Appier, an AI-powered SaaS platform, released its latest whitepaper mapping out the core capabilities of agentic AI today in a marketing context.
The paper distinguished between large language models (LLMs) – which serve as an “engine” through reasoning and content generation, and agentic AI – which takes on the role as a proactive “pilot”.
The key difference between agents and LLMs like ChatGPT and Claude, Appier outlines, is an agent’s ability for always-on learning and continuous decisioning.
In this sense, LLMs exist as a reactive tool that waits to be told what to do, whereas agents act autonomously, having a goal and figuring out how to achieve it.
The paper also outlines a phenomenon that’s emerged in marketing called the “Autonomy Gap”, where human analysis alone is incapable of keeping up with the large quantities of customer data that now exists.
Some specific marketing tasks appropriate for agents, the paper identifies, includes finding audience clusters based on behavioural signals, generating and testing creative variations, running experiments across channels, recommending budgets or targeting, and orchestrating multi-step journeys.
What Characterises Agentic Systems?
The Appier paper outlined a number of characteristics unique to agentic marketing.
This includes the ability of agents to have autonomy and operate across numerous platforms, to adapt to differing conditions in real-time, to proactively plan through breaking down objectives into actionable tasks, to work towards goals through a complex journey, to coordinate with other agents, and to continuously learn and improve.
These factors distinguish agents from other forms of AI, while the report found that the key functions of agentic AI in marketing include:
- Autonomous planning and execution – understanding objectives and putting them into actionable tasks and priorities.
- Real-time optimisation – identifying audience clusters, generating and testing creative variations, and executing budgets and targeting in real-time.
- Learning continuously – adapting behaviour over time based on performance outcomes and changing conditions.
- Carrying out multi-step orchestration – managing complicated customer journeys and coordinating across different channels.
