Repository navigation
💡 Progressive Workflow Suggestions: Avatar 'What's Next' Pathways for Day-2 Retention #517
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Uh oh!
There was an error while loading. Please reload this page.
Summary
After completing any workflow, the avatar suggests 2-3 logical next steps based on what was just accomplished — "Great brainstorming session! Teams typically organize these ideas into a PRD next. Want me to help with that?" This creates natural workflow chains that guide non-technical users from one-time use to habitual engagement, addressing the Day-2 retention problem that first-session onboarding alone cannot solve. Suggestions are gentle, contextual, and never persistent — the avatar celebrates completion first, then offers a path forward.
Market Signal
Duolingo's streak mechanics, LinkedIn's profile completeness prompts, and Notion's template suggestions demonstrate that post-action progression cues are proven retention drivers in consumer products. ChatGPT's new voice agentic features (Sep 23, 2026) focus on task execution but offer no post-task guidance — the conversation ends when the task ends, leaving users to figure out what to do next on their own. Microsoft's 2026 Work Trend Index identifies the biggest productivity gap as not individual task completion but workflow continuity across related tasks. No AI agent connects completed work to logical next steps via avatar conversation.
User Signal
The onboarding initiative (#333-338) covers the first session extensively with a 5-step tutorial producing a real artifact. But the PRD's daily usage target (≥1 session/day within 30 days) requires users to return independently on Day 2+. Existing discussions cover workflow templates (#244) as starting points but nothing addresses what happens after a workflow completes. The gap between "completed first workflow" and "daily user" is where adoption stalls — users finish their task, close the app, and don't know what else TalkTerm can help with. Progressive suggestions bridge this gap by showing users the breadth of TalkTerm's capabilities through their own work context.
Technical Opportunity
BMAD workflows already have a natural progression: brainstorming → PRD → architecture → UX design → epics → stories → implementation. This progression map is implicit in the planning artifacts (
_bmad-output/planning-artifacts/) but not exposed to users. Implementation: define aworkflow_progression.jsonmapping each workflow type to 2-3 logical successors with descriptions and conditions ("if brainstorming produced ≥10 ideas, suggest PRD; if <10, suggest another brainstorming technique"). Avatar presents suggestions as ActionCards (UX-DR2) in the left panel after the output summary (FR26). Starting the suggested workflow carries forward the output artifacts as context. Touches the agent bounded context (main/agent/) for progression graph and the renderer (components/) for suggestion display.Assessment
Adversarial Review
Strongest objection: Post-workflow suggestions could feel pushy or patronizing — like an upsell prompt at checkout. Non-technical users who just completed a challenging workflow might want to stop, not be nudged toward more work.
Rebuttal: Suggestions are offered once, gently, after the output summary (FR26) — not as a modal, not as a blocking prompt, and not as a notification that follows the user. The avatar frames suggestions as a helpful observation: "By the way, teams usually..." not "You should do X next." Dismissing is effortless (ignore, say "no thanks," or close the panel) and the suggestion never repeats for the same session. Critically, the avatar acknowledges completion first: "Great work on that analysis! Here's your output." Only then, after a natural pause, does it offer next steps. Tone is celebratory, not demanding — like a knowledgeable colleague sharing context, not a product manager pushing KPIs.
Suggested Next Step
Map the BMAD workflow progression graph (brainstorming → PRD → architecture → UX design → epics → stories) with branching paths and conditional transitions. Design the avatar suggestion conversation script with the celebratory-then-suggestive tone pattern. Define success metrics: Day-2 return rate for users who receive suggestions vs. those who don't (A/B framework for post-launch measurement).
All reactions