Practice Hub App Overview
Rehearse the hard conversation before it’s real — Practice Hub is your AI practice coach for de-escalations, tough 1:1s, interviews, service recovery, and presentations, scored against a rubric peers actually wrote.
What is the Practice Hub App?
Practice Hub is the Practice layer of the workforce platform — the step between Learn (a Training course) and Do (the real conversation). An employee opens an authored scenario, holds a turn-by-turn conversation (typed or spoken) with an AI persona that plays the other person, and ends with a scored scorecard: per-criterion rubric scores, delivery metrics, and narrative coach’s feedback. Most scenarios are authored by peers — your best rep, nurse, or shift lead encoding how they actually handle a situation — and curated by L&D.
Core Value Proposition:
- 🎭 Rehearse under pressure — practice the behavior against a realistic AI counterpart, not a slide deck, before you face the real person.
- 🧑🤝🧑 Peer-authored first — your best performers encode tacit know-how into scenarios the whole team can practice; L&D curates the best.
- 📊 Capability, not completion — every attempt is scored against a weighted rubric with evidence quotes, so “I demonstrated it” replaces “I clicked through it.”
- 🔁 Closes the loop — passing an official scenario writes back to Training’s certification spine, connecting Learn → Practice → Do in one platform.
Key Concepts (How It Works)
If you only read one section, read this — it explains the model behind the screens.
The three layers: Learn, Practice, Do
Practice Hub is deliberately not Training. Training does Learn — watch a lesson, pass a knowledge quiz; it’s central, authoritative, and the compliance system of record. Practice Hub does Practice — rehearse the behavior under pressure and get scored on it. The boundary between them is about authorship and trust model, not quiz-vs-roleplay: knowledge compliance stays a Training quiz, while behavioral compliance and soft skills live in Practice Hub. The two connect — a passing official Practice Hub attempt records a completion back in Training (the Do layer is where you then use the skill for real, in apps like Service Desk, Comms, or Shifts).
Scenario, attempt, scorecard — the three core objects
Everything in the app is built from three primitives:
- Scenario (
PracticeScenario) — the authored unit: the situation and learner objective shown to the learner, one or more AI personas to roleplay against, and a weighted rubric the attempt is scored on. This is the reusable “definition.” - Attempt (
PracticeAttempt) — one learner’s run of a scenario. It holds the turn-by-turn transcript and moves through statusesin_progress → completed → scored(orabandoned). - Scorecard (
PracticeScorecard) — the graded result of a scored attempt: per-criterion scores with evidence quotes, a weighted overall, pass/fail, delivery metrics, and narrative feedback.
A scenario is the template; an attempt is one instance of practicing it; a scorecard is how that instance was graded.
What makes a scenario community, curated, or official (and what can certify)
Every scenario has a visibility that reflects how much it’s been vetted — this is the peer-content lifecycle, not a permission setting:
- Community — freshly peer-authored. Auto-screened by the AI quality-gate (safety, PII, off-topic, too-thin), but not yet human-reviewed.
- Curated — L&D reviewed it in the Curation queue (rating, usage, AI-check) and promoted it as a good scenario.
- Official — the organization stands behind it.
The load-bearing rule: only an official scenario can be marked certifiable. This is enforced in the model itself (a certifiable_requires_official validation), so community and curated content can inform practice but can never certify — the guardrail that keeps peer-authored content safe next to compliance requirements. Separately, status (draft / published / archived) controls whether a scenario appears in the library at all; an unpublished draft is visible only to its author and admins.
What makes someone a member, manager, or admin
Practice Hub has three role tiers, each determined by the platform permission you already hold — not a Practice-Hub-specific setting:
- Member — anyone with the app enabled. Can browse the library, run attempts, see their own attempts, rate scenarios, and (when peer authoring is on) author community scenarios.
- Manager — anyone who is
manager_or_above?or the app-admin. Gets the Assignment readiness board, Team skills radar, and (when manager review is on) the Reviews queue. - Admin / app-admin — anyone who is
admin_or_above?, plus anyone granted the per-app App Administrator capability forpractice-hubspecifically. Owns Curation, Analytics, and Settings.
