Concept — The Automated Intervention Layer
A cross-cutting capability that wraps the entire Applied Learning Cycle and fires on the absence of expected behavior at each phase — not its presence. This is the operational expression of the core design principle.
“% reporting use of a strategy within 24 hours, 48 hours, 5 days?” — v3 strategy. Those thresholds are interventions waiting to be built.
Trigger table
| Phase | Stuck signal | Triggered response |
|---|---|---|
| Prepare | Pre-work incomplete <24 hrs before engagement | Auto-nudge participant; flag on IS + LA dashboards |
| Practice | No action plan submitted within 24 hrs | Auto-nudge participant; IS sees gap on dashboard |
| Apply | No application within 24h / 48h / 5d | Microlearning + reminder cascade; if attempted+failed → targeted support [G3] |
| Reflect | Survey not complete | Auto-reminder; certificate gated behind survey; LA + IS notified |
| Adjust | Drift in repeat application (week 2–6) | Spaced nudges; LA reinforcement prompt; IS supplemental release |
| Adjust | At-risk cohort patterns across engagements | PM flags to IS + Account Owner; high-perf cohorts → case study |
The three application branches (the heart of it)
At the Apply gate, behavior splits three ways — each with a distinct response:
- Applied quickly (≤48h) → reinforce + extend (unlock extension resources).
- Delayed (>48h) → trigger nudge cascade (microlearning + reminder, [G2]).
- Attempted but failed → targeted support ([G3]: microlearning + IS office hours).
This branching is what separates a reminder system from a coaching system.
Role-specific responses
The same stuck signal routes differently by role (from the user-stories doc):
- Strategist → targeted follow-up (async reply to reflection, 1–3 scheduled nudges [G2], microlearning rec [G3])
- Learning Admin → reinforcement prompt
- Participant → nudge or microlearning
Empathy in the loop
The trigger decides when the layer fires; empathy decides the tone of what it says. Every absence the layer detects is also an emotional moment for a real, tired, time-pressured teacher — so the same trigger can land as support or as surveillance depending entirely on the wording. This is the bridge between the mechanism above and the personas & empathy map (visual page).
| Phase | Trigger (absence) | What the teacher feels | Tone that honors it | Message |
|---|---|---|---|---|
| Prepare | Pre-work incomplete <24h | “here we go again” / behind | low-pressure, time-honest | “3 focus points, ~2 minutes — from a real classroom.” |
| Practice | No action plan <24h | exposed, unsure | low-stakes, no score | “Just one class. Jot what you’ll try — no wrong answer.” |
| Apply | No application 24h/48h/5d | busy + guilty | care, not a countdown; offer an out | “When you get a quiet minute this week, try it once. Not this week? That’s okay.” |
| Apply | Attempted but failed | discouraged | normalize, route to help | “Tried and it didn’t land? That’s the job. Here’s a hand.” (→ targeted support, never a failing grade) |
| Reflect | Survey not complete | done / depleted | brief, appreciative | “Two quick prompts when you’re ready — they help your strategist help you.” |
| Adjust | Drift in repeat application | fading momentum | warm, spaced, celebratory of progress | “You’ve done this 3 times now — it’s becoming yours. One more this week?” |
The non-negotiables that keep a nudge on the right side of that line: - Care language, never deadline language. “Best while it’s fresh,” not “overdue.” - Always an out. A participant self-snooze (“not this week”) is worth more for trust than any backend cap. See nudge fatigue. - Distinct copy per Apply branch. A teacher who tried and failed must never get the same message as one who ghosted — collapsing them is the fastest way to lose trust. - The IS sees patterns, the teacher feels privacy. Stuck-detection drives cohort weather for the strategist, not a watchlist — the teacher is never named on a screen. (Crossover thread “don’t make us feel judged” — see empathy.)
Design rules
- Fire on absence, not presence. The layer watches for expected behavior that didn’t happen.
- Bounded nudging. The IS schedules 1–3 follow-up nudges [G2] — not an unbounded drip. See nudge fatigue question.
- Every phase pairs a produced behavior with a trigger. If a phase has no defined trigger, the cycle contract is incomplete.
Dependencies
- Needs the 24h / 48h / 5d timing data → time-to-apply metrics.
- Needs a microlearning library to point at → Content Creator (roles).
- The “Mark complete” IS action opens the application-tracking window that starts the Apply-phase timers.
Open questions
- How are 24h/48h/5d windows anchored — from “mark complete” timestamp, or per-participant?
- Are nudges in-app, email, or both? (v2 shows “auto-nudge: what in-app · flag to IS + LA”.)
- See open-questions.md.
Related
- Source: §6.1 of recommendations doc; intervention band in Figure 1
- Built in Phase 3 of the roadmap, though the Apply-gate nudges land in Phase 1.