Turn Complex Help Scout Tickets into Research-Backed Customer Responses

24 min setup
No coding required
Runs automatically
Route Help Scout tickets tagged as "research needed" to Perplexity's AI research engine for comprehensive analysis. Perplexity pulls real-time data on product specifications, competitive benchmarks, troubleshooting protocols, and industry standards, then generates detailed draft responses. The researched solution appears as a note in your Help Scout ticket, ready for agent review and customer delivery.

How it works

10 steps to start turning Help Scout tickets into research-backed responses automatically

Conversation Tags Updated

Why this matters: This trigger captures tag changes on Help Scout conversations, enabling agents to signal that specific tickets need AI research simply by applying a tag rather than manually searching for information.

Check for Duplicate Processing

Why this matters: Counts how many times the "updated by mesa" tag appears on the conversation to determine if this ticket has already been processed, preventing duplicate research and wasted API calls when tags are modified multiple times.

Filter: Skip Already-Processed Tickets

Why this matters: Acts as the decision gate that ensures research only runs on tickets that haven't been processed yet, maintaining workflow efficiency and preventing redundant AI research on the same ticket.

Get List of Conversation's Threads

Why this matters: Retrieves all messages in the conversation thread to provide complete customer context for AI research, ensuring understanding of the full issue, previous troubleshooting attempts, and conversation history.

Map (Extract Message Bodies)

Why this matters: Extracts just the message text from all conversation threads into a comma-separated format that can be passed to AI, creating a readable conversation history without metadata clutter.

Loop: Find 'Research Needed' Tag

Why this matters: Identifies and processes conversations specifically tagged with "research needed," enabling the workflow to selectively research only tickets agents have flagged rather than processing all tag updates.

Research Issue with Perplexity AI

Why this matters: Uses Perplexity's web-connected AI to research the customer's issue with current, reliable sources, generating a structured draft response that agents can review and refine before sending to customers.

Add Note to Conversation

Why this matters: Saves the AI-generated research draft directly to the Help Scout conversation as an internal note, making it immediately accessible to the assigned agent without requiring context switching to external tools.

Map (Extract Current Tags)

Why this matters: Creates a comma-separated list of all existing tags on the conversation so the "updated by mesa" marker can be added without removing other tags that agents have applied.

Mark Ticket as Processed

Why this matters: Adds the "updated by mesa" tag to mark this conversation as processed, preventing duplicate research if tags are modified again and providing agents with a clear indicator that AI research has been completed.

Frequently asked questions

How do I customize the research output format?

Edit the "Create Chat Completion" step and modify the "Output Format" section of the prompt. You can change the structure, add/remove sections, adjust detail level, or specify different types of information to include based on your support team's needs.

Can I use different AI models or adjust research depth?

Yes, the model parameter can be changed from "sonar-pro" to other Perplexity models. Temperature (currently 0.2) controls consistency vs creativity—lower values (0.1) produce more consistent factual responses, higher values (0.5+) produce more varied responses. Adjust based on your accuracy vs flexibility needs.

What if Perplexity can't find relevant information?

The AI will indicate when information is uncertain or unavailable in its response, typically in the "Notes & Caveats" section. The agent can then do additional manual research or escalate to specialized team members. Consider adding a fallback that tags tickets as "needs-specialist" if confidence is low.