Scenarios And Progress

Scenarios let learners practise decisions, conversations, judgement, escalation, compliance steps, or workplace responses.

In-course scenario practice requires an active paid plan and a published course version. The workspace Free Trial does not include scenario authoring or in-course practice; public offerings and public AI Practice trials follow the separate rules below. If enabling is blocked, the course keeps its current setting and explains whether to upgrade or publish first. Enabling unlocks setup; it does not generate or release scenarios. Complete the scenario scope and objective focus before generating, then review and approve the results before learner delivery. Disabling remains available, and existing scenarios are preserved if plan eligibility changes.

Create A Scenario

Click Generate Scenarios to open its options. Choose a Scenario Type: Workplace Practice (the default) or Interview. Workplace Practice rehearses the objective in its configured work context; Interview uses a candidate and hiring interviewer. The type sets the practice format for that generation; your course scope and objective focus still supply the subject, audience and boundaries. Changing the selector does not change existing scenarios. The type and instruction version are saved with each generated draft. Set Maximum Scenarios, then click Generate inside the options panel. Interview creates a standalone hiring conversation; Workplace Practice retains course scenario sequencing. Review drafts before making them available to learners.

1. Open Dashboard > Scenarios or open the course and go to its scenario workspace.
2. Choose the course.
3. Add a scenario manually, or generate draft options from approved course context.
4. Define the learner role, workplace situation, goal, risk points, and expected outcome.
5. Review or write the learner narration script used by the scenario speaker control.
6. Build each scenario step with a learner prompt, good-response signals, serious mistakes, optional choices, and optional example answers.
7. Use Order Matters only when signals must be satisfied in sequence.
8. Review source metadata, facts used, assumptions, and reviewer notes where available.
9. Save or approve the scenario when it is ready for learner practice.

Standalone Scenario Delivery

Some scenarios can be published as scenario-only practice outside the course player. Configure standalone delivery, access type, URL slug, and published status only after review.

Public scenario offerings can be shared or monetised independently of the organisation’s course-authoring subscription. Public launch does not require an active paid course-authoring plan or the course’s Scenario Learning Enabled switch. The course must still be published with approved scenarios, and the public Expert profile must be active, enabled and configured to allow the visitor. Public Free access and members of the owning organisation use organisation funding. Other permitted non-free visitors must sign in and have personal paid AI credit; trial credit does not qualify. Launch checks credit but does not itself charge for AI usage.

Public AI Practice Trials

Public AI Practice on /learning/ai-roleplay is a temporary try-before-sign-up experience, not learner history. Its Coaching and Roleplay trials in Chat and Live Voice keep the visible conversation only in the current browser tab. Refreshing that tab can restore the visible trial, but closing or clearing the tab, opening the URL in another browser context, or signing in does not restore or transfer it. A Live Voice reload never restores an active microphone or connection.

Public trials still apply eligibility checks, abuse protection, anonymous quotas, rate limits, billing/usage measurement and privacy-safe operational telemetry. Those controls do not retain the visible dialogue as an attempt or turn history. Authenticated learner AI Practice remains account-owned and durable; it never imports an anonymous public trial.

Both practice approaches show two participant cards: the learner and either the assigned roleplay counterpart or their coach. The learner card uses their saved account profile photo, with their sign-in photo as a fallback. If no photo is available, it shows their initials rather than a fictional character portrait. The assigned counterpart card keeps its profile portrait. There is no separate avatar or voice picker: roleplay uses the approved voice saved with that counterpart profile. The speaker control sits with that other person’s details, so learners can listen to the intro instead of reading it. Audio starts only when the learner clicks the control.

Scenario generation selects an explicit fictional counterpart from the system-managed catalogue alongside the roleplay context. Each catalogue counterpart has a locked synthetic portrait and approved voice; generation selects its ID rather than inventing a new identity, portrait, or voice. The selected counterpart stays fixed for the learner attempt.

The counterpart card shows their scenario role, including their job title and company when supplied by the scenario. Their name, portrait and voice come from the assigned profile. Chat and Live Voice receive both participants’ scenario roles so the counterpart responds from their own position in the conversation.

Chat labels make the speaker clear: You and the assigned counterpart’s name in Roleplay, and You and Coach in Coaching.

