Value-Share Contracts

Imagine a vendor that supplies an AI tool to a contract manufacturer. The tool drafts inspection reports. The vendor does not want its entire payment to depend only on the sale being signed. It wants part of the payment to depend on results the tool actually produces. That basic wish is the appeal of a value-share arrangement, in which payment is tied to the value that is generated.

The difficulty appears at the moment of settlement. If the customer's data cannot separate what the AI contributed from everything else happening in the business, both sides can end up arguing about who deserves credit. A simple yes-or-no payment rule gives them no way out. As you read, hold this question in mind: what kind of payment rule lets two parties agree on a fair amount even when they cannot agree on a precise measurement?

Before naming any structure, follow what a payment rule does with a measurement. Suppose the contract has one trigger point: a measured outcome at or above that point pays the full amount, and a measured outcome below it pays nothing. Now suppose the measurement is noisy, so the true result is uncertain by a small margin. A reading just below the trigger and a reading just above it sit almost on top of each other, yet the payment differs completely between them.

The two readings differ by a tiny amount of measurement error, but the payment jumps from nothing to the whole amount. Whichever side has the more favorable reading will defend it, and the other side has no middle ground to accept. This is the fragility that the source attributes to all-or-nothing payment: it is easy to write down and easy to check, but when the measurement is messy, any binary test forces a verdict one side will reject.

The problem is not that the all-or-nothing rule is wrong about what happened. The problem is that the rule converts a question of degree into a question of kind. If the customer's data cannot cleanly isolate the AI's contribution, the honest answer is often that the AI contributed something, though not a precisely known amount. A binary rule has no slot for that answer, so it pushes both parties toward claiming the entire outcome or disclaiming it entirely.

A graduated design does not ask whether the value showed up at all. It asks how much of the measured result was achieved and pays accordingly. Because the payment grows with the measured result, a reading that is somewhat short of the top outcome still earns a partial payment, and a reading that clearly improves earns more. Nothing in the payment depends on a single knife-edge threshold.

The diagram shows the difference in shape. The all-or-nothing column has a jump between the middle and high results. The graduated column rises in steps, so a small change in the measured result produces a small change in payment. That smoothness is what allows both parties to accept a fair amount even when the exact measurement is contested.

This article follows the case of an inspection-report drafter used by a contract manufacturer. The source describes how the way the outcome payment was structured kept the arrangement from collapsing into a dispute. Walk through the settlement decision as a path of choices, starting from the data the customer has.

The source states that the graduated structure prevented a dispute when the customer's data could not isolate the AI's contribution cleanly. The path above shows the logic in plain steps. When the data is clean enough, payment can follow the measured degree of result. When it is not, the contract already holds agreed tiers, so the settlement does not have to be won by one side. Both routes end at a payment the parties can accept, which is the point of the design.

It helps to keep the two ideas distinct. A value-share contract describes the relationship: pay is linked to the value generated. A graduated outcome payment describes how that link is calculated and settled. One is the goal and the other is the working rule that makes the goal achievable when the data is not clean.