Marketing workflows still rely on predefined segments, fixed journeys, and static messages. Those approaches assume predictable customer paths, but customers don't follow fixed scripts: they wander, double back, and change their minds. Many current tools are brittle, costly to maintain, and they don't know what other campaigns are doing to the same customer at the same time.
Campaign Agent takes a different approach: you define the strategy and brand constraints, and the agent builds and runs the campaign. It handles three core responsibilities:
- Content personalization: generates brand-compliant, multichannel, multilingual content variants grounded in CMS assets and brand guidelines, with marketer review and approval.
- Send prioritization and arbitration: evaluates each customer's intent and context in real time to decide which campaign should reach them and suppresses competing or redundant sends.
A marketer prompts Campaign Agent with a campaign goal (for example: grow subscriptions). The agent produces a complete campaign brief: objectives, KPI target, live dates, eligible audience (including inclusion/exclusion rules), budget, and initial content variants. Content assets are generated for multiple channels (email, RCS, MMS, SMS) and languages (English, French, Spanish) and are checked against brand guidelines. Marketers can iterate and approve at each step before the campaign goes live.
Once live, Campaign Agent continuously combines the marketer's strategy with each customer's real-time signals (browsing history, propensity models, past purchases) and makes three kinds of moves:
- Prioritization: selects which campaign should reach an individual customer and suppresses those that don't fit the moment. For example, it may hold a seasonal campaign or welcome journey in favor of a more relevant promotion.
- Assembly: chooses the exact language, content variant, channel, and send time for the customer (for example, Variant 3 via SMS in the morning in the customer's language).
- Optimization: adjusts budget allocation, channel weights, offers, send times, or audience size while the campaign runs, including pausing underperforming elements or expanding with lookalikes.
The piece is authored by Brett Bothe and Max Kiely on the Salesforce blog.