Demo a personal AI assistant

Status: Current | Last reviewed: 2026-10-20

Create a Scene that follows a personal AI from its owner's phone to a contact center, then into browser chat and a human representative's Salesforce console. The included return journey supplies editable dialogue and all nine chapters.

Add the return journey

  1. Open a Scene agent in Director and select Stage.
  2. Under Personal AI assistant, select Add Ralph Lauren return journey.
  3. Keep or remove earlier chapters using Chapter options. Adding the journey preserves them.
  4. Rename the Scene and edit the script to suit your demonstration.

The starter uses Alex Morgan, an assistant named Ari, and a damaged sweater with missing tags. Its return options are example demo content; edit them for your scenario. The generated Ari portrait is included.

Make a fictional-brand copy

Use Duplicate in the Agents library to create an independent Scene with its own recordings. In the copy, change the brand and product details, assistant organization and website, dialogue, authentication text, and Salesforce Stage → Branding. Update the records, sidebar, incoming-work context and wrap-up summary to match the new request. The wrap-up's Customer Intents field can describe the new request separately from its case reason. Then choose Generate missing takes to prepare the revised speech.

The Alder & Branch · Personal AI return exception example follows Alex's return of an unworn Weekender Canvas Jacket that is too small. It was delivered 45 days ago, outside the fictional brand's 30-day return window, with tags still attached. The representative approves store credit.

This copy includes a 1.5-second pause between the IVR transfer prompt and the first Service AI greeting. To adjust it, edit the IVR-to-AI handoff Pause moment's duration. The preceding prompt has no additional trailing gap.

The copy includes opening narration, a narrated Agent Assist spotlight, and callouts for identity verification and the owner's final choice. Edit the Narrator persona in Voices, and its lines in the script. Keep Who hears this line? → Audience only · narration so the explanation remains separate from the call participants.

Configure the experiences

In Stage → Personal AI assistant, edit the assistant name and portrait, owner, contact center, Service AI and human representative. You can also change the session identifier, six-digit code, website, queue explanation and permitted resolutions. The app appears on the phone home screen when enabled. At the start of the journey, its selected icon grows slightly before the app opens.

Use Assistant experience to select the chapter's view:

View Use it for
Owner’s phone · Personal assistant Owner requests, thinking, status, authentication and decisions
Assistant call · IVR and Service AI IVR navigation and the conversation between two AIs
Assistant browser · Session and chat Website, six-digit code and the continuing chat
Agent Assist → Salesforce Human Messaging Session, policy options and Conversation Wrap Up

Use Experience label to name the perspective shown at the top of each assistant phone, call or browser chapter.

Choose Ralph Lauren as the Salesforce preset for the starter layout. The left sidebar identifies the personal AI below the owner's profile and above insights. The center contains the Messaging Session conversation. Agent Assist shows the permitted resolutions and transfer summary on the right. The conversation uses Salesforce's message bubbles, timestamps, reply toolbar and End Chat control. Select Details to see the session's status, start time, origin and related case. Human messages are blue, earlier Service AI messages are dark gray, and incoming messages from Ari are light gray. In Ari's browser, her outgoing messages are dark gray.

Changing the session code updates the starter's spoken sessionCode detail. Leave an Enter six-digit code moment's code blank to use that setting. An explicitly entered code overrides it. A wrong or incomplete code prevents the chat from connecting and produces a timeline notice.

For a Connect to the same session browser action, enable Animate connection. Set Pause before click and Connecting duration; the moment grows to fit them. The completed code enables the button, which visibly presses and shows Connecting… before opening chat. A cursor cue can target assistant.browser.connect; set its click time to the same delay.

For a policy spotlight, select Visual cue → Spotlight, pick the resolution cards, and adjust Background dimming, Zoom to spotlight, and Ease in and out. Use Advanced timing to anchor cues to the narration or a screen action, so they follow timing edits.

Edit messages, thinking and actions

Use + → Personal AI assistant between script moments:

Edit Voice & timing or Edit moment to change when each event occurs. Owner messages type into the composer before sending. The owner's message box wraps and expands as Alex types, up to six lines; longer drafts scroll within it. It shrinks again after sending. While recording interactions, Enter sends and Shift+Enter inserts a new line. Thinking and working status remain visible for their moment's duration. Working actions appear near the portrait at the top. Edit customerFirstName in Details to change the greeting without changing the full customer name on records. The starter sends AI chat messages every 1.5 seconds. Ari types words quickly into the browser composer; the other AI shows a thinking bubble before replying. Human responses type character by character after a short pause. Both composers show a caret, briefly hold the full draft and visibly press Send before the message appears in the conversation. For a service-chat message, enable Type before sending and set Pause before typing (seconds). The moment's Duration includes both the pause and typing; the message sends at its end. Leave Queue explanation empty to hide the additional queue note.

For direct authoring, pause playback and enable Record phone interactions or Record workspace interactions. Open Ari on the home screen or send a message in the preview. These actions become saved moments at the playhead. Add the assistant's scripted reply and thinking moment afterward.

Prepare voices and IVR audio

  1. Open Voices. Choose the personal assistant's voice under Caller and the contact-center AI's voice under Agent.
  2. For the Automated phone service persona, choose Amazon Polly, an Amazon connection, region and voice. Standard · Classic IVR supplies a conventional IVR voice; Neural is also available.
  3. Select Generate missing takes to prepare speech, DTMF and preset sounds. Saved takes are reused in playback and MP4 exports. Recorded-audio moments require a file upload in their moment settings and are skipped by voice generation.

The Virtual Keypad visibly presses each authored digit while its prepared tone plays. The call transcript also shows Ari dialing and pressing the digits. The starter's transfer to the Service AI has no hold music. Its Omni-Channel offer uses the same incoming ring as phone demos. To use your own IVR prompt or music, add a sound, choose Recorded audio / hold music, and upload an audio file up to 8 MB and 120 seconds. Use the sound's volume control to balance it against speech.

Rehearse the handoff

The starter follows this sequence:

  1. The owner asks Ari to arrange a return; Ari thinks, replies and starts a call.
  2. Ari presses IVR keys and reaches the Service AI. After Ari's request, an clearly labeled AI Activity card shows the secure order lookup, its running status and expanded result immediately after Ari gives the order number.
  3. The Service AI requests authentication permission. Ari checks with the owner, showing a thinking bubble before the question. One second after the owner agrees, the secure text arrives. Completing authentication returns to the call.
  4. The Service AI confirms verification and offers a digital continuation.
  5. Ari opens its standalone browser and enters the six-digit code. The Service AI confirms the linked chat over voice, then ends the call. The starter displays ralphlauren.com/chat/a2a without a voice-call banner.
  6. The AIs discuss the damaged item and missing tags. The chat enters a queue with eight humans and two AIs ahead.
  7. Omni-Channel rings and displays the incoming chat for 4.5 seconds, then shows acceptance. Its context contains the name, personal-AI identity, order and reason for transfer. The human types the permitted resolutions.
  8. Ari shows a thinking bubble before asking its owner; the owner chooses an in-store return for store credit.
  9. Back in Salesforce, Ari types and confirms the choice. After the final exchange, a cursor clicks End Chat. Conversation Wrap Up uses the Allianz Call Wrap Up layout: case details, a bulleted conversation summary, recommended actions, Add Task and Mark as Complete.

The wrap-up recommends Flag this conversation as initiated by a customer's AI assistant.

Seek between chapters to check continuity, then choose Export. Scenes play saved scripts and simulated actions; they do not contact a retailer or navigate the assistant browser to a live website.

See also