Agent
An AI entity assigned to each user, learning preferences through interactions and optimizing messaging in real time. Agentic Infrastructure
The system of agents, content, and labels that enables true 1:1 personalization at scale. Agentic Learning
Continuous adaptation through trial, feedback, and reinforcement, balancing exploration and exploitation. Alternates
Variant options for a single message component (e.g., three different CTAs) that agents mix and match. Components
Building blocks of a message (Greeting, Value Proposition, Offering, Incentive, CTA, Tone, Other). Content as Infrastructure
Approach where messages are not campaigns but a reusable, labeled library that fuels learning. Content Coverage Map
A structured inventory mapping product features and workflows to Topics, showing where content exists and where gaps remain. Contextual Bandit
A machine learning framework where agents select actions (e.g., messages) based on user context, balancing personalization with learning. Evergreen Content
Always-on messages (e.g., feature education, FAQs) that form the bulk of agent learning. Exploration vs. Exploitation
The balance agents strike between testing new content and using proven, high-performing variants. Inner Voice Statement
User-centered phrasing of a value proposition (e.g., “I want to feel in control of my finances”). Labels
Semantic tags (e.g., “Convenience,” “Urgency”) applied to alternates, enabling agents to reason and compare systematically. Message Group
A set of related messages made by combining alternates and labels, tested together for personalization. Offering
The specific product, feature, or benefit highlighted in a message. Reinforcement Learning (RL)
Learning approach where agents adjust strategies based on observed user behavior and outcomes. Reward Function
The feedback mechanism agents use to determine whether a message or action led to a desirable outcome. Surfaces
Channels or placements where messages appear (push, email, SMS, in-app, web). TACIR (Target Agent-Customer Interaction Rate)
The ideal frequency of interactions per week between agents and users, used as a diagnostic for content sufficiency. Thompson Sampling
A probabilistic method agents use to select actions, balancing exploration of new options with exploitation of known successes. Three-Fold Agentic Content Strategy
Framework dividing content into Evergreen (60%), Triggered (30%), and Tactical (10%) categories. Topic
A high-level communication category (e.g., “Welcome,” “Churn Rescue”) that bundles offerings, audiences, and contexts. Triggered Content
Event-based messages delivered in response to user actions (e.g., cart abandonment, subscription pause). Value Proposition
The core reason users should care about an offering, often tied to human values (e.g., Control, Trust).