The Contract Stopped Being the Moat

For twenty years, enterprise software had a structural advantage that had nothing to do with product quality.

You signed a three-year deal. Migration was painful. Nobody wanted to run a vendor review. Renewal happened because not renewing was more work than renewing.

That advantage is disappearing, and there is now a number attached to it.

What the research actually says

Venture firm Madrona surveyed 150 enterprise IT professionals. Two findings matter for anyone selling software right now.

The first is that budgets are healthy. 74% plan to expand AI spending in the next twelve months, and the rest plan to hold steady. Nobody is cutting. Total enterprise technology spend is on track for $4.25 trillion this year, almost all of it driven by AI.

The second finding is the one to sit with. 77% of enterprises re-evaluate their AI vendors every six months, or on a rolling basis with no fixed cycle at all.

Madrona’s own summary is blunt: this creates a “fast in, fast out” dynamic, and the re-evaluation cadence is relentless.

Read those two findings together. More money is moving, and it is staying in one place for less time.

There is a third number worth adding. Fewer than half of enterprise AI pilots make it into full production. That is actually an improvement, because MIT reported last year that 95% of enterprise AI projects failed to show a return. The bar has moved from catastrophic to merely difficult.

Fast in, fast out

The old enterprise motion had a shape. Long sales cycle, hard to win, then years of quiet revenue once you were in.

The new shape is the opposite. Easier to get in, because buyers are experimenting and budget exists. Much harder to stay.

This is why some AI companies are going from zero to $10 million in three months. That number is real. It is also less durable than it looks, because the same conditions that let you in that fast let the next vendor in just as fast.

If you sell to enterprises today, you are not defending a contract. You are defending a position that gets reviewed twice a year by someone whose job is to check whether something better exists.

Most revenue teams have not updated for this. They still treat the signature as the finish line. Pipeline reviews, forecasting, comp plans, the whole apparatus is built around closing. Very little of it is built around the six-month review nobody told you was happening.

The pricing problem underneath it

There is a second finding that explains part of the instability.

Andreessen Horowitz surveyed 50 technical AI buyers. More than half said they want fees tied to the work produced or to outcomes, rather than to usage like tokens consumed.

That is a bigger statement than it looks. Charging by seats or by usage is a SaaS-era model, and it works when the buyer already knows they need the thing. Once a company has decided it needs email or storage, the only question is how much.

AI does not get that assumption. The buyer is still deciding whether it works at all. So a bill that arrives measured in tokens asks them to convert consumption into value themselves, every month, while they are already unsure.

The a16z partners writing it up put it well. Pricing around the recognisable work makes the product “economically valuable to both sides.”

If your invoice is denominated in something your buyer does not care about, you have quietly made your own renewal harder.

What actually holds now

If the contract is not the moat, something else has to be.

Being inside the workflow, not beside it. Software that sits next to a process gets reviewed as a line item. Software the process runs through gets reviewed as a risk. That is not lock-in through pain, it is depth through use.

Owning an outcome with a number on it. Not “improves productivity.” Tickets closed, reports produced, leads qualified. Something the buyer already counts and already reports upward. When your value shows up inside a metric they defend in their own reviews, you are defended too.

Multiple people who would notice. Single-champion accounts have always been fragile. In a six-month review cycle they are close to indefensible, because one person changing jobs resets the whole relationship.

Evidence you can hand over. Your buyer has to justify keeping you to someone. If they have to build that case from scratch every cycle, most will not. Give them the numbers before they ask.

What to do this quarter

Assume a review is happening whether or not anyone tells you. Build a light quarterly check into account management. Not a QBR deck. A short conversation about what result they got and what they would say if asked to defend the spend.

Find out what your value is denominated in. Ask three customers what they would point to if their CFO questioned the line item. If they answer with a feature, you have work to do. If they answer with a number, keep that number.

Count your relationships per account. One contact is a countdown. Two is thin. Named relationships across the people who use it, pay for it and are accountable for the outcome is the minimum for anything you want to survive a review.

Stop reading fast growth as proof of safety. Revenue that arrived quickly can leave quickly. Those are the same property, not two different ones.

The uncomfortable summary is that enterprise buying got faster in both directions. Getting in is easier than it has ever been. Staying is harder than it has ever been. Most teams have rebuilt for the first half of that sentence and not the second.

Common questions