Scenario generation decides who speaks first from the specific roles, situation and conversational purpose, without fixed rules for particular domains. In chat, a counterpart-first scenario shows their opening dialogue before you answer. In Live Voice, Start conversation plays that opening in the assigned counterpart’s voice, then starts listening. Learner-first scenarios start listening immediately. The optional intro explains the situation separately from the conversation. Reconnecting restores up to six recent delivered conversation turns to the new Realtime conversation. The model uses that context with the active scenario move to make a short in-character handoff. In Roleplay, the counterpart then asks a natural question to continue the conversation; in Coaching, the coach returns to the current practice move. It is not a new conversation turn and does not affect progress. Resetting starts the opening again.

In Roleplay, speak naturally: your counterpart responds when you finish, including to short replies. Scenario practice currently accepts English responses; unsupported-script input is not sent as a conversation turn. Listen to intro is a separate audio player for the optional setup, spoken in the assigned counterpart’s approved voice. It is available in Chat and Live Voice, before or during a session, and never selects a communication mode or starts a live connection. Use the microphone button to mute or unmute; Pause session stops the connection between turns. The counterpart follows the same character, facts and disclosure rules in chat and live voice, and responds in character rather than explaining what you should say. Resetting asks for confirmation once; cancelling keeps the current session.

Roleplay uses a separate counterpart action set: respond in character, hold the current beat, advance, close the conversation, express an authored disposition change, or redirect an unrelated message in character. Safety checks remain a guard around the conversation rather than a coaching response from the counterpart.

Live Voice keeps its active controls above the conversation transcript, so learners can review the exchange while they practise. On desktop, the larger circular player floats over the page and can be dragged directly; dragging changes its position for the current visit. In standalone practice and the course player, setup and messages scroll within the practice area while the centred player stays visible, including on mobile. The text composer is hidden while Live Voice is selected. Its icon, colour and motion show readiness, listening, speech, response preparation, speaking, failure and completion. The player has accessible labels. A readable error is shown only if voice fails.

In coaching, use the left and right arrows on the coach card to browse characters. Each coach shows a short description of their voice. The arrows cycle through all available characters, and the portrait and voice change together. The selection is shared across scenario practice and the course player in this browser, including after a refresh; it is not an account-synced preference. It controls live coaching speech and briefing audio independently of the course narrator. Change it before connecting, or return to chat before choosing another character.

The participant cards are followed by a short setup: the learner sees their next move, chooses Coaching or Roleplay, then chooses Chat or Live Voice. The wider scenario guidance is available from Open Scenario Brief, which opens a scrollable right-side drawer. Your Move remains visible; the goal, context, guidance and watch-outs are compact icon-led sections that expand on demand. After the first submitted turn, the practice approach becomes a fixed icon-and-label badge for that attempt and remains locked after refresh. Chat remains the default view. Both coaching and roleplay support switching between chat and live voice while retaining completed turns and progress; roleplay retains the same assigned counterpart. Finish a spoken turn before switching, because unfinished speech is not a saved turn. The separate short voice-message recording control remains available in the composer. During an active voice connection the coach character arrow selector is locked; the cards condense to a persistent header. Learners can reset practice to return to the opening state. Reset remains available when progress exists after Live Voice is paused, disconnected or completed, including when reopening saved practice. It is temporarily disabled during a voice response or transition. Confirmation permanently clears the current objective’s scenario chat and progress; cancelling keeps them. Live Voice uses the organisation’s Live Voice Practice capability, and public roleplays also require the owning Expert profile’s Live Voice setting—there is no per-scenario setting. The assigned counterpart profile determines the roleplay voice. The active participant card highlights while it is that person’s turn. Returning to chat is temporarily disabled while a voice response or transition is in progress. Switching communication mode does not change the scenario’s authored content.

In coaching Live Voice, you can keep speaking while the coach prepares its response. Additions and corrections before playback starts amend the same answer; the coach prepares a replacement using the original answer and your additions in order. During playback, possible speech briefly pauses the coach while it is checked. Background noise and filler let the same reply continue; confirmed speech interrupts it without adding an interruption notice to the conversation. If a pending response cannot be safely replaced, voice stops and shows a reconnect option. Work already performed may still use AI credit. During longer waits, the coach may make a brief remark based on your latest statement and recent conversation. The AI chooses whether anything useful can safely be said, rather than cycling through fixed phrases. These remarks stop when you speak or the answer is ready and are not saved as coaching messages.