So “manager” here means has direct reports or is an app admin — a permission, not a job title — and the app-admin is a targeted grant that lets someone run Practice Hub without being a full org admin.
Live roleplay vs recording analysis — two different things
Practice Hub has two ways to get a scored attempt, and they behave very differently:
- Live roleplay (
source: live, the default) — you converse with the AI in real time. These attempts count: they can be assigned by a manager, reviewed and overridden, and write back to Training on a certifiable pass. - Recording analysis (
source: recording_upload) — you upload a real call, meeting, or rehearsal recording (or pull an AI Notepad meeting transcript); it’s transcribed, speaker-labelled, and scored against a scenario’s rubric. These are private self-development: they never enter a manager’s review queue, never satisfy an assignment, and never write back to certification (an unverifiable transcript shouldn’t certify). Uploads are off by default and gated behind an admin toggle because they carry real consent/privacy weight.
Conversation vs delivery, and voice vs text
Two independent settings shape how a scenario is practiced:
- Interaction mode — conversation (default: back-and-forth with a counterpart) vs delivery (the learner holds the floor for a presentation, pitch, or speech, and the AI persona acts as an audience — reacting and asking questions, never a counterpart).
- Modality — a content choice about typing vs talking:
learner_choice(voice available, learner may type),voice_only(spoken practice is the point), ortext_only(written-comms practice — voice is never offered). Delivery scenarios default to voice_only; text_only is only ever set deliberately.
Whether spoken practice actually runs depends on both the scenario’s modality and the tenant’s live-voice toggle: voice appears only when live_mode_enabled is on for the business and the scenario is voice-capable (not text_only). Live speech-to-speech spends metered voice minutes, which is why it’s admin-gated.
How an attempt is scored
Scoring is an LLM grading the transcript against the rubric, engineered for a score you can trust:
- Evidence-first — the judge must quote the learner’s own lines for each criterion before scoring it (those quotes render on the scorecard as reviewable “game film”).
- Anchored — when a criterion has score anchors (what a 25/50/75/95 looks like), the judge places you in a band instead of inventing a scale each time.
- Median-of-3 — three independent evaluations run at temperature 0; the final per-criterion score is the median and the overall is the weighted average, so an outlier run can’t swing the result.
- Pass/fail — the weighted overall is compared to the scenario’s pass threshold.
- Delivery metrics — alongside the rubric, the scorecard reports light delivery signals (word count, filler words, and — on voice replies — words-per-minute and talk-to-listen ratio) plus total practice time. They inform; the rubric drives the score.
At a Glance
| ⏱️ Setup Time | 🔗 Integrations | 📱 Mobile Ready | 🗣️ Languages |
|---|---|---|---|
| ~5 minutes | 6+ apps | ✅ Yes | 15 |
Perfect For:
- 🛎️ Frontline & deskless teams — de-escalation, service recovery, and safety conversations rehearsed before the real customer or shift.
- 🎓 L&D and enablement — turn tacit know-how from top performers into curated, certifiable practice instead of another deck.
- 👔 New managers & internal candidates — practice difficult 1:1s, feedback, and promotion interviews with rubric-based feedback on structure, clarity, and pace.
