Adobe CX Coworker connector

Overview

The Adobe CX Coworker connector for Workfront Fusion lets you call Adobe CX Enterprise Coworker from an automation scenario. Instead of building API calls and parsing JSON yourself, send a plain‑language prompt and let Coworker interpret the request, run the applicable skills, and return a conversational result you can use later in the flow.

The connector provides two modules:

• Run a Prompt — Starts a new Coworker request. By default this module waits for Coworker to finish and returns the result, including both the natural‑language summary and any structured outputs the skill produced.
• Get Execution Status — Polls an earlier request to check progress or retrieve a result if you started the prompt asynchronously. Use this when you want non‑blocking execution or to coordinate long‑running checks across systems.

Impact

The Coworker connector dramatically reduces the complexity of integrating AI‑driven insights and data validation into automation scenarios. Use cases include:

• Automatically validating data (for example, dataset quality checks) and routing results to the right teams.
• Comparing analytics systems (AA vs CJA) and creating tickets or notifications when variance thresholds are exceeded.
• Generating plain‑language summaries of diagnostics, then creating follow‑up tasks or notifications based on recommendations.
• Embedding Coworker checks inside onboarding, QA, or deployment workflows so non‑engineers can trigger sophisticated validations without writing queries.

Because Coworker can use the account’s available tools and skills, the connector lets scenarios call domain‑aware capabilities (data validation, upgrade validation, insights) while keeping the scenario logic simple.

Tryable (quick start)

1. Create a new automation scenario in Workfront Fusion / Adobe Automation and add the Adobe CX Coworker app.
2. Use the Run a Prompt module. In the prompt text box, write a clear instruction such as:

   Validate the dataset Electronics Sample 1000 and return a summary of any fields with high null rates or obvious invalid values.

3. Run the module and inspect the output. You’ll typically get a conversational summary plus structured fields (quality scores, top invalid values, suggested next steps).
4. If you prefer asynchronous execution, set the Run a Prompt module to return immediately (where supported) and capture the executionId from the module output.
5. Add the Get Execution Status module downstream and pass the executionId to poll for completion. When status is finished, read the result and branch your scenario (create a ticket, send an email, or update records) based on the returned insights.

Setup

• Make sure your Adobe org has access to Adobe CX Enterprise Coworker and that the account credentials used in Automation have the necessary permissions.
• In Workfront Fusion add and authorize the Adobe CX Coworker connector following the standard app provisioning flow.
• Decide whether your scenarios should be synchronous (wait for results) or asynchronous (poll for status) and design the flow accordingly.
• Familiarize yourself with the Coworker skills available to your account (Data Validation, AA→CJA validation, insights, etc.) so prompts are aligned with capabilities.
• See the official connector docs for up‑to‑date module reference: Adobe CX Coworker modules.

Best practices & lessons learned

• Write prompts that are explicit and start with the object name when applicable (for example, “Validate the dataset Electronics Sample 1000” rather than just “Validate Electronics Sample 1000”). That helps Coworker reliably identify resources.
• For dataset or data checks remember Coworker typically analyzes a recent sample (for example, the last ~1,000 rows). Use follow‑up prompts or additional queries for exhaustive checks where needed.
• Use structured outputs (executionId, result JSON) from the connector to drive deterministic branching in scenarios rather than parsing free text alone.
• If you need long‑running validations, prefer the asynchronous pattern: submit with Run a Prompt, store executionId, then poll with Get Execution Status. This avoids timeouts and keeps scenarios resilient.
• Validate and human‑review suggested actions before making destructive changes. Coworker’s recommendations are a guide and sometimes require business context or further verification.
• Log prompts and results in your automation audit trail so you can iterate on prompts and improve consistency over time.

Common example flows

• Data quality alert: New dataset ingestion triggers a flow → Run a Prompt to validate key fields → If Coworker flags critical issues, create a high‑priority ticket and notify the data owners.
• Analytics upgrade check: Scheduled job triggers an AA→CJA comparison prompt → Run a Prompt to validate report suite vs data view → If matching rate is below threshold, generate a diagnostics report and assign remediation tasks.
• Automated summary & action: After a nightly job, Run a Prompt to summarize anomalies across multiple systems → Use Get Execution Status to wait for completion → Post summary to a Slack channel and create follow‑up tasks for items marked critical.

Troubleshooting tips

• If prompts return unexpected results, rephrase the request to be more explicit and include exact resource names or IDs.
• Confirm the Coworker account has access to the datasets, report suites, or data views you reference; missing access is a common cause of failures.
• Monitor connector rate limits and design backoff/retry logic for high‑volume scenarios.

Further reading

Full module reference and implementation details are available in the Workfront Fusion (Adobe Automation) docs: Adobe CX Coworker modules.

Got a connector pattern you want to share? Post your scenario in the Experience League Community forums to help others automate with Coworker.