The great AI transformation is here. Everyone is trying to figure out how they can be the first to successfully redefine their business with AI. And for good reason: The last transformation of this scale, the digital transformation, created new billion-dollar companies and totally retired existing ones.
But there is a difference between being first and being successful. Being early to AI transformation is no guarantee of success, and companies feel that right now. We’ve all read the MIT study headline—95% of generative AI pilots are failing to generate ROI.
At Pendo, we’ve spent 12 years improving the world's experience with software. And while everything is new, some things look pretty familiar.
AI systems are built without a data-driven approach. Adoption suffers because change management isn't addressed. The voice of actual users is ignored. No one is doing the hard work to improve new AI solutions that are subpar at launch—the same issues that software leaders have faced for decades.
There are many reasons AI isn’t meeting expectations, but don’t blame AI. The real issue is that companies aren’t building a system to measure and scale user success. The right system is intelligent, designed to make sure the AI you build and buy is successful, and has true business impact.
That system at the heart of this quarter's product release. Let’s dig in.
A system to launch agents users love
Right now, everyone is launching agents, but few are succeeding. There are two major issues.
1. Agent strategy has no inputs
The goal shouldn’t be just to launch a new agent; it should be about changing how people behave. Helping users achieve better outcomes faster. Improving how employees accomplish tasks and helping your customers be more productive. You can’t do that unless you know where users get stuck, what their existing workflows look like, and what their most important use cases are. If you don’t have these answers, you lack the right inputs for your agent strategy.
We see our customers using Pendo Analytics in a new way. Mapping customer behavior before making major decisions about where to inject agentic experiences.
2. Companies can’t improve the agents they deploy
No software is perfect on day one, and neither is your agent. You have to understand how people are leveraging the agent, see where they get stuck, and monitor the relationship between agent and traditional software usage across complex workflows. At Pendo, we have 12 years of helping a billion end users succeed with software and we’ve learned one thing very clearly. Without visibility into usage and data, you’re just guessing.
With Agent Analytics teams can finally see key AI agent performance metrics and adoption trends. Not just who clicked a button, but what happened next. Are users getting value? Are they abandoning mid-task? Are specific prompts failing to land?
And when things aren’t working, you’re no longer left guessing. Agent Mode lets you ask AI questions like, “Where is adoption dropping off?” and get immediate answers in plain language. You can even generate in-app onboarding flows from a simple prompt, guiding users past confusion and into action with the help of AI guides creation.
A system to improve productivity with AI
AI has broken the old cadences of work. User requests come faster, and change happens faster. If your team is still operating on SaaS speed (manual analysis, building dashboards, buried metrics), they can’t keep up.
Every software experience is moving this way. But to make our customers successful, Pendo must be far ahead of this curve. We don’t want to layer in AI and agents to help you here and there, we want to redefine what it’s like to use Pendo - agent first. We want to do it in a way that delivers in-depth insights to power users. We also want to help more casual users, who just want to quickly get questions answered without feeling intimidated or lost.
Agent Mode does exactly this. At its heart, it’s a new, more powerful agent that can access and take actions across all of Pendo’s products. It’s fun to talk to, extremely smart (and getting smarter), and will proactively surface opportunities and risks.
Instead of spending hours unraveling usage patterns, product teams get a built in data analysis. You provide instructions through a conversational interface or pre-built template, and the agent uses logic to form plans and access the right Pendo data and tools through Model Context Protocol (MCP) to produce results. It then follows up with strategic questions to improve your results.
Talk about a system for productivity!
A system to use data for predictive outcomes
Great agents are built on great data sets. But powerful data sets can have an impact in other ways too. They can even predict the future. I mean that literally.
A handful of Pendo customers have leveraged usage data to build predictive models. Powerful signals that predict which accounts might churn, or which are ready to buy more. They did this through an in-house data science team. Those who can do it have amazing results. But the majority of our customers are limited in what they can pull off, even if they have the right team. Building models, tweaking them, or delivering more specific in-depth models takes a huge investment of time and resources.
We want all of our customers to experience this value. In July of 2025, Pendo bought Forwrd. That acquisition has quickly turned into a new on demand predictive model building feature called Pendo Predict.
Predict allows you to quickly build predictive models with your Pendo Data. You choose which data sources to integrate alongside your Pendo data, like CRM or other integrations, then build models inside Predict. This data is then sent back into your CRM, where your customer-facing teams can put it to use.
Because predictive insights now live alongside behavioral data and user journeys, you can immediately trigger the right motion. Whether routing a support follow-up, launching a targeted rescue guide, or syncing signals into Snowflake for deeper modeling.
You can even see how AI agents themselves are contributing to churn, whether through ineffective prompts, confusing flows, or a complete lack of engagement. This isn’t just churn detection. It’s churn intervention and automation, embedded directly into the product experience.
Build the system, succeed with AI
It’s going to be a telling next few years. The companies that can redefine how they build, how they work, and adapt faster are going to win. It’s going to take a lot of effort, but those who build the best system for success will increase their odds. We’re excited to help you build that system and support you as you roll out agents, redefine your product lifecycle, and use more AI in your processes.
Explore everything new in the Autumn 2025 Release.
