Resolver AI is a retrieval-augmented generation (RAG) Zendesk app that helps agents draft accurate, policy-aligned customer responses in seconds—directly inside the ticket.

Instead of “guessing,” Resolver AI retrieves the most relevant Zendesk Help Center articles and synced internal documents and uses them (along with the latest customer query and optional agent instructions) to generate a grounded, high-quality draft response.

Ideal for reducing handle time, improving consistency, and ensuring every response stays aligned with approved support content.

⸻

How It Works (RAG)

1. Retrieve
Resolver AI searches your synced knowledge sources (Help Center + documents) to find the most relevant content for the ticket.

2. Generate
Using those retrieved snippets, Resolver AI generates a clear, context-aware draft reply based on:- The latest customer query / message- Relevant Help Center articles- Synced internal documents (FAQs, policies, SOPs)- Optional agent instructions (tone, constraints, next steps)

3. Review & Send
The draft appears in the Zendesk editor so agents can review, edit, and send—keeping full control of the final response.

⸻

Key Features

1. Knowledge-Grounded Responses (RAG) – Drafts are grounded in your existing documentation so agents answer accurately and consistently.
2. Optional Agent Instructions – Short prompts like "be concise" or "offer troubleshooting steps" guide tone and content.
3. Ticket-Aware Drafts – Uses the latest customer query so replies stay relevant.
4. Clean Formatting – Produces readable, well-structured drafts that need minimal edits.
5. Feedback Loop for Continuous Improvement – Admins can review agent responses to identify knowledge gaps and capture override responses where needed. As documentation improves, Resolver AI’s accuracy and relevance improve significantly over time.
6. Adoption Tracking – Applies a resolver_ai tag when used so you can track usage in Zendesk reporting.

⸻

Admin Feedback Loop (Improve Answers Over Time)

Resolver AI includes an admin-friendly feedback loop that helps you:

• Review the AI-generated answers produced for support tickets and questions

• Approve responses that are accurate, or override them with the correct “final answer” when needed

• Capture approved/overridden responses so Resolver AI can use the best-approved wording in future similar situations

Over time, this process helps Resolver AI deliver more accurate and consistent answers—based on what your admins have confirmed as correct—without requiring agents to do anything extra.

⸻

Data Handling & Privacy

• Only the necessary ticket content (latest public customer message + optional agent instructions) and relevant knowledge snippets are used to generate a draft.
• Resolver AI does not permanently store ticket content or PII.
• Data is never used to train shared models.
• Processing occurs in a secure backend and returns results directly to Zendesk.

⸻

Support

For help with setup, document syncing, or troubleshooting, visit MyCXApps Support.