🌿 AI Agents

An AI agent acts as an autonomous computational system that can determine its behaviour based on its environmental context, including data input. It is also capable of adapting its behaviour over time.

An autonomous, goal-directed AI system consisting of AI agents is known as Agentic AI.

Whilst traditional generative AI models require user interaction by way of prompts, Agentic AI models are digital ecosystems which use LLMs to take initiative, including making decisions and adapting to inputs and the environment appropriately.

According to Amazon Web Services (quoted in Oxford AI Course), there are 4 key features of Agentic AI systems:

• Proactivity: the ability to anticipate needs, identify patterns and address issues without human prompting.
• Adaptability: the ability to understand changing environments and adapt accordingly. An example might be understanding a particular customer's concerns and including context sensitive information in a reply.
• Collaboration: the ability to collaborate with other AI systems and humans.
• Specialisation: agentic systems use specialised AI agents with narrow areas of expertise.

AI agents communicate with each other using the 🌱 A2A protocol.

It is Agentic AI where the big efficiency gains are expected and the current focus of most organisations seriously investing in AI. 🥶 Agentic AI is already changing the workforce - Stave, Kurt and Winsor

Agentic AI is different from but complementary to Generative AI. Agentic AI can operate as a central brain which determines what tasks are needed and assigns them to the necessary tools, which may include Generative AI.

For more information, see https://www.oracle.com/uk/artificial-intelligence/agentic-ai/ (contains steps to getting started with Agentic AI).

For the development of Agentic AI, see 🌱 From Machine Learning to Agentic AI