While 92% of organizations have increased AI investments in 2025, over three-fourths of enterprises report that their agents haven’t moved beyond the pilot phase.

Executives are now asking themselves (and their teams) two questions: 

  1. How do we scale these agents we’ve launched? 
  2. How do we prove they’re delivering real value to users and the business?

For internal and external agents, teams are struggling to understand what users are doing with these tools and whether agents are being used the way they were intended. 

That’s why we built Pendo Agent Analytics (now free for all Pendo customers) to answer those questions and give you the tools you need to scale agents with confidence.

Through many conversations with our customers, and our own work launching agents at Pendo, we’ve identified three key phases of insight required to move from experiments to enterprise-ready deployments: understanding how your agents are used, pinpointing friction and failure points, and proving value across your user journeys.

Agent Analytics is built to guide teams through each phase.

Phase 1: Understanding how your agents are used

Who’s using them, and how often?

The first step to a high-impact agent is visibility. Agent Analytics shows you: 

  • Exactly who’s using your agents and how often
  • How many unique visitors and accounts are interacting with the agent
  • How many conversations and prompts they’re generating
  • How often they return to your agent and your overall product

All of this data can be filtered by role type, user segment, or timeframe so you can get a clear view across different personas or teams.

What are they using the agent for?

Adoption data is just the beginning of understanding usage. Teams also want to know what challenges their agents are actually being used to solve. 

Agent Analytics automatically surfaces emergent use cases by analyzing conversation data and categorizing it into themes. Out-of-the-box, you can instantly see which types of use cases are trending, which drive the most engagement or retention, and where users are finding the most value.

Is the agent being used how you intended?

If your agent was built or trained for specific tasks, like expense management or onboarding support, Agent Analytics helps you confirm whether users are actually using it for those intended purposes. 

You can track key use cases over time, see how frequently they’re accessed, and monitor trends in performance or frustration. And when you want more context, you can click directly into Session Replay to see the full user journey—including how users interacted with the agent, where they struggled, and what happened next.

“It’s incredibly useful to see the common categories and, crucially, to drill down into examples and even Replays of specific sessions of interest. This tool will make it significantly easier to iterate and improve our AI features.”



—Enterprise Power-user

Read how Pendo’s own product team used Agent Analytics and Session Replay to uncover unmet user needs, and improve our internal agent roadmap.

Phase 2: Identifying performance issues

Even the best agents will stumble sometimes. Or, users will open your agent and struggle to get the most out of it. The challenge is spotting those moments early and acting quickly, so users don’t give up on it forever.

That’s why we’ve built issue detection and frustration signals into the core of Agent Analytics. First, identify moments where users express anger with your agent through what Pendo has coined Rage Prompts: repeated rephrasing, all-caps messages, profanity, or other signs of dissatisfaction. These frustration signals are captured at the agent and use case levesl, helping you easily identify problems long before they surface in a support ticket.

You’ll also see where your agent is failing to answer effectively, and whether those failures are concentrated around specific topics, segments, or workflows. When patterns emerge, you can flag the issue and link it directly to a Jira ticket, turning an ambiguous frustration into a fixable engineering task.

Phase 3: Proving your AI investment pays off

Visibility and issue detection are vital. But leadership—especially for those spending R&D dollars—will ultimately ask a harder question: Is this agent driving measurable business results?

With Pendo, you can finally connect agent interactions to business outcomes. Because Agent Analytics is integrated with Pendo’s traditional product analytics, you get a hybrid view of user behavior before, during, and after engaging with your agent.

See what users do before and after agent interactions

Using Paths, you can understand what users were doing just before they turned to your agent, and what they did afterward. Did the agent help them complete a task, or avoid asking for help? Did users move forward in a workflow, or abandon your app altogether? Without a central tool to manage AI and traditional product insights, answering these questions is incredibly time-consuming and manual.   

Compare the “old” way and “agentic” way of using your app

If your agent was designed to streamline or replace a specific workflow, Agent Analytics gives you a direct way to validate that shift. You can compare how users completed a task—like booking a trip or filling out a form—through traditional, click-based paths versus how they now complete it with the help of your agent.

In a single dashboard (no SQL or stitching together tools required), you’ll be able to see whether agent-assisted outcomes are more efficient, whether conversion rates improve, and how behaviors differ across experiences. It’s the clearest view yet into whether your AI-powered workflows are truly better than what came before.

See exactly what a hybrid agent dashboard looks like in this Agent Analytics product tour.

Real-world use case: Monitoring Agentforce outcomes for Sales teams

For companies that’ve deployed internal agents to help teams like Sales move faster, visibility into usage and impact is key. With Agent Analytics and Pendo installed on your Salesforce instance, you can get exactly that.

Imagine your reps are using an Agentforce agent to complete tasks like updating deal records, finding collateral, or preparing for calls. With traditional Pendo Analytics capturing product usage data inside Salesforce and Agent Analytics capturing agent interaction data, you can see exactly which tasks the agent is assisting with, whether it’s accelerating completion time, and how adoption varies across regions or roles.

It’s a full picture of how your internal AI tools are being used, and whether they’re truly improving productivity and cutting costs.

Move your business beyond the AI pilot phase

Agent Analytics is now free and you can get started today. If you’ve launched an agent, but still feel like you’re guessing what’s working or digging through hours of log traces in Dev tools, it’s time to get a PM-first solution. Don’t be one of the 77% still struggling to scale.

As our own AI PM and Agent Analytics power-user, Niamh Jones, has said, “Agent Analytics is much easier for someone non-technical to really understand what’s going on within their agent. You know quickly, whether an agent is doing a good or bad job.”

👉 Are you a customer? Get started with Agent Analytics for free. Or, get a demo today.