Attrition value is the third rail of AI measurement. Understate it and you're accused of missing the point. Overstate it and you're accused of recklessness. Doing it responsibly requires a framework most organizations don't have.
The Umbral Intelligence approach treats attrition value as a distribution, not a number. For a given role and a given level of AI integration, we produce a range — a conservative estimate, a central estimate, and an optimistic ceiling — anchored to published automation baselines and discounted for organizational friction.
What we never do is publish the optimistic ceiling as the number. That's the difference between a research product and a marketing one. The range exists so leaders can make decisions with appropriate uncertainty — not so vendors can cite the top of it.