EdTech Operator
Platform-agnostic AI co-pilot for any learning platform — works on top of Kajabi / Teachable / Thinkific / Skool / Mighty Networks / Canvas / Moodle / custom. Four flows on one dashboard: AI tutor that knows your course content with inline citations a click jumps to the right lesson, auto-grading that scores against your rubric and writes personalized feedback in your voice (encouraging / direct / technical), 30-student cohort with AI-flagged at-risk students + drafted nudges, and instructor-side content generation that turns a transcript / deck / PDF / audio into a starter pack (lesson outline + 5 quiz questions across difficulty tiers + 3 discussion prompts + summary card + rubric). Vertical toggle re-keys the entire UI for 4 buyer types: Course Creator (default) / Test Prep / Corporate Training / K-12 Academy.
How it works
1. AI tutor with inline citations + instructor escalation
Pick a course (Intro to Product Management / GMAT Quant / Anti-Bribery Compliance / Beginner BJJ). The student-side view shows lessons, progress, and a chat panel. Type a question — the tutor’s answer streams paragraph-by-paragraph with `[Lesson 3 ↗]` citation chips that scroll-jump and gold-flash the lesson list when clicked. Edge case: ask something the course doesn’t cover and the tutor says so + routes to the instructor with a 1-line summary; the side panel shows the instructor’s Approve / Edit / Send workflow.
2. Auto-grading with rubric + voice tone + bulk batch
Switch to Assignments — grid of 25 submissions with status (ungraded / graded / flagged). Pick one — rubric + submission text + tone selector (Encouraging / Direct / Technical). Click Auto-grade — scores against the rubric, highlights specific lines/sections, writes per-criterion feedback in the chosen voice. Bulk grade all 25 — counter ticks live (11.0 hours human grading → 22 minutes AI-assisted, 7 flagged for needs-attention review).
3. Cohort risk + content generation + audit-everything
Switch to Cohort — 30-student table sorted by completion-likelihood (6 at-risk 30-50, 10 mid 60-75, 14 high 85-95). Click an at-risk student — per-module progress bars + AI-recommended next 3 lessons with reasoning + drafted personalized nudge (‘Hey Sarah, noticed you’ve been quiet for 8 days. Module 5 is where most folks stall — want a 15-min check-in with the coach?’). Switch to Content — pick a source (transcript / deck / PDF / audio) — starter pack streams in (lesson outline with timestamps, 5 quiz questions across easy/medium/hard with answer keys + 1-line rationale per option, 3 discussion prompts, summary card, rubric). Each output has Regenerate-with-feedback (‘make this harder’ / ‘this is ambiguous’). Download as JSON for import into your platform. Audit log captures every action.
60-second Loom
75-90s flow: Course Creator dashboard with left-rail queues + 4-course catalog → click Intro to Product Management → tutor chat → ask MVP-vs-prototype question → AI streams answer with `[Lesson 3 ↗]` citation chip → click citation, lesson list scrolls + gold-flashes → ask out-of-scope question → tutor says ‘not covered, routing to instructor’ → side panel shows draft + Approve/Edit/Send → switch to Assignments → pick 1 submission → tone Encouraging → Auto-grade → rubric + per-criterion feedback streams in → Bulk grade all 25 → counter ticks 11.0 hrs to 22 min, 7 flagged → switch to Cohort → 30-student table → click an at-risk student → progress bars + drafted nudge → switch to Content → pick a transcript → starter pack streams in (outline + 5 quiz Qs + 3 prompts + summary card + rubric) → toggle vertical to Test Prep → entire dashboard re-keys (PMP / GMAT / Series 7 / MCAT, ‘cohort’ becomes ‘section’, ‘instructor’ becomes ‘tutor’) → mobile view at 375px shows stacked layout with bottom-tab nav.
See the dollar impact for your numbers
5 inputs → get hours saved/yr, dollars saved/yr, and payback vs. our $9,990 – $25k range. Results show on this page and the breakdown plus build spec hit your inbox.
Detailed breakdown plus the build-spec PDF have been sent to your email.
Who it’s for
Course creators on Kajabi / Teachable / Thinkific / Skool / Mighty Networks running cohorts. Test-prep companies (LSAT / GMAT / bar / real estate license / Series 7) building adaptive assessment runners. B2B training platforms delivering compliance / sales / skills training with completion accountability. K-12 / academy platforms (martial arts, music, tutoring centers) needing student management + engagement automation. Same architecture serves all 4 buyer types via the vertical toggle.
Tech under the hood
Frequently asked questions
Course Creator AND Test Prep AND Corporate Training AND K-12 Academy from one demo?
Yes. The architecture is single-tree multi-vertical: one component tree, one set of flow panels, four seeded vocabulary maps. Toggling re-keys the entire UI without re-rendering — instructor becomes tutor / L&D admin / teacher; cohort becomes section / team / class; course catalog re-frames per buyer type (e.g. Test Prep ships PMP / GMAT / Series 7 / MCAT instead of PM / Compliance / BJJ). Persisted in localStorage so a refresh keeps the chosen vertical.
Where does the LLM actually live? Is the citation real RAG?
The demo streams pre-written paragraphs from the seed (deterministic playback, zero credentials, zero cost). The ‘RAG’ is a small in-memory keyword-match: `lib/mocks/rag.ts` tokenizes the question, scores each lesson by overlap against the lesson’s `summary` (1×) plus its `keyConcepts` (2× — keyConcepts are curated, so they get extra weight); top-N lesson IDs become the citations and scroll-jump to that lesson. Pre-canned tutor scenarios already include `[CITE:lessonId]` markers in their answers, parsed at render time into clickable chips. Production RAG would use embeddings + a real vector DB (pgvector or similar) for semantic similarity beyond keyword overlap.
Does it integrate with Kajabi / Teachable / Thinkific / Skool / Mighty / Canvas / Moodle?
Not in the demo — out of scope per the brief. In production we wire to the relevant LMS / community platform via REST: Kajabi + Teachable + Thinkific + Skool + Mighty Networks for course-creator cohorts; Canvas + Moodle for higher-ed and corporate-training LMS; custom platforms via webhooks + REST. Adding a new platform adapter typically takes 3-5 days; the tutor + grading + cohort + content-generation layer is platform-agnostic (the demo’s `download as JSON` button is the production hand-off contract — generated content lands in your platform’s import format).
How does the bulk-grade time-saved math work?
Real, not hallucinated: `humanHours = ungradedCount × 30min / 60` (industry-standard ~30 min/essay), `aiMinutes = ungradedCount × 1min` (the AI-assisted human-review pass averages ~1 min per submission). The ticker increments live as each submission completes via the NDJSON stream. Edge cases (per-submission score below 55 or above 95) auto-flag for needs-attention review — the bulk-grade summary surfaces those first, sorted by score-distance from the rubric mid-point. Production builds replace the seeded ~30 min benchmark with the customer’s actual baseline.
Want one of these in your business?
We’ll scope, build, and hand off a production-ready system in 3–4 weeks.