Among users who have reached independent execution on at least one reusable workflow, which specific onboarding designs that expose “next-level” capabilities (e.g., timed walkthroughs that appear only after N successful runs, tiered checklists for adding parameters or data sources, or mentor-style recommendations from power users’ workflows) most reliably convert early workflow maturity into cross-task generalization—measured by growth in the number of distinct tasks covered per user—without increasing abandonment or error rates?

Answer

Three patterns look most promising, mainly in combination:

1. Run‑gated, optional “upgrade” walkthroughs

• Trigger only after N successful runs of the same reusable workflow.
• Very short (1–3 steps), focused on one upgrade: add a parameter, attach a data source, or save a variant.
• Clear skip/exit; default keeps the existing flow.
• Best design: inline hint on run completion ("Want to reuse this for similar tasks?") that opens an overlay showing the current workflow with suggested fields highlighted.
• Effect: raises distinct-task count by turning a single-task flow into a small family (e.g., multiple clients, channels) with minimal added errors, because users start from something they already trust.

2. Tiered checklists tied to real runs, not setup

• Lightweight checklist visible near the workflow run button, unlocked over time:
  1. "Make key fields parameters"
  2. "Add a second input source"
  3. "Create a variant for a different audience/tool"
• Each item is clickable, runs a tiny guided edit, and can be ignored with no penalty.
• Checkmarks are based on observed behavior (e.g., user actually added a parameter), not clicks in an abstract tutorial.
• Effect: steady increase in distinct tasks per user as workflows cover more audiences, channels, or data without a big spike in abandonment; users act when they feel the gap ("I wish I could reuse this for X") and see a matching checklist item.

3. Mentor-style, asset-based recommendations with strong filters

• Surface 1–3 power-user workflows as patterns after the user has run their own flow several times.
• Match on task and structure ("other people who automate this step also do…"), not just popularity.
• Provide a “copy structure, keep your content” action that imports steps/parameters into the user’s workflow instead of pushing them to a new template.
• Effect: boosts cross-task coverage when users are close to a new use case but don’t know how to structure it; copying structure limits error risk because users keep their known prompts and data while borrowing decomposition.

Design features that matter for not increasing abandonment or errors

• Triggers are based on success (N correct runs, low recent corrections), not first use.
• Interventions are optional, local, and reversible (undo / keep old version).
• One upgrade per nudge; no long tours.
• Measured by: increase in number of task types per user, with flat or improved correction and rerun rates.

Relative reliability (from current mixed evidence and analogous patterns)

• Most reliable: run‑gated upgrade walkthroughs + tiered checklists, triggered on stable, repeated use of a workflow.
• Useful but more situational: mentor-style recommendations, especially in orgs with visible power users and shared assets.
• Least suitable for this stage: broad, first-session feature tours or generic galleries; they tend to add confusion and drop-off without clear gains in distinct-task coverage.