Introducing Generative Analytics: Turn conversation insights into a growth lever

AI agents create unprecedented access to conversation insights. At Decagon, we’ve committed to giving teams the most comprehensive analytics suite, so they can leverage those signals. We started by introducing Watchtower, an always-on QA tool that scores conversations against a team’s custom rubrics. We then introduced Duet, so teams could ask any analytics question and get instant answers.

Today, we’re introducing the next evolution of our analytics platform: Generative Analytics. Generative Analytics gives teams greater access to the custom insights they care about most. With Generative Analytics, teams can: 

• Create charts just by asking Duet in natural language
• Fill in data gaps by tracking conversation outcomes
• Empower smarter business decisions with deep dives into revenue and agent health 

Create charts just by asking

With Duet, hundreds of teams already create custom reports grounded in their conversation data. The reports anchor weekly meetings and tell teams what is most important to their customers, how their agent is behaving, and what innovation they should invest in next. 

Now, teams can turn any Duet insight into a custom chart in natural language. No matter how deep the cut, teams can prompt Duet to create custom data representations, including time series, multi-series bar charts, derived trends, and distributions. Every chart downloads as a titled, padded PNG that works in light or dark mode, ready to drop into a deck, an email, or Slack.

As the charts come through, teams can dig deeper to get the answers that matter most. Duet acts as an always-available data scientist to answer what changed last week, why a metric moved, and where to look next.

Fill in your data gaps

Duet has access to all conversation data, but it needs help to identify the moments and insights that matter most to each team. Action Trackers act as a new mechanism for teams to curate custom datasets that Duet can pull from.

Action Trackers work as a trigger-based data capture, so teams can track outcomes at any point in the customer conversation. Teams can also define what type of data to capture, whether they need to know if an action was or was not completed, or if they want to sum up payments made. Once an Action Tracker is built, it lives in a shared library that any Agent Operating Procedure (AOP) can pull from.

For one major BNPL provider, an Action Tracker follows the entire payment flow to give the business a precise view of exactly where customers drop off before a payment completes. Meanwhile, a major airline uses Action Trackers to monitor successfully completed transactions like flight cancellations and refunds. 

Configurable charts let teams visualize any cut of that data with pre-built components like a bar or pie chart, and metric highlights. Dynamic filtering keeps every view current, so teams can drill into the metrics that matter most when they matter most.

Empower smarter business decisions

With the right data available, teams can monitor outcomes to inform how best to improve the customer experience. 

At Power, a leading home electronics retailer, Action Trackers are used to inform high priority agent updates and to improve revenue outcomes. One Action Tracker follows the success of Power's product-recommendation flow. It showed that the agent couldn't find a suitable product for 26% of customers and prompted the team to reprioritize engineering efforts to fix the gap. Another Action Tracker monitors offer creation and finalization, providing the team with insight into where the agent loses a sale. This analysis guided the team towards updates to the agent logic.

"With Generative Analytics, we stopped waiting on someone else's dashboard to answer our questions. Every team can just ask what they want to know about their own conversations, and get a real answer that directly pushes their work forward." –Adonis Hallberg, AI Team Lead at Power

‍

Book a demo to learn more about how you can put your conversation data to work.