Agentic AI system

Course Q&A Agentic AI
,
1

I’ve completed the excellent course and I have two questions. 1. I find some Agentic systems have two roles (system role and user role), but some only have user role, so what’s the reason? 2. After completing graded labs, I can’t modify any codes to test different results. Thanks

2

Hi Yiyi,

The default system role of LLMs is often set to ‘be a helpful assistant’. This setting can be very useful for interactions that are not too specialized. When a specialized topic comes up, e.g., when an LLM is to be used as a coding model, it can be very useful to set the system role to something like ‘You are an expert developer’, or the like. This activates that part of the LLM where attention to coding is high, and you can get more specialized answers.

3

Hi @Yiyi ,

I am not sure about your comment cannot modify code, once that in some lessons the learner motivated to change options, parameters, to see different results. In addition, you can download the entire lesson and run in your local computer.
Keep learning!

1 Like
4

hi @Yiyi

  1. When an Agentic System Uses System and User Roles (Dual Roles)

This type of agentic system allows sophisticated autonomy and complex workflows

  • System Role (The ‘Manager’ or ‘Persona’): Defines the AI’s identity, constraints, goals, and expertise (For example “You are a helpful travel assistant, prioritize sustainable options”)
  • User Role (The ‘Request’ or ‘Data’): Provides the specific task or data (for example “Book a trip to Kyoto for next month”).
  • When to Use: For complex, multi-step tasks, project management, personalized customer service, or anything requiring planning, context, and goal-directed action (e.g., analyzing market trends, drafting campaigns)
  1. When an Agentic System Uses Only the User Role (Single Role)

In a user only agentic system, the system functions more like a powerful tool, and not just as aautonomous agent.

  • User Role (The ‘Command’): The user provides direct input, and the AI responds based on its training and context, without a pre-defined, persistent persona or complex goal.

  • When to Use:

    • Simple, Direct Queries: Asking for facts, summarizing text, or generating content (for example simple Q and A based system asking what is the capital of Thailand.)

    • When Simplicity & Predictability is Key: For users who prefer the AI to act purely as a responsive tool, minimizing unexpected autonomous behavior.

    • Basic Tool Functions: When the AI is just fetching data or performing a single, well-defined action without needing to “manage” a complex project.

Main Difference: Autonomy versus Responsiveness

  • Dual Roles: Creates an agent that plans, decides, and acts autonomously (like a manager deciding which technicians to use).

  • Single Role (User Only): Creates a tool that responds to direct commands (like a technician doing a specific job)

Regards

DP

5

Good explanations, Thanks!

Deepti Prasad <notifications@dlai.discoursemail.com> 於 2026年1月21日週三 下午5:51寫道：

1 Like