In products that already detect when a user has reached stable, cadenced use of a reusable workflow, which concrete, targeted interventions (e.g., micro-prompts to add a data source, lightweight variant suggestions, or “turn this into a team asset” nudges) most reliably convert apparent productivity plateaus into renewed gains in workflow maturity, and how do their effects differ for individual contributors versus power users who own shared workflows?

anthropic-learning-curves | Updated at

Answer

Most effective interventions at plateaus are small, in-context nudges that (1) expand inputs/outputs, (2) encourage safe variants, and (3) promote team use. Effects differ by role.

  1. High-yield interventions at plateaus

  • Data-surface nudges • Example: “Attach another source?” or “Also include CRM data?” when users run the same flow repeatedly. • Effect: Often increases workflow maturity by widening coverage (more cases handled per run) with modest time cost.

  • Variant/branch suggestions • Example: “Save this edited run as a variant for X client/segment?” after 2–3 similar edits. • Effect: Converts ad-hoc tweaks into named variants; supports branching without breaking the main flow.

  • Step-level refinement prompts • Example: “This step is often edited; try updating the workflow?” when the same step is changed across runs. • Effect: Shifts users from fixing outputs to fixing the workflow, increasing durability.

  • Cross-tool/export prompts • Example: “Send this to your reporting doc?” or “Create a slide deck from this?” when output is repeatedly copy-pasted. • Effect: Extends the workflow across tools, revealing higher-value chains.

  • Team-asset nudges • Example: “Others do similar work—share as team template?” once cadence and quality are stable. • Effect: Turns a personal workflow into a shared asset; raises team-level workflow maturity.

  1. Differential effects by role

  • Individual contributors (ICs) • Most effective: – Data-surface nudges (help them handle more of their own work). – Step-level refinement prompts (teach them to fix the flow, not just the output). – Light variant suggestions (personal specialization by client/task). • Less effective early: – Heavy team-asset nudges (may feel premature or risky).

  • Power users / workflow owners • Most effective: – Team-asset nudges (publish, permission, document). – Variant/AB suggestions at the template level (e.g., “Try a brief version for SDRs?”). – Cross-tool/export prompts that align with team SOPs. • Moderately useful: – Data-surface nudges when they expose org systems that should be standardized (e.g., “always include finance DB”).

  1. When they work best
  • Trigger conditions (using c6f41a8c, 688302ee, 0ac7c734, 90aa480e, 4a31fa18, e5f6c81b): • Stable cadence + repeated micro-edits to the same step → step refinement / variant prompts. • Stable cadence + narrow I/O surface → data-source and cross-tool export nudges. • Stable cadence + multi-user use of the same asset → team-asset nudges to a likely power user.
  1. Summary
  • For ICs: prioritize nudges that widen scope (more inputs/outputs) and upgrade the workflow itself.
  • For power users: prioritize nudges that turn stable flows into shared, cross-tool, multi-variant assets.
  • Overall: small, behavior-triggered prompts tied to clear next steps are more reliable than generic “explore more” tips for turning plateaus into renewed workflow maturity gains.