Turn negative social mentions into Zendesk tickets

By General Input

Every 30 minutes, scan Brandwatch for negative mentions of your brand and open a Zendesk ticket for the ones that look like real customer complaints.

Integrations

  • Brandwatch
  • Zendesk

Type

Agentic Task

Categories

  • Customer Support
  • Marketing

Build me an agent that turns negative Brandwatch mentions of our brand into Zendesk support tickets, so the support team can respond to public complaints alongside their normal queue.

Trigger: cron, every 30 minutes.

On each run, the agent should call Brandwatch Retrieve Mentions for our configured project and brand query, filtered to negative sentiment, scoped to the last 30 minute window (with a small overlap to avoid missing mentions on boundary). Use cursor pagination if there are more than a page of results.

For each mention, the agent should judge whether it is an actionable customer complaint, meaning a real product or service issue from someone who appears to be a customer or user, that warrants a support response. It should skip general criticism, jokes, news or media commentary, competitor or analyst takes, and spam. When in doubt, lean toward skipping rather than filing a ticket.

Before creating a ticket for an actionable mention, the agent should call Zendesk Search Tickets to look for an existing ticket that already references the mention's original post URL. If one exists, skip it. This prevents duplicates across runs.

For each new actionable mention, the agent should call Zendesk Create Ticket with: a subject derived from the mention (a short summary of the complaint, prefixed with the source platform, e.g. "[Twitter] Checkout failing on mobile"); a body containing the full mention text, the author handle and display name, the source platform, the author's reach or follower count, the post's published timestamp, and a link back to the original post; priority set to high when the author's reach or follower count is greater than 10,000, otherwise normal; and tags including social-listening, the source platform slug (for example twitter, reddit, instagram), and negative-sentiment.

At the end of each run, log a brief summary: how many mentions were pulled, how many were judged actionable, how many tickets were created, and how many were skipped as duplicates.

Make the Brandwatch project, query (or query group), the follower threshold for high priority, and the list of tags configurable so the user can adjust them without editing the workflow.

Related prompts

Explore more prompts
Pause marketing emails to escalated customers, then restore themLook up any customer's email delivery history in one placeResolve unsubscribe tickets without a MailForge loginCatch angry support tickets before the customer escalatesAccount takeover ticket triage console for support teamsTurn your top support questions into explainer videosA support answer desk that drafts replies from your docsAnswer hard Zendesk tickets with your best past repliesDebug a customer's E2B sandbox from their support ticketCustomer log timeline for support agents, without Kibana