A customer asks your AI agent how to change their subscription. The answer explains the steps, but the customer still gets stuck.
Your support team handles similar questions throughout the week. Some need a clearer explanation. Others reveal a problem in the product itself.
You update the guidance. Now you want to know whether it helped—and whether the same friction is showing up elsewhere.
Every conversation offers a chance to improve something: an answer, a support process, or the experience that prompted the customer to reach out.
Today, we’re introducing Rippit’s Decagon integration to help you connect those signals.
Sync Decagon conversations into Rippit, investigate them alongside customer conversations from sources like Zendesk and Intercom, and turn your findings into structured analysis. Then build an Agent App that keeps checking as new conversations arrive.
See the Decagon integration documentation for setup, data handling, and sync details.
Chat and voice conversations with resolution, CSAT and quality reviews
Why add Rippit if you already use Decagon?
Decagon’s AI conversations and Watchtower reviews provide valuable context about the experience your AI agent delivers. Rippit brings that context into a broader workflow across your customer conversation data. For a side-by-side look, see Rippit vs. Decagon.
| What Rippit adds | Why it matters |
|---|---|
| Broader conversation coverage | Investigate patterns across Decagon conversations and human support interactions from other connected sources. |
| Structured, reusable analysis | Classify conversations by issue, root cause, or outcome, then save the results in columns your team can filter and chart. |
| Managed recurring work | Turn an investigation into an Agent App with a defined scope, schedule, and delivery destination. |
For example, start with subscription questions handled by your AI agent. Compare the problems customers describe with those reaching your human support team. Then track whether changes to your guidance or product reduce the friction.
The question becomes broader than how an individual interaction went: What keeps making this difficult for customers, and what should we change?
Bring Decagon conversations into your analysis
Rippit’s Decagon integration imports full conversation content on a schedule. Each Decagon conversation becomes a conversation in Rippit, including:
- The messages: Customer and AI messages in order, with their respective authors.
- Conversation context: Summary, notes, resolution, flow type, deflection status, CSAT score, tags, and metadata.
- Voice recordings: Audio from voice conversations is downloaded and stored. Rippit transcribes the stored recordings.
- Watchtower reviews: Review results, flags, rationales, rubric scores, and category-level choices and scores are stored with the conversation.
That means your investigation can draw on both what was said and the review context already available in Decagon.
You can examine a pattern, inspect the underlying conversations, and decide which additional classifications would help your team understand it.
Start with a question that needs the broader picture
Imagine customers have been struggling to change their subscriptions. Your AI agent handles many of those questions, while others reach your human support team.
You want to know:
“Analyze subscription-change conversations from Decagon and our connected support sources over the past 30 days. What are customers trying to do, where do they get stuck, and which problems appear across both AI and human support?”
Here is how the workflow comes together.
1. Read the conversation behind the outcome
A conversation’s status is useful context. The exchange itself helps explain what happened.
Did the customer understand the instructions? Did they say the steps failed? Did the answer address the original question? Was the obstacle a confusing policy, a missing capability, or a suspected bug?
Rippit lets you investigate those questions alongside the available resolution fields, customer feedback, and Watchtower reviews.
Then broaden the investigation to your connected human support conversations. Look for recurring issues, differences in the explanations customers receive, and approaches that appear to help.
Comparing themes across sources can reveal a shared problem even when the conversations involve different customers.
2. Turn the findings into reusable analysis
Ask Rippit to classify the selected conversations using the distinctions your team needs. For this investigation, those might include:
- What the customer wanted to change.
- Where they encountered difficulty.
- The likely cause, with supporting evidence.
- Whether the customer explicitly confirmed success.
- Whether the outcome remained unclear.
- The improvement the conversation suggests.
Save those classifications as structured columns. Then compare the most common obstacles, break the results down by source, and examine how patterns change over time.
Keep “confirmed success” separate from “no further response.” That distinction gives your team a clearer view of what the conversation actually establishes.
The resulting analysis stays available for follow-up questions, worksheets, and charts. As you learn more, you can refine the categories without starting the investigation over.
3. Make improvement tracking a standing job
Once the analysis is useful, turn it into an Agent App.
For example:
“Every Monday morning, review the past week’s subscription-change conversations from Decagon and our connected support sources. Summarize the leading points of friction, highlight examples where customers remained stuck, and compare the themes with the previous week. Send a brief with conversation links to our customer experience Slack channel. If no significant changes appear, say so.”
Rippit builds a flow you can inspect and adjust: the records it reads, the analysis it performs, the format of the brief, its destination, and its schedule.
Test the findings, refine the instructions, and activate the app. Your team can then review fresh evidence as part of its regular improvement process.
Agents that scan, analyze, and act
The same approach supports several recurring jobs:
| Agent | Scan | Analyze | Act |
|---|---|---|---|
| AI experience review | Decagon conversations and available Watchtower reviews | Identify recurring confusion, unresolved requests, and patterns worth investigating | Send a brief with supporting examples |
| Contact-driver analysis | Decagon and connected human support conversations | Find the underlying reasons customers ask for help across sources | Deliver a ranked summary to support and product |
| Improvement tracking | Conversations before and after a documented change | Compare recurring complaints and evidence of customer difficulty | Report what improved and what still needs attention |
| Product feedback | Customer requests across AI and human conversations | Group missing capabilities, suspected bugs, and usability problems | Send a recurring product feedback digest |
Get started
Connect Decagon, let your conversations sync, and start with one recurring customer problem:
“What do customers keep getting stuck on in our Decagon conversations, and does the same issue appear in our human support conversations?”
Review the evidence. Save the classifications that make the problem clearer. Then turn the investigation into an Agent App that keeps your team informed.
Your AI agent keeps helping customers. Every conversation can help your team make the next experience better.
.webp)
