This recipe focuses on making expansion suggestions feel like helpful tips rather than intrusive sales.
Generic upsell banners rarely hit the mark. Customers respond better when suggestions match what they already do and where they feel friction. A usage-based recommender uses actual product activity to decide when and what to propose.
Define an agentic AI Skill that triggers based on a question of an end user that relates t the feature that you want to upsell, if the user belongs to an Audience that has no access to said feature. From these, the skill can show a list of recommended_offers. If you want to go wild, you can even encode simple rules to trigger readiness, such as sustained usage near limits or repeated manual workflows that a premium feature automates.
Ground the skill on flat Topics such as Product-bundles, Usage-thresholds, and Expansion-playbooks. These Topics contain your own definitions of when to suggest which product, avoiding speculative recommendations. The AI then explains suggestions using language taken from these Topics.
For accounts that meet readiness criteria, you can use two paths. The first is internal: create tasks for account owners with a summary of usage and suggested talking points. The second is customer-facing: show in-app notifications to segmented users in the portal, such as “You are close to your current limit; here is an option to remove this cap.”
Use audience segmentation based on upsell_readiness, feature_usage, and segment to keep notifications relevant. For high-touch segments, you may choose to suppress in-app offers and rely only on human outreach using the AI-generated context. For lower-touch segments, in-app offers can be the primary channel.
Conclusion
A usage-based upsell and upgrade recommender turns product data into considerate, context-aware suggestions. With an AI Skill grounded in your own expansion playbooks, plus tasks, notifications, and segmented in-app prompts, it lets growth feel like a natural extension of value rather than a generic sales push.