Unit 5: AI Agents
5a. Differentiate traditional workflows from AI agents.
How does a traditional workflow follow a fixed path from trigger to action?
What makes an AI agent different from a rule-based or AI-enhanced workflow?
Which tasks fit a workflow, and which tasks fit an agent better?
If a task changes during execution, which approach gives you more flexibility, and why?
Traditional workflows move through a set sequence of steps. A trigger starts the process, conditions decide the path, and actions carry out the work. In the earlier unit on workflow logic, you saw that a workflow is built to follow rules. An ai agent is different because it can interpret a goal, decide what to do next, and adapt its steps while it works. So, a workflow is best when the path is known in advance, but an agent is better when the task needs judgment or can change during execution.
A key difference is control. In a traditional workflow, you define the steps before it runs. In an agent, you give a goal and the agent helps plan or adjust the