What a financial product brings a user is net value: the outcome it delivers, minus the effort the user puts in, minus the risk the user bears as a result; and all of this rests on the premise that the user has reason to trust it. The outcome is constrained by the market: only what the market can provide can be promised. Effort and risk are constrained by the user: a design that exceeds what the user can bear will not be carried through. How much the product can do on the user's behalf is constrained by AI. Part Four sets out design at four layers: the value layer decides what to promise, the contract layer agrees on what terms to act, the implementation layer ensures how the promise is delivered, and the trust layer decides whether the promise can be accepted.
Each of the three terms is governed by one set of constraints. How large the outcome can be is decided by the market: time compensation and risk compensation are available to everyone, while information advantage is used up by competition. How much effort and risk can be borne is decided by the user: attention is limited, and risk tolerance has three faces. How much of the user's effort the product can take on is decided by AI: scope of application, reliability and the cost of checking set its upper limit.
The first three parts each set out one set of constraints; this part discusses the design choices made within them. Each chapter again starts from a concrete situation and states which constraint each design requirement comes from.
Under the three sets of constraints, designing a financial product comes down to four sentences: promise only value that can be delivered and verified; assign powers according to consequences, and align the agent's income and goals with the user's; make the implementation auditable, with assurance proportionate to risk; and earn, with diagnosable evidence, trust proportionate to how trustworthy the product is.
The market decides what can be offered; the user decides what is wanted, what can be borne and what grounds for belief are needed; AI decides how much can be done on people's behalf. A good financial product is the one that, under these three sets of constraints, maximizes the net value the user receives.