Turn Complex Help Scout Tickets into Research-Backed Customer Responses

When any conversation's tags are updated in Help Scout—whether you add, remove, or change tags—this trigger captures the tag update and conversation details. It monitors your Help Scout account continuously and passes the full conversation data to the next step for tag verification. The trigger automatically connects to your Help Scout account once you authorize MESA, with no manual configuration needed.

The step loops through all tags on the conversation and counts specifically how many match "updated by mesa"—a tag that gets added at the end of the workflow. If the count is 0, it means this ticket hasn't been processed yet; if the count is 1 or higher, it's already been researched. This count gets passed to the next filter step to decide whether to continue or stop the workflow.

The filter checks if the count from the previous step equals exactly 0. If yes (the ticket hasn't been processed), the workflow continues to retrieve conversation threads; if no (the ticket was already researched), the workflow stops here. This is a hard-coded filter with no configuration needed—it automatically protects against reprocessing tickets you've already researched.

The step fetches every thread (message) in the conversation using the conversation ID from the trigger. This includes the initial customer inquiry, any follow-up questions, and previous agent responses that might provide additional context about what the customer needs. Each thread's body text gets extracted and prepared for the next step to combine into a format the AI can process. The step automatically uses the conversation ID to pull the correct thread data from Help Scout.

The Map step iterates through all the thread bodies retrieved in the previous step and combines them into a comma-separated text string. For example, if there were 3 messages in the conversation, this step extracts each message body and joins them together, creating a single text field that contains the entire customer inquiry from start to finish. This formatted text gets inserted into the Perplexity AI prompt for research.

The step loops through all tags on the conversation and filters specifically for the tag "research needed" (this exact tag name is required). When it finds a match, it allows the workflow to continue to the Perplexity research step; if no match is found, the workflow stops here. The loop structure allows you to process multiple matches if somehow the tag appears more than once, though typically it will match once or not at all.

This step sends the conversation context to Perplexity AI with detailed research instructions.

Critical configuration: You must customize the AI prompt by replacing placeholder text:

1. Replace "[INSERT COMPANY NAME]" with your company name (appears 3 times)
2. Replace "[INSERT COMPANY WEBSITE]" with your website URL
3. Replace "[INSERT WHAT COMPANY DOES]" with your product/service description

The AI receives {{loop_3.comma_separated}} (all conversation messages) and uses model "sonar-pro" with search_mode "web" to research the issue. It's instructed to:

1. Research relevant information from reliable sources (vendor docs, official standards, technical sources)
2. Include product specifications, competitive comparisons, troubleshooting protocols, industry standards
3. Output a structured draft with: Summary, What We Found (facts + links), Steps to Resolve, Notes & Caveats, Sources
4. Use neutral, precise language appropriate for support replies
5. Cite all sources for non-obvious claims

Search capabilities: Perplexity searches the web in real-time (unlike ChatGPT's training cutoff), enabling research on current versions, recent outages, latest documentation, and time-sensitive technical issues. The temperature "0.2" setting ensures factual, consistent responses rather than creative variation.

This step creates a Help Scout note on conversation {{helpscout.id}} with text {{perplexity.response}} (the AI's structured research draft). Internal notes are visible only to agents, not customers. The agent can review the research, refine the response to match their voice, add personalization, and reply to the customer with confidence backed by thorough research.

Agent workflow: After this note is added, agents see it in Help Scout's conversation view, can edit it, and use it as the foundation for their customer reply rather than starting from scratch or doing manual research.

The step loops through all the conversation's current tags and extracts them into a comma-separated list. For example, if the ticket has tags "research needed, billing, priority," this step creates a list containing all three tags so they can be maintained when adding the "updated by mesa" tag. This ensures you don't lose your existing tag structure when the workflow marks the ticket as processed.

This step updates the conversation's tags to include all existing tags ({{loop_4.comma_separated}}) plus ", updated by mesa". This additive approach preserves any tags agents applied while adding the processing marker. The "updated by mesa" tag is checked in step 2 to prevent re-processing.

Tag visibility: This tag appears in Help Scout alongside other tags, providing transparency to agents about which tickets have received AI research.