When users appear to hit a productivity plateau—stable cadence of a reusable workflow but flat or shrinking task-cycle time—what concrete, in-product behaviors (e.g., branching into variants, adding data sources, delegating steps to teammates) most reliably distinguish (a) a true ceiling in workflow maturity from (b) a latent growth phase where users are quietly building toward higher-value, cross-tool workflows that current metrics fail to capture?

anthropic-learning-curves | Updated at

Answer

Signals are mostly about variation, extension, and orchestration vs. rigid repetition and avoidance.

(a) Behaviors that suggest a true ceiling in workflow maturity

  • Runs stay identical
    • Same template, same parameters, same input types; almost no edits over many runs.
  • Narrow I/O surface
    • Inputs from one source only; outputs copied to a single destination; no new file types or fields.
  • Declining micro-edits
    • Fewer prompt changes, no step reordering, no added steps despite minor issues.
  • Bypassing adjacent features
    • Rare use of attachments, data connectors, or API/integration hooks that are nearby in the UI.
  • Solo, non-handoff use
    • No sharing, assignment, or comments; workflow never appears in team spaces.
  • Error avoidance instead of repair
    • When failures occur, users revert to manual work or ad-hoc prompts instead of evolving the saved workflow.

(b) Behaviors that suggest a latent growth phase (hidden upward curve)

  • Branching and variant creation without big time gains
    • Users clone workflows, create labeled variants, or A/B steps while cycle time stays flat.
  • Broader data and tool surface
    • Growing use of attachments, new data sources, or structured fields; outputs flow to more destinations or formats.
  • Localized step edits
    • Users repeatedly tweak the same step, add small pre/post steps, or change ordering while keeping the core flow.
  • Cross-tool orchestration traces
    • Short, regular in-product sessions tightly coupled in time with edits or activity in other tools (docs, BI, CRM), plus heavy copy/paste in/out.
  • Rising team signals
    • Increased sharing, comments, @mentions, assignments on runs; more distinct users invoking the same asset.
  • Asset-level evolution
    • Periodic template edits that propagate to many runs; new fields added; naming tightened (e.g., “QBR deck v3 – CRM+CSAT”).
  • Stable or slightly worse time with richer outcomes
    • Time per run flat or up a bit, but more sections, audiences, or channels covered in the output.

Practical rule: if behavior is dominated by unchanged repetition with shrinking experimentation, treat the plateau as a real ceiling. If you see structured variation, new connections to data/tools, and growing team touchpoints without matching time savings yet, treat it as a latent growth phase and support it with better cross-tool and team-level views.