Review Learner Progress

Admins and trainers can review scenario conversations, completion status, missed signals, covered signals, and risk flags from the scenario workspace, course scenario threads, and analytics views.

In Dashboard > Analytics, choose a From and To date and select Apply to update every analytics tab and its learner drilldown with the same reporting period. Scenario Hotspots show the beats and expected signals learners miss most often; use View Learners to see each affected learner’s misses and whether they later achieved that signal.

Use this review to decide whether a learner needs coaching, a retry, a micro-lesson, assessor review, or no action.

If Many Learners Miss The Same Signal

Review the scenario wording, source material, prerequisite content, or micro-lessons before treating the issue as individual learner performance.

In learner AI Practice, the floating Live Voice player opens centered within the practice content container. Desktop learners can drag it within that container; returning to Live Voice restores the centered starting position.

Scenario practice request-rate limits apply to anonymous visitors. Signed-in session owners are exempt from application rate limits on scenario launch, chat/voice turns and Live Voice connections. Access, applicable course or public-offering eligibility, wallet billing and duplicate-request protections still apply.

Public Paid practice requires paid AI credit in an external learner’s personal wallet. Signing in removes anonymous rate limits but does not make paid practice free. If personal paid credit is exhausted, starting practice or connecting Live Voice shows a message explaining the organisation’s paid setting and asking the learner to add credit. Trial credit cannot fund Public Paid practice. Public Free practice and members of the owning organisation use organisation funding.

When your personal paid AI credit is insufficient, the error alert offers Add AI Credit, which opens learner billing. Anonymous visitors can choose Sign Up to Add AI Credit and continue to billing after authentication.

Live Voice coaching uses Realtime by default. After you finish speaking, the coach or counterpart can begin a natural reply while the learning system evaluates your turn. Its confirmed result guides a later reply and step state; it does not rewrite audio already underway. The chat records the actual delivered transcript. If the Realtime response cannot complete, reconnect and retry.

With Realtime coaching and roleplay, small noises do not cancel the stream. Confirmed speech interrupts it. Interrupted turns do not insert an interruption message.

If your microphone stops providing audio during Live Voice, reconnect after checking microphone access. If Finish turn reports that no speech was detected, check your microphone and speak again, or pause the session.

For provider-funded practice, insufficient AI credit means the provider needs to add credit. Contact the provider admin for more information; topping up your personal wallet does not resolve a provider-wallet shortfall.

In Realtime coaching, the coach briefly signals when it is your turn after you start. Small sounds that are rejected as noise do not restart the preparation of your coaching feedback.

Coaching and roleplay accept brief transcribed words when speech was detected, including “Yes”, “No”, “Stop”, “Wait” and “Why?”. Empty input, hesitation-only sounds and common laughter-only transcripts such as “haha” are ignored. Words mixed with laughter still count as speech. Before playback, accepted words amend your answer; during playback, they interrupt the current reply. “Stop” interrupts the reply; use Pause session to disconnect voice.

In both coaching and roleplay, Live Voice keeps the conversation moving while it evaluates a turn. Realtime follows the scenario role and safety boundaries immediately; later learning guidance shapes a following reply. Roleplay stays in character and never introduces coaching advice, hidden criteria, facts or commitments that are not part of the scenario. The conversation transcript contains only delivered turns.

Both modes keep listening while preparing a reply. If you add or correct something, your words are combined with the pending answer and its undelivered reply is replaced. Discarded replies do not advance the scenario or change the roleplay counterpart’s disposition.

Live Voice starts preparing a reply as soon as your completed speech is accepted, without an added 1.5-second wait. You can still add or correct something while it prepares. More separate backend requests may be needed if you continue after preparation starts.

Scenario Brief and Listen to Intro sit beside the practice options, wrapping on smaller screens. In Chat, the message box stays at the bottom while the conversation above it scrolls.

Listen to Intro is separate from the live conversation opening. Anonymous public trials can use it while their browser trial remains valid and the course and public profile remain available. Starting Live Voice still uses Realtime for its opening, not the intro narration player.