How It Works
The Learn → Practice → Do loop
┌─────────────────────────────────────────────────────────────────────────┐
│ LEARN → PRACTICE → DO │
├─────────────────────────────────────────────────────────────────────────┤
│ ┌──────────────┐ ┌──────────────┐ ┌──────────────┐ │
│ │ LEARN │───▶ │ PRACTICE │───▶ │ DO │ │
│ │ (Training) │ │(Practice Hub)│ │ Service Desk │ │
│ │ lesson+quiz │ │ AI roleplay │ │ Comms/Shifts │ │
│ └──────────────┘ └──────┬───────┘ └──────────────┘ │
│ │ │
│ passing an official + certifiable │
│ scenario writes back a completion ▼ │
│ ┌──────────────────┐ │
│ │ Training cert │ │
│ │ spine (record) │ │
│ └──────────────────┘ │
└─────────────────────────────────────────────────────────────────────────┘
A single practice attempt
┌─────────────────────────────────────────────────────────────────────────┐
│ ONE PRACTICE ATTEMPT │
├─────────────────────────────────────────────────────────────────────────┤
│ Open scenario ─▶ Start practice ─▶ turn-by-turn conversation │
│ (situation + (type or with the AI persona(s) │
│ objective) speak) │ │
│ ▼ │
│ End & score ─▶ Scored scorecard │
│ • per-criterion + evidence│
│ • pass/fail vs threshold │
│ • delivery metrics │
│ • coach's feedback │
│ ▼ │
│ Ask the coach · "Show me a 95" · Drill a gap │
└─────────────────────────────────────────────────────────────────────────┘
The peer-content promotion ladder
┌───────────┐ AI quality-gate ┌───────────┐ L&D review ┌───────────┐
│ COMMUNITY │ ─────screens────▶ │ CURATED │ ───promotes───▶ │ OFFICIAL │
│ peer-made │ │ L&D-picked│ │ org-backed│
└───────────┘ └───────────┘ └─────┬─────┘
│
only OFFICIAL may be
marked CERTIFIABLE ▼
writes back to Training
Integration ecosystem
┌─────────────────┐
│ PRACTICE HUB │
└────────┬────────┘
│
┌────────────┬─────────────────┼─────────────────┬────────────┐
▼ ▼ ▼ ▼ ▼
┌───────────┐┌───────────┐ ┌───────────┐ ┌───────────┐┌───────────┐
│ Training ││Recognition│ │ Service │ │ Skills & ││AI Notepad │
│ (cert ││(creator │ │ Desk │ │ Certs ││(meeting → │
│ write- ││ points on │ │ (launch + │ │ (skill ││ scenario/ │
│ back) ││ promotion)│ │ signals) │ │ links) ││ analysis) │
└───────────┘└───────────┘ └───────────┘ └───────────┘└───────────┘
Key Features
🎭 AI Roleplay Runtime
Start a scenario and hold a real conversation with the AI persona(s).
| Feature | Description |
|---|---|
| Single & panel personas | A scenario can put up to three personas in the room (an interview panel, a buying committee); the AI picks who responds each turn and bubbles attribute the speaker. |
| Type or speak | Reply by typing, or record by voice — speech is transcribed, and the persona can reply with a synthesized voice. |
| Delivery mode | For presentations, pitches, and speeches, the learner holds the floor and the AI persona is the audience — reacting and asking questions on the wrap. |
| Live voice mode | An optional real-time speech-to-speech mode (admin-enabled, uses metered voice minutes) with a session cap. |
| 15 languages | Persona replies, transcription, voice playback, and scorecard feedback all follow the attempt’s chosen language. |
| Adaptive difficulty | The more times you’ve passed a given scenario, the harder the persona pushes on your next attempt. |
Use Case: A frontline associate opens “Calming an upset customer,” speaks their way through a de-escalation, and reviews the scorecard before facing the real thing.
💡 Pro Tip: Record by voice rather than typing — voice replies are the only way the scorecard can compute your speaking pace (WPM) and talk-to-listen ratio.
📊 Scored Scorecards & The AI Coach
Every finished attempt produces a scorecard — and the coach doesn’t stop there.
| Feature | Description |
|---|---|
| Evidence-first rubric scoring | Per-criterion 0–100 scores, each grounded in quotes from your own transcript, combined by weight into a pass/fail overall. |
| Coach chat | Ask follow-up questions about your scorecard — the coach explains its reasoning (scores are final; it clarifies, it doesn’t re-litigate). |
| “Show me a 95” | For any criterion you scored below the top band, see a verbatim exemplar of what a 95 would have sounded like and why. |
| Micro-drills | Turn a weak criterion into a focused ~3-minute drill scenario and jump straight into practicing just that skill. |
| Improvement deltas | “Your best yet” / “Up N from your last try” badges track progress across attempts. |
Use Case: A learner scores 62 on “acknowledged the frustration before problem-solving,” taps Show me a 95 to see a model opening, then drills that one criterion until it lands.
💡 Pro Tip: When a scorecard flags a weak criterion, use the Drill this button — it clones a private, single-criterion practice scenario so you can rep the exact skill without re-running the whole conversation.
