TL;DR
- Raw helpdesk data connected to Claude gives you ticket content, but not insight. You need an enrichment layer first.
- Revelir Insights connects to Claude via MCP, giving Claude access to both raw Zendesk/Salesforce data and a full AI enrichment layer in one connection.
- CX leaders can ask questions like "What drove negative sentiment last week?" and receive synthesised, evidence-backed answers.
- The sentiment arc (how a customer felt at the start versus end of a conversation) surfaces retention risks that standard resolved/unresolved ticket statuses cannot.
- No code is required. The MCP connection handles the integration.
Why Can't You Just Connect Claude Directly to Zendesk?
You can, and many teams do. But a standard Zendesk MCP connection gives Claude access to raw ticket data: subject lines, agent replies, timestamps, and status fields. What it does not give Claude is meaning.
Ask Claude "Why are customers frustrated this week?" via a raw Zendesk connection, and it will scan ticket text and attempt to infer patterns. The answers are inconsistent, unstructured, and unverifiable. There is no sentiment score to filter on, no standardised contact reason taxonomy, no churn risk flag.
This is the core problem with connecting AI to unstructured support data: the AI is doing interpretive work that should have already been done at the data layer. As CMSWire notes, the questions CX leaders ask often reveal a deeper problem: the data they have does not actually answer the questions they need answered.
The fix is not a better prompt. It is a better data layer.
What Is an MCP Connection and Why Does It Matter for CX?
MCP (Model Context Protocol) is an open standard that allows AI models like Claude to connect to external data sources and retrieve structured information in real time. Think of it as a live API bridge between your data and Claude's reasoning engine.
For CX leaders, the practical implication is straightforward:
- Without MCP: You export data, paste it into Claude, and ask questions. This is manual, stale, and limited by context window size.
- With MCP: Claude queries your live data source directly, retrieves relevant records, and synthesises answers on demand.
The quality of the answer Claude returns is entirely dependent on the quality of the data source it is connected to. A raw helpdesk connection returns raw helpdesk data. An enriched AI layer returns structured, labelled, queryable insight.
What Does "Enriched" Ticket Data Actually Mean?
Enriched ticket data means every conversation has been processed by an AI layer before Claude ever sees it. Standard enrichment fields include:
| Field | Raw Helpdesk Value | Enriched Value |
|---|---|---|
| Ticket status | Resolved / Open | Resolved, but sentiment ended negative |
| Contact reason | Freetext subject line | Standardised AI-generated tag (e.g. "Refund - Delay") |
| Customer sentiment | Not available | Scored at conversation start and end |
| Churn risk | Not available | Flagged based on sentiment arc + tone shift |
| Custom metrics | Not available | Binary, multi-option, or tag-based per your definition |
This is the difference between a filing cabinet and a structured database. Claude can reason over the latter far more effectively.
According to Clootrack's customer feedback analytics guide, the practical goal of feedback enrichment is to identify recurring concerns, rising expectations, and high-volume topics across channels including support tickets, app reviews, and social media. Enrichment at the ticket level is the foundational step that makes this cross-channel analysis possible.
How Does Revelir Insights Connect to Claude?
Revelir Insights is an AI insights engine that enriches every support ticket with structured metadata, then exposes that enriched data layer to Claude via MCP. Critically, it is a superset of a standard Zendesk MCP connection. One connection gives Claude both the raw helpdesk data and the full AI enrichment layer. No separate Zendesk connection is needed.
What CX leaders can ask Claude once connected:
- "What drove negative sentiment last week?"
- "Which contact reason is growing fastest this month?"
- "Show me tickets where customers started positive and ended negative."
- "What do refund-related tickets have in common this quarter?"
Each answer is backed by real ticket data, not a statistical summary. Every insight is traceable to a specific conversation.
The sentiment arc: the metric most CX platforms miss
Standard helpdesk reporting tells you a ticket was resolved. Revelir Insights tells you the customer started frustrated and ended neutral. At scale, that distinction matters enormously. A technically resolved ticket with a negative sentiment arc is a retention risk. A batch of tickets where 15% of customers started positive and ended negative is a signal that something in your service delivery is actively eroding trust.
Research from Teravictus confirms that integrating multiple behavioural signals improves churn prediction accuracy by 6-52% compared to analysing any single metric in isolation. Sentiment arc is exactly this kind of compound signal.
Is This Actually No-Code? What Does Setup Look Like?
For CX leaders without engineering resources, the setup path is:
- Connect your helpdesk (Zendesk, Salesforce, or any API-compatible platform) to Revelir Insights.
- Revelir enriches every incoming ticket automatically. Historical backfill is also available.
- The MCP connection to Claude is configured once by Revelir. No code is written on the client side.
- Ask questions in plain English through Claude.
The contact center AI insights this enables, including churn risk detection, contact reason trending, and sentiment arc analysis, are available immediately after setup. No dashboards to configure, no SQL to write, no analyst to brief.
Frequently Asked Questions
Does this replace my Zendesk MCP connection?
Yes. The Revelir MCP connection is a superset of a standard Zendesk connection. It includes all raw ticket data plus the enrichment layer, so a separate Zendesk connection is not needed.
What helpdesks does Revelir support?
Revelir integrates with any helpdesk via API, including Zendesk and Salesforce. If your platform has an API, integration is possible.
How is sentiment arc different from standard CSAT?
CSAT is a post-ticket survey that a fraction of customers complete. Sentiment arc is AI-generated from every conversation, at both the start and end, giving 100% coverage with no survey dependency.
Can Claude answer questions about individual tickets?
Yes. Claude can surface specific conversations as evidence when answering broader questions, making every insight traceable to real customer interactions.
Is this suitable for regulated industries like fintech?
Revelir Insights is already in production at Xendit, an Indonesian fintech. Every AI evaluation includes a full reasoning trace (model, prompt, documents retrieved) to support compliance and audit requirements. As CX Network highlights, AI governance requires exactly this kind of auditability.
What kinds of custom metrics can I define?
Revelir supports binary (yes/no), multi-option, and tag-based custom metrics. Examples include escalation flags, product feedback categories, or compliance-relevant conversation attributes.
How quickly can this be deployed?
For teams already on a supported helpdesk, deployment timelines are measured in days, not months.
About Revelir AI
Revelir AI is an AI customer service platform that operates across three layers: an autonomous Support Agent that resolves tickets end-to-end, RevelirQA, an AI scoring engine that evaluates 100% of conversations against your own policies, and Revelir Insights, an AI insights engine that surfaces what is driving contact volume and customer sentiment. The platform is in production with enterprise clients including Xendit and Tiket.com, and is built for global enterprises operating at scale. Founded in Singapore by a YC alumnus, Revelir is designed for CX and Support Operations leaders who need to move beyond CSAT sampling and manual review.
Ready to ask your support data anything? Learn more or speak with the team at revelir.ai.
References
- Teravictus. CX Leader's Guide to AI Churn Prevention. https://www.teravictus.com/blog/cx-leaders-guide-ai-churn-prevention
- CMSWire. 5 Questions CX Leaders Are Asking and What They Really Mean. https://www.cmswire.com/customer-experience/5-questions-cx-leaders-are-asking-and-what-they-really-mean/
- Clootrack. Customer Feedback Analytics Guide. https://www.clootrack.com/cx-guide/customer-feedback-analytics-guide
- CX Network. AI Governance: A CX Leader's Guide to Responsible AI Implementation. https://www.cxnetwork.com/artificial-intelligence/articles/ai-governance-leaders-guide-responsible-implementation