How Campaign Agent makes decisions

Campaign Agent builds and optimizes campaigns through two types of intelligence: design-time intelligence builds the initial campaign, and run-time intelligence optimizes it after launch.

Design-time intelligence

When you provide a prompt, Campaign Agent consults a knowledge base to ensure your campaign is founded on proven best practices.

Learn from annotated campaigns

Campaign Agent's design-time intelligence is built on annotated datasets — thousands of high-performing campaigns executed within Bloomreach that teams have systematically analyzed and labeled.

Annotation means each campaign component — audience definitions, trigger events, timing delays, and content structure — has been tagged with information about what it does and how it performs. This structured knowledge helps Campaign Agent understand not only individual elements but also how they work together.

Context-specific insights

Campaign Agent combines large language model (LLM) knowledge with insights tailored to your context:

Industry: Fashion retailers send reactivation campaigns with discount offers, while B2B software focuses on feature updates and usage tips.
Region: European campaigns typically send during business hours (9 AM–5 PM local time) with formal messaging, while US campaigns favor evening sends with a casual tone.
Use case: Abandoned Cart campaigns can repeat multiple times per week, while newsletters are sent on a fixed schedule.

Learn more about how to provide effective context in your prompts. For details on design-time intelligence boundaries, see the AI optimization section in Campaign Agent limitations.

Data validation

Before generating a campaign brief, Campaign Agent checks whether your project data supports the campaign you're building. It reads your Data Mapping configuration, data structure, and existing catalog names to verify that the required data is available.

If it detects missing or unmapped data, it warns you and guides you to Data Mapping to resolve the issue before proceeding.

Run-time intelligence

Campaign launch starts the learning process. Run-time intelligence automatically refines campaigns based on real-time customer behavior.

Dynamic wait-time adjustment

Dynamic wait-time adjustment replaces fixed delays with intelligent timing for Abandoned Cart and Browse Abandonment campaigns. Campaign Agent analyzes each customer's context — cart value, purchase history, browsing behavior — to determine the optimal follow-up moment.

This optimization is enabled automatically during campaign brief generation for supported use cases. To learn more about wait-time options, see Wait node in Campaign Agent.

🚧
Important
Dynamic wait-time adjustment requires Contextual Personalization, a Premium tier feature. Contact your Customer Success Manager to learn more.

Optimization cycle

Optimization operates in three steps:

Decide: Campaign Agent chooses the optimal wait time based on customer context.
Act: Campaign Agent calculates the delay and sends the message.
Learn: Campaign Agent analyzes the outcome and feeds results back into the model.

For details on run-time intelligence scope and limitations, see the AI optimization section in Campaign Agent limitations.