Decision debt
The maintenance burden created when an AI system's thresholds and decision boundaries are set once and never revisited, even as data, models, costs and risks change.
Decision debt is the upkeep an organization quietly owes once an AI system is allowed to act on a probability. Somebody chose that a score of 0.85 means "approve" and 0.60 means "send to a person". That choice was reasonable on the day it was made. It then keeps running while the customer mix shifts, the model is updated, the cost of a mistake changes and the people who set the number move on.
The debt is not in the model. It is in the decision boundary around the model: who owns it, when it was last tested against real outcomes, and what evidence would make the organization change it.
Why it matters
A governance question that stops at "who approved this model?" misses the point. The more operational question is "who owns this decision boundary, when was it last tested, and what would make us change it?" Teams that cannot answer it are accumulating decision debt, and like any debt it is cheap to ignore until it is expensive to repay.
How to keep it small
- Give every threshold policy a named owner and a review date.
- Feed real outcomes back so calibration can be re-measured.
- Keep a path for abstention so uncertain cases are not forced into a yes or no.
Read more in OpenAI Decisions API turns probability into policy.
Related terms
Threshold policy
The explicit rules that turn a model's probability or score into an action — reject, send to a human, or act autonomously — owned and reviewed separately from the model itself.
Calibration
How well a model's stated confidence matches reality: if it says 90% on a hundred comparable cases, roughly ninety should be correct. Once probabilities drive actions, it becomes an operating metric.
Abstention
A deliberate option for an AI decision system to decline to decide and return uncertain or high-consequence cases to a person, rather than forcing every input into yes or no.