🧠 Personalized Coaching
Coaching that remembers you and grounds itself in your workplace.
| Feature | Description |
|---|---|
| Coach memory | A coach profile builds up across your scored attempts (focus areas, strengths, a narrative) — and you can see and edit everything it “knows,” or reset it. |
| Workplace personalization | With the admin toggle on, coaching draws on cross-app signals (performance-review themes, recognition strengths, skills gaps) as reference material — it still scores only the transcript. |
| Company grounding | Admins can add a company context pack (products, common objections, policies, tone) so personas and scoring reflect your business, not a generic one. |
| Skill radar & recommendations | Attempts map to a 12-competency taxonomy; a “Your skills” radar shows your weakest areas and recommends scenarios to practice, with the reason why. |
| Weekly plan | A “Your plan for this week” card (and an optional Monday Inbox digest) suggests up to three things to practice, weakest-competency-first. |
💡 Pro Tip: Open the “What your coach knows” card on your dashboard — you can remove any focus area or strength that doesn’t fit, and the coach will note it again only if it shows up in future practice.
✍️ Scenario Authoring & Curation
Peers are the primary authors; L&D curates.
| Feature | Description |
|---|---|
| Peer authoring | Any employee (when enabled) can draft a community scenario: situation, objective, difficulty, tags, one-or-more personas, and a weighted rubric. |
| AI-assisted drafting | Describe the scenario in a sentence and the builder drafts a complete community scenario — personas, rubric, and score anchors — for you to refine. |
| AI quality-gate | New community scenarios are automatically screened for safety, PII, off-topic content, and thin rubrics before curation. |
| Curation queue | L&D ranks community/curated scenarios by rating, usage, and AI-check, and promotes the best to curated → official. |
| Starter library | Hundreds of ready-made starter scenarios (across sales, service, leadership, healthcare, safety, and more) can be imported from the MangoApps template gallery. |
| Ratings | Learners rate scenarios 1–5; ratings feed curation ranking. |
Use Case: A top performer authors “Handling the ‘it’s too expensive’ objection,” L&D sees it climbing the ratings, curates it, and eventually makes it an official, certifiable scenario.
🎬 Recording Analysis (Private Self-Development)
Bring a real conversation in for coaching — off by default, admin-enabled.
| Feature | Description |
|---|---|
| Upload & score | Upload a call, meeting, or rehearsal recording; it’s transcribed, speaker-labelled, and scored against a chosen scenario’s rubric. |
| Companion transcript | Attach a Zoom/Teams speaker-named .vtt/.srt for exact diarization and real names — the pipeline skips re-transcription. |
| Meeting-sourced analysis | Analyze an AI Notepad meeting transcript directly, with no file re-upload. |
| Private by default | Analyses never enter the review queue and never certify. You can opt in to share a specific analysis with your own manager. |
| Retention & purge | The raw file auto-deletes after a configurable window (default 30 days); the transcript and scorecard are kept. |
| Trend view | A Recording Analyses tab charts your recent analyses so you can see progression over real conversations. |
💡 Pro Tip: Attach the speaker-named transcript Zoom or Teams exports next to your recording — it gives the analysis real names and cuts processing time dramatically versus letting the AI guess speakers from a raw audio file.
📋 The Assignment Loop
Turn a working engine into used software — assign practice and track readiness.
| Feature | Description |
|---|---|
| Assign to people or a cohort | A manager assigns a scenario to specific users or a whole department, with a due date. |
| Inbox nudges | Assignees get an Inbox notification; a daily reminder job re-nudges the overdue (throttled so it doesn’t spam). |
| Readiness board | A funnel view — assigned → started → passed, plus overdue — with re-nudge, extend-due-date, and cancel. |
| Auto-status | Assignments move to started when the learner begins and passed when a scored attempt clears the threshold. |
👥 Manager Review & Team Readiness
| Feature | Description |
|---|---|
| Review & override | When enabled, a manager reviews a scored (live) attempt and can record an override that becomes the final score, with a note. |
| Team skills radar | On the readiness board, a per-competency view of the team’s average, with “below the bar” names and one-click assign. |
Note: recording-upload analyses are private and never appear in the review queue.
