Get started with Coworker Chat in Playground

Use Coworker Chat in Playground to explore how natural-language prompts can help you learn, investigate, and refine your work. Playground provides sample ways to start a conversation so you can quickly understand what Coworker Chat can do.

AI Coworker Demo: weBank Credit Card Upgrade Campaign

This exercise covers Real-Time CDP and Adobe Journey Optimizer.

You’re a CX Enterprise strategist at weBank, a digital-first financial services company. Use the AI Coworker to go from raw data to a live campaign — entirely through conversation.

Part 1: Understand Your Business

1.1 — Get the lay of the land

Prompt

Give me an overview of my sandbox — how many audiences, datasets, and destinations do I have?

Expected result

Coworker returns a full sandbox inventory in seconds:

1.2 — What audiences already exist?

Prompt

List all my audiences sorted by size

Expected result

Coworker lists 21 audiences sorted by size, from 8,012 profiles down to 0:

All birthday audiences currently show 0 because they are time-dependent.

1.3 — What data feeds these audiences?

Prompt

What datasets do I have that contain customer data?

Expected result

Coworker identifies two key business datasets and related system datasets:

System Journey Optimizer datasets for engagement signals include email tracking, push tracking, consent, and message feedback.

Part 2: Discover the Right Fields

2.1 — Find account value fields

Prompt

What fields are available related to account value or balance?

Expected result

Coworker surfaces the primary field and related fields in the weBank: CRM dataset:

2.2 — Find credit card fields

Prompt

Find fields related to credit card or card type

Expected result

Coworker identifies these fields in the weBank: CRM schema:

2.3 — Geographic fields

Prompt

Is there a field for home state or region?

Expected result

Coworker identifies geographic fields in the weBank: CRM dataset:

Part 3: Find the Right Audience for a Reactivation Campaign

3.1 — Ask the AI to help you find the right audience

Prompt

Help me find or create the right audience for a credit card reactivation campaign — I want to target customers who were scored as high or medium propensity for a credit card upgrade but haven’t taken action yet

Expected result

Coworker finds two candidate audiences that have never been targeted in a journey or destination:

Recommended options:

Combined size: approximately 5,308 profiles

3.2 — Create the precision audience

Prompt

Create an audience of customers who are in the “Propensities: Credit Card Upgrade High” audience AND have an account value over $250,000 AND are opted in to email

Name it Platinum Upgrade - High Value Email Eligible → Select Batch → Approve Plan

Expected result

Coworker layers ML propensity, financial value, and channel eligibility into one precision segment — work that would normally require a data team ticket.

3.3 — Estimate size & waterfall

Prompt

How large is that audience? Show me the waterfall breakdown.

Expected result

Coworker returns a real-time profile count estimate and a visual waterfall chart:

This shows exactly which condition is the biggest filter and whether to loosen the criteria.

3.4 — Check for similar audiences

Prompt

Are there any existing audiences similar to the one I just created?

Expected result

Coworker confirms no single combined audience exists, but all building blocks are already present:

Coworker confirms there is no duplication risk and recommends proceeding.

Part 4: Build the Journey

4.1 — Get journey ideas grounded in your data

Prompt

Based on popular use cases in financial services, suggest some quick-win journeys I can create with the audiences I have

Expected result

Coworker suggests five concrete journey ideas grounded in actual audiences:

4.2 — Create the credit card upgrade journey

Prompt

Create a journey using the audience “Platinum Upgrade - High Value Email Eligible”. Send a push notification introducing the weBank Platinum Card benefits. Wait 3 days. If the customer hasn’t applied, send an SMS with a limited-time offer of 0% APR for 12 months on balance transfers. Wait 5 more days. If still no application, send a final push notification with a personal banker invitation.

Review plan → Approve Plan → Grant permissions

Expected result

Coworker creates a full 3-touch, 2-channel journey from a single natural language description, including wait timers, condition checks, and escalating offers.

4.3 — Bonus: Create a journey from an image

Prompt

Create this journey from the image I uploaded

Expected result

Coworker reads the uploaded image — including nodes, channels, wait times, and conditions — and generates the full journey in Journey Optimizer. This turns a planning artifact directly into a deployable journey.

4.4 — Check for conflicts

Prompt

Are there any active journeys that could conflict with the journey I just created?

Expected result

Coworker automatically checks all active journeys for audience overlap, schedule collisions, and channel saturation.

Part 5: Turn Journey Drop-Offs into a New Audience

5.1 — Identify the drop-off point

Prompt

Show me the journey I just created and identify which step has the biggest drop-off

Expected result

Coworker analyzes journey step events and surfaces the node with the highest exit rate. For example:

68% of profiles who received the first email never opened it and exited before the SMS step.

5.2 — Create a drop-off audience

Prompt

Turn those journey drop-offs into a new audience — people who entered the Platinum Upgrade journey but never reached the SMS step

Name it Platinum Upgrade - Email Non-Responders → Select Batch → Approve Plan

Expected result

Coworker creates a net-new audience based on journey behavioral data — profiles that entered the journey but did not progress to the SMS step. This audience is immediately usable for a follow-up campaign.

5.3 — Reactivate the drop-offs

Prompt

Create a reactivation journey for the “Platinum Upgrade - Email Non-Responders” audience. Try a different approach — start with a push notification highlighting a limited-time 75,000 bonus points offer, wait 2 days, then send a direct mail invitation to visit their local branch.

Review plan → Approve Plan → Grant permissions

Expected result

Coworker creates a dedicated recovery journey using different channels and messaging for non-responders. You have now closed the loop on a conversion leak — entirely through conversation.

Part 6: Automated QA & Validation

6.1 — Data freshness and ingestion health check

Prompt

Are there any data quality issues with the datasets feeding my audiences? Check ingestion health for weBank: CRM and weBank: Customer Actions

Expected result

Coworker reports ingestion health for each dataset:

Recommendation: Review TTL settings or check the source system for stale records.

6.2 — Validate the journey before publishing

Prompt

Review the Platinum Upgrade journey for errors or quality issues — check for missing wait timers, dead-end paths, missing channel configurations, and audience eligibility gaps

Expected result

Coworker inspects the journey structure and flags issues such as:

Part 7: Monitoring & Governance

7.1 — Track audience health over time

Prompt

Show me how the “Propensities: Credit Card Upgrade High” audience has changed over the last 30 days

Expected result

Coworker returns a time-series view of audience growth or decline with trend analysis, helping you identify whether the ML model’s scoring is drifting or upstream data changes are affecting qualification.

7.2 — Diagnose a change

Prompt

Why did the “Account Value: +500K” audience drop recently?

Expected result

Coworker performs causal analysis by decomposing the change across:

This replaces 1–2 days of analyst triage.