20 ways to build better AI/ML products like Pendo’s PMs

For too long, product management has been a series of redundant, manual tasks. Sorting through mountains of usage data, reading user feedback, translating in-app messaging—the list could go on forever. 

When widespread adoption of conversational and generative AI shook up historical workflows, product managers (PMs) found creative ways to simplify processes and add AI capabilities to their products. The dust settled, revealing new ways to build and use AI-powered tools.

Nobody understands a company’s market, users, or product better than its PMs. Because of this, they’re uniquely positioned to impact the company’s strategic outcomes—and many are doing this by building AI-powered tools or weaving AI into existing ones. New tools come with new challenges, and the traditional PM playbook won’t cut it in the age of AI. 

To help you build the most impactful AI products, we compiled 20 best practices from Pendo’s own internal product managers. Consider this our Pendo on Pendo guide to building AI tools, across four strategic areas.

Research and development

R&D is an integral piece of building new products and improving existing ones. It’s also manual, time-consuming, and expensive. In 2022, R&D performed in the U.S. cost an estimated $885.6 billion—up $84.1 billion from 2021. 

To perform more effective and efficient R&D, Pendo PMs are using and building AI-powered tools with support from product analytics, customer intelligence, and session replays to:

1. Visualize and analyze user journeys to identify where AI can enhance the user experience (UX).
2. Use dashboards to share qualitative, quantitative, and visual data with their AI development team to align PMs and engineering.
3. Prioritize AI feature development and enhancements with product usage data, user feedback, validation, and session replays.
4. Iteratively improve AI features based on user interactions and feedback.
5. Identify opportunities to streamline workflows with AI-driven automation.
6. Simulate user interactions before launching a product to find and resolve potential issues.

Data analysis

Before building analytics into products became the status quo, PMs didn’t have enough data to inform decisions. Now, they’re faced with the opposite problem: Too much data, not enough information. 

Data can only become information with context. At Pendo, we use AI to make it easier to understand and act on data, whether we’re analyzing qualitative inputs from review sites to deeply buried usage patterns. Here’s how Pendo PMs analyze large batches of data with AI, for our AI products:

1. Measure engagement with AI features to understand adoption, value, and impact on your business outcomes.
2. Create cohorts based on user behavior and metadata to tailor AI features for different user groups.
3. View where users click and interact the most with your AI features.
4. Analyze the journey users take when interacting with your product’s AI components.
5. Compare different user cohorts via segments to determine how AI features influence user and account retention rates.

Customer intelligence

Knowing each customer comes with the territory of being a small-scale startup. As your business expands, there’s simply not enough time to maintain personal relationships with each user. 

Your qualitative data (e.g. from NPS responses and in-app feedback) can be the golden ticket to planning, developing, and refining AI tools. To build better AI products, PMs should:

1. Gather user feedback on AI features to understand satisfaction and areas for improvement.
2. Understand user sentiment outside of your application with AI-powered analysis across social media, reviews, emails, and other third-party websites or applications. 
3. Collect feedback on AI functionality directly from users while they’re interacting with the product.
4. Inform stakeholders of progress on AI development and timelines.

Pendo on Pendo tip: To standardize, summarize, and find ideas from large volumes of qualitative data, PMs should use customer intelligence tools like Pendo Listen. Learn how you can use AI/ML tools for better feedback collection. 

User onboarding and adoption

Features are only as successful as their adoption rates. Onboarding, walkthroughs, and resource centers should be part of every new feature and tool you launch—especially for AI/ML tools that introduce new workflows, technologies, and ways of working. 

To ensure your AI features are as valued and adopted as possible, PMs should: 

1. Develop in-app content to teach users how to effectively use new AI features.
2. Track time to first use in order to monitor how quickly and widely users adopt AI features.
3. Identify and address barriers to adoption to help users get the most value from AI enhancements.
4. Use data to customize AI-driven recommendations and content for individual users.
5. Anticipate user needs and provide proactive support for AI features.

If you fail to plan, you’re planning to fail—especially when it comes to sparkly new AI features. Bring these 20 tips into your product planning, and see what you can do with Pendo AI.