is the enterprise pilot booked as revenue. Pilot budget and operating budget move through different approvals and different review cycles. What the six-month review asks for is not a demo, it is the usage record from the months in between.
So the thing to fix now is instrumentation, not the price sheet. Who uses it and how often, which task inside the account it displaced, and what evidence the owner of that task would cite for keeping it. That 65% on end-user adoption points exactly here. Six months without logs is a blank page at the renewal meeting.
Splitting experimental spend from operational spend before announcing ARR belongs to the same discipline. Combine the two lines and the growth rate looks larger, but you lose the ability to tell which line disappears in six months. Note that the ChartMogul figures describe 2025 while the Madrona survey describes the past twelve months, so they are adjacent readings rather than the same period.
Two fields to check in the contract: when the renewal date is, and where the record you will bring to it is accumulating. With both blank, the renewal conversation is a first sale all over again.
Sources