📈 Analytics & Insights (Admin)
Standard Analytics:
- Adoption, pass rate, and top scenarios (server-rendered, with CSV export).
- Shelf health — published scenarios bucketed active / going-stale / never-run over a 30/60/90-day window, so L&D can see which scenarios nobody is running.
- Practice by department — participation, attempts, pass rate, and overdue assignments per department.
- Assignment adoption funnel.
- AI usage — LLM calls, tokens, and cache reads by stage, business-scoped.
🔗 Integrations
| Integration | Capability |
|---|---|
| Training | Passing an official + certifiable scenario writes an external completion record to Training’s certification spine. |
| Recognition | Optional creator incentives — award Recognition points to an author when their scenario is promoted (admin opt-in). |
| Skills & Certifications | Scenarios link to skills (Workforce Readiness Graph), powering precise gap → practice recommendations. |
| Service Desk / Onboarding / Training | A “Practice this” launch widget surfaces relevant published scenarios from inside other apps. |
| AI Notepad | Turn a real meeting transcript into a practice scenario, or analyze the meeting directly against a rubric. |
| Performance / Recognition / Skills signals | Feed personalized coaching (when the personalization toggle is on). |
🤖 AI Agent (Ask AI)
Practice Hub ships a read-only help agent in the Ask AI sidebar (admin-enabled). It never starts or scores a roleplay — it helps you around the engine:
- Find scenarios — search published scenarios by keyword, category, or difficulty.
- Explain a scenario — its situation, objective, personas, and rubric.
- Review your own attempts — your recent attempts and their scores (never anyone else’s).
- Explain scoring — how the rubric, pass threshold, delivery metrics, and Training credit work.
🔔 Notifications
| Type | Description |
|---|---|
| Assignment due | An Inbox action-request when a scenario is assigned to you, with daily overdue re-nudges. |
| Weekly practice plan | An optional Monday Inbox digest of your top few things to practice. |
| Quality-gate flag | The author is pinged when the AI quality-gate flags their community scenario, with the notes and an edit link. |
| Promotion ping | An author is notified when their scenario is promoted up the ladder. |
| Manager review request | The learner’s own manager is notified when a certifiable attempt is scored (when review is on). |
User Roles & Permissions
| Role | Capabilities |
|---|---|
| Member (any employee) | Browse the Practice Library, run attempts, see My Attempts, rate scenarios, use the AI coach/chat/drills, and — when peer authoring is on — author community scenarios (Authored by me). |
| Manager | Everything a member can do, plus the Assignments readiness board, team skills radar, and — when manager review is on — the Reviews queue to review and override scorecards. |
| Admin / App-admin | Everything above, plus Curation (promote community → curated → official), Analytics, and Settings. |
How We Compare
See how MangoApps Practice Hub stacks up against leading AI roleplay platforms:
| Feature | MangoApps Practice Hub | Yoodli | Second Nature | Zenarate |
|---|---|---|---|---|
| Peer-authored-first scenarios | ✅ | ⚡ | ❌ | ❌ |
| Community → curated → official lifecycle + AI quality-gate | ✅ | ❌ | ❌ | ❌ |
| Weighted rubric scoring | ✅ | ✅ | ✅ | ✅ |
| Delivery / presentation practice to an AI audience | ✅ | ✅ | ⚡ | ⚡ |
| Upload & score a real call/meeting recording | ✅ | ✅ | ⚡ | ⚡ |
| Broad soft-skill + compliance + frontline breadth (not just sales) | ✅ | ⚡ | ❌ | ⚡ |
| Native to a full employee platform (LMS, Comms, Recognition, Shifts) | ✅ | ❌ | ❌ | ❌ |
| Webcam / video-avatar analysis | ❌ | ✅ | ⚡ | ❌ |
| Legend: ✅ Included | ❌ Not Available | 💰 Paid Add-on | ⚡ Limited / Different focus |
Why MangoApps Practice Hub?
- 🔗 Unified platform — practice is native to the same system as your LMS, comms, recognition, and shifts, so Learn → Practice → Do is one loop, not three vendors.
- 🧑🤝🧑 Peer moat — a real authoring-and-curation lifecycle (with an AI quality-gate) instead of a peer-sharing bolt-on, so the shelf reflects how your best people actually work.
- 🤖 AI-native coaching — evidence-first, anchored, median-of-3 scoring plus a coach that remembers you and grounds itself in your company — built in, not bolted on.
Competitor capabilities compared against publicly published product information (Yoodli, Second Nature, Zenarate).
Getting Started
For Employees
- Open Practice Hub from the app sidebar to land on your dashboard.
- Go to Practice Library, open a scenario, and read the situation and objective.
- Click Start practice, reply each turn (type or speak), then End & score my practice.
- Review your scorecard — ask the coach a question, tap Show me a 95, or Drill a weak criterion.
For Managers
- From Assignments, assign a scenario to specific people or a department with a due date.
- Track progress on the readiness board (assigned → started → passed) and re-nudge the overdue.
- Check the Team skills radar to see where the team is below the bar and assign practice in one click.
- If manager review is on, use the Reviews queue to review and, if needed, override a scored attempt.
For Administrators
- Enable Practice Hub for the business, then open Settings and confirm the toggles (peer authoring, manager review, AI agent — all on by default; recording uploads, live voice, and creator incentives are off by default).
- Use Load sample data to seed starter scenarios, or import from the template gallery.
- Optionally add a company context pack (products, objections, policies, tone) so coaching reflects your business.
- Use the Curation queue to promote the best peer scenarios, and Analytics to watch adoption and shelf health.
Best Practices
- ✅ Practice by voice when you can — it unlocks pace and talk-to-listen metrics that typing can’t.
- ✅ Curate actively — promote highly-rated, well-used community scenarios so the official shelf reflects reality; watch shelf health for scenarios nobody runs.
- ✅ Assign, don’t just publish — a working scenario nobody is told to run is shelf-ware; use the Assignment Loop with due dates.
- ✅ Reserve certifiable for official scenarios — the platform enforces it, but treat official promotion as a real quality bar.
- ✅ Add a company context pack so personas raise your actual objections and scoring reflects your policies.
- ✅ Keep recording uploads consent-aware — they’re private self-development; leave the toggle off unless your team understands two-party-consent expectations.
Frequently Asked Questions
Q: Who creates the practice scenarios?
A: Peers are the primary authors. Any employee (when peer authoring is on) can draft a community scenario, with AI assistance if they want it. L&D curates the best and promotes them to official — and only official scenarios can certify.
Q: How is Practice Hub different from Training?
A: Training does Learn — watch a lesson, pass a knowledge quiz. Practice Hub does Practice — rehearse the behavior under pressure and get scored on it. Behavioral and soft-skill practice feeds Training’s certification spine rather than replacing it; knowledge checks stay in Training.
Q: How does scoring work?
A: An AI grades your attempt transcript against a weighted rubric, quoting your own lines as evidence for each criterion and running three independent evaluations to take the median. You get per-criterion scores, delivery metrics (word count, filler words, and WPM/talk-to-listen on voice replies), narrative feedback, and pass/fail against the scenario’s threshold. A manager can review or override a live attempt.
Q: Can a peer-authored scenario certify me?
A: Only official scenarios can be certifiable. Community and curated scenarios are for practice and rating — a model-level guardrail keeps peer-authored content from certifying next to compliance requirements.
Q: Does it use my camera?
A: No. Practice Hub is audio-first — it uses your microphone and the transcript. There’s no webcam or eye-contact analysis.
Q: What happens to a recording I upload?
A: It’s transcribed and scored privately against a rubric — it never enters a manager’s review queue and never certifies. The raw file auto-deletes after your admin’s retention window (default 30 days); the transcript and scorecard are kept. You can choose to share a specific analysis with your own manager.
Related Resources
- Training App Overview — the Learn layer; Practice Hub complements it and feeds its certification spine.
- Recognition App Overview — peer recognition; the source of optional creator incentives when a scenario is promoted.
- Apps Overview — the full marketplace app catalog.
Practice Hub — rehearse it here, so you’re ready when it’s real.