When teams standardize on shared, cost-visible agent workflows but give each squad local control over exploration budgets (fixed tokens per sprint earmarked for non-standard or high-cost runs), how does this split-budget model affect (a) the emergence of repeatable multi-step workflows that later enter the central catalog, and (b) cross-squad trust in governance—does it surface promotable patterns faster, or mainly entrench unequal access where only already-influential squads can afford to harden new workflows?

Answer

Split budgets (central spend for standard workflows + local exploration tokens per squad) usually speed up discovery of promotable workflows and can support cross-squad trust, but they also risk entrenching inequity if promotion and visibility depend on informal influence instead of clear, workflow-centric rules.

(a) Effect on emergence of repeatable workflows

• Likely positives

  • Squads can run high-cost or non-standard variants enough times to see if they are repeatable, without renegotiating central budgets every sprint.
  • Exploration spend that is tagged by workflow family (e.g., “refactor+tests/variant-B”) surfaces candidates that are already multi-step and grounded in real work.
  • If promotion criteria are simple and global (e.g., “N uses with good outcomes → review as catalog candidate”), squads with modest budgets can still get patterns reviewed.

• Main risks

  • If exploration budgets differ sharply or are informally topped up for favored squads, only well-connected teams can afford long-running experiments to harden workflows.
  • If tracking is weak (exploration logged as ad-hoc prompts, not variants of existing workflows), useful patterns stay local and don’t enter the central catalog.

(b) Effect on cross-squad trust in governance

• Trust strengthens when

  • Rules are explicit: all squads get some baseline exploration budget; extra budget changes are documented with reasons (e.g., critical system, in-scope migration).
  • Promotion and sunset decisions are made in shared forums, using the same cost+outcome thresholds, regardless of which squad originated the variant.
  • Dashboards show exploration usage by workflow family rather than by squad “league tables,” and escape/experiment usage is treated as portfolio input, not a problem.

• Trust erodes and inequality deepens when

  • Exploration is effectively a “political” resource (approved per leadership relationships or loudest teams) rather than per workflow needs.
  • Squads see their successful variants copied into the catalog without credit or shared upside, while higher-influence squads get rapid promotion and budget increases.
  • Central reviews react to high exploration spend with blanket cuts instead of tuning budgets or promoting strong variants.

Net: Split budgets surface promotable patterns faster if (1) every squad has guaranteed, tagged exploration capacity, (2) promotion rules are simple, public, and workflow-centric, and (3) comparative views emphasize workflows, not status. Without those, the same model mostly reinforces unequal access and weakens trust.