When pay-as-you-go teams standardize a few high-cost, high-ROI cost-visible agent workflows (e.g., large-context refactor+tests, incident deep-dive) and then later tighten portfolio budgets, which governance pattern better preserves durable adoption: (a) relaxing caps on these named workflows while cutting or sunsetting cheaper, low-ROI ones, or (b) enforcing uniform percentage cuts across all workflows—and how do these choices affect developers’ willingness to rely on the expensive workflows in ambiguous, real-world cases?

Answer

Pattern (a)—protecting or even relaxing caps on a few named high-cost, high-ROI workflows while aggressively pruning cheaper, low-ROI ones—tends to better preserve durable adoption and developers’ willingness to rely on expensive workflows in ambiguous real-world cases.

Uniform percentage cuts (pattern b) usually:

• Reintroduce token anxiety around the very workflows you most want used in hard, ambiguous situations.
• Signal that leadership treats all spend as roughly interchangeable, which pushes developers to reserve expensive workflows for only the most extreme, clearly justifiable cases.

Why pattern (a) works better for durable adoption:

• It makes the governance story legible: “These specific expensive workflows are strategically important and protected as long as they keep paying off; low-ROI workflows are what get cut first.”
• Developers learn that using sanctioned high-cost workflows in messy, borderline cases is expected behavior, not a personal risk.
• Portfolio reviews can focus on: (1) defending the high-ROI workflows with data, and (2) systematically sunsetting or simplifying low-ROI patterns to fit within the tighter budget.

To avoid perverse effects with pattern (a):

• Require each protected high-cost workflow to carry explicit value tags (e.g., MTTR, defect escape rate, refactor lead time) and appear in regular portfolio reviews.
• Make the rules explicit: “If we must cut again, we first simplify steps inside these workflows; only then do we restrict when they’re allowed.”
• Pair cuts to low-ROI workflows with alternatives (cheaper variants, clearer guidance) so people don’t feel that budget tightening means “avoid agents in ambiguous work.”

Net: Under tighter budgets, uneven, value-aware cuts that explicitly protect a small set of named, high-ROI workflows better sustain both durable adoption and developers’ willingness to use those expensive workflows in real-world, ambiguous cases than uniform percentage cuts across the board.