CX leaders can now query their entire Zendesk ticket history in plain English and receive synthesised, evidence-backed answers in seconds. By connecting Revelir Insights to Claude via the Model Context Protocol (MCP), teams move beyond static dashboards and get a live, conversational interface over their enriched support data. This is not a reporting shortcut; it is a structural upgrade in how customer feedback analysis works at the enterprise level.

Why Is Asking Questions the Wrong Way to Use Your Zendesk Data?

Most CX teams do not have a data problem. They have an access problem. Zendesk holds an enormous volume of signal: ticket tags, satisfaction scores, agent notes, reply times, and raw conversation text. The bottleneck is that extracting meaning from that data requires either a data analyst with SQL access or a manual ticket review process that covers, at best, five percent of conversations.

The result is that CX strategy gets made on thin evidence. A weekly report shows CSAT dipped. Nobody knows exactly why, or which product area is driving it, or whether the issue is worsening. By the time a root cause is identified, the window to act has closed.

"Dashboards tell you what happened. The question is whether your platform can tell you why, and what to do next."

This is the gap that conversational querying over enriched ticket data closes. Instead of navigating filters and charts, a Head of CX types a question and gets a direct answer, the same way they would ask a well-briefed analyst.

What Makes Revelir Insights Different From a Standard Zendesk MCP Connection?

A raw Zendesk MCP connection gives Claude access to ticket fields: subject, status, assignee, tags, and CSAT ratings. This is useful, but it is a narrow slice of the actual customer experience signal available in a conversation.

Revelir Insights acts as an enrichment layer that sits on top of that raw data. Before Claude ever sees a ticket, Revelir has already processed it and added:

• Customer Sentiment (Initial): How the customer felt at the start of the conversation, not inferred from the tag or the outcome, but from the actual opening message.
• Customer Sentiment (Ending): How they felt by the last exchange. A technically resolved ticket where sentiment ended negative is a retention risk, and Zendesk's default view will never surface it.
• Reason for Contact: AI-generated contact reason tags that go beyond whatever manual tagging convention an agent applied in the moment.
• Custom Metrics: Binary, multi-option, or tag-based metrics defined by the business. Churn risk signals, escalation flags, product feedback categories, and more.

When Revelir Insights connects to Claude via MCP, this enriched layer travels with it. Claude is not reading raw ticket text. It is reading structured, AI-analysed signal. This is a qualitatively different input and it produces qualitatively different answers.

How Do You Actually Set This Up in Claude?

The setup is a one-time configuration. Once complete, querying your Zendesk data becomes as natural as sending a message ^[1].

1. Activate Revelir Insights on your account. Revelir connects to your Zendesk instance via API and begins enriching conversations as they come in. Historic backfill is available on Professional and Enterprise plans.
2. Enable the Revelir MCP connection in Claude. In Claude's settings, add the Revelir MCP endpoint. Credentials are generated inside the Revelir platform. This single connection surfaces both your raw Zendesk data and the full AI enrichment layer ^[2].
3. Verify the connection. Ask Claude a test question, for example, "How many tickets came in last week and what were the top three contact reasons?" A valid response with specific numbers confirms the connection is live.
4. Start asking real questions. No SQL. No filter navigation. No dashboard exports.

What Questions Can You Actually Ask, and What Does a Good Answer Look Like?

The value of customer sentiment analysis software is only realised if the questions are specific enough to be actionable. Generic questions produce generic answers. Here are high-value question patterns with the reasoning behind each:

Trend Questions

• "Which contact reason grew the most in the last 14 days compared to the prior 14 days?"
• "Is the volume of refund-related tickets increasing week over week?"

These surface emerging issues before they show up in CSAT data. By the time CSAT drops, the problem has already been affecting customers for weeks.

Sentiment Arc Questions

• "What percentage of tickets this week started with positive sentiment and ended with negative sentiment?"
• "Which agents had the highest rate of sentiment deterioration in their conversations?"

This is where the sentiment arc becomes a direct customer feedback analysis platform capability. A resolved ticket with a negative ending sentiment is a churn risk that a CSAT survey may never capture because the customer simply does not respond to the survey.

Root Cause Questions

• "What topics are mentioned most often in conversations where sentiment ended negative?"
• "Which product area generated the most complaints last month?"

QA and Coaching Questions

• "Which agents have the most tickets flagged for escalation risk?"
• "Are there common phrases in low-scoring conversations?"

Customer service QA automation becomes genuinely useful when it can be queried contextually, not just reviewed as a score export.

What Are the Limitations to Know Before You Start?

Conversational querying is powerful, but it is not magic. A few honest constraints to keep in mind:

• Quality of enrichment determines quality of answers. If Revelir Insights has not had time to process a large ticket backfill, questions about historical periods may return incomplete data. Allow enrichment to complete before running retrospective analyses.
• Claude's context window applies. Very broad questions across very large datasets may require scoping. Asking about a specific week or category will return sharper answers than asking across all time ^[2].
• Custom metrics must be configured first. If your team wants to query for a specific business metric, for example escalation risk or first-contact resolution, that metric needs to be defined in Revelir Insights before it can be queried. Out of the box, sentiment, contact reason, and standard Zendesk fields are available immediately.

Frequently Asked Questions

Do I still need a separate Zendesk MCP connection if I use Revelir Insights?

No. The Revelir Insights MCP connection is a superset of a standard Zendesk MCP. It includes everything a raw Zendesk connection provides, plus the full AI enrichment layer. One connection is sufficient.

How quickly does Revelir Insights enrich new tickets?

New tickets are enriched in near real-time as they close. For live querying of recent conversations, enrichment latency is typically minimal. For historic backfill on large datasets, processing time depends on volume.

Is this only useful for large CX teams?

No. The value scales with ticket volume. Teams processing hundreds of tickets per week benefit from eliminating manual report-building. Teams processing thousands of tickets per week, like Xendit and Tiket.com, benefit additionally from population-level pattern detection that is impossible to do manually.

Can I use this as a customer feedback analysis platform for product decisions?

Yes. Revelir Insights tags contact reasons and custom product feedback categories across 100% of tickets. Asking Claude "what product issues are customers raising most often this month" gives a product team an unfiltered view of the customer voice without any survey bias.

Does the platform support languages other than English?

Yes. Revelir Insights has proven multilingual support and runs in production across Indonesian-language, high-volume environments. Enrichment and querying work across languages.

How is this different from Zendesk's own AI features?

Zendesk's native AI features operate within the Zendesk interface and are primarily focused on agent productivity. Revelir Insights provides a portable enrichment layer that can be queried through Claude, giving CX leaders a flexible, conversational interface independent of any single helpdesk's native reporting constraints.

What helpdesks does Revelir support beyond Zendesk?

Revelir connects to any helpdesk via API, including Salesforce Service Cloud. The MCP integration with Claude works across helpdesk sources, so teams on multiple platforms get a unified querying experience.

References

1. How to answer common customer inquiries with Claude · Missive Blog (missiveapp.com)
2. Claude best practices 2026: the complete power user guide (www.the-ai-corner.com)