AI Operations · Glossary
Routine Adaptive Traversal
Routine Adaptive Traversal is the property of an authored routine that allows the agent to adapt its path mid-execution based on context, rather than following a fixed sequence or a rule-based branch.
A routine is not a flowchart. A flowchart commits at design time to every branch. An authored routine traverses adaptively, taking the next step that the current context warrants under the authored policies that govern the routine.
Adaptive Traversal is not rule-based branching. Rule-based branching enumerates conditions in advance. It is not state-machine transition. State machines fix the transition function. Adaptive Traversal evaluates the right next step against the agent's per-decision context, assembled by Dynamic Context Engineering, and governed by the policies in scope.
The capability is authored, not emergent. Domain experts compose routines to permit traversal. They mark steps where backtracking is allowed (see Backtracking). They link routines where adaptive crossover is permitted (see Routine Linking).
The result is behaviour that responds to context without collapsing into unbounded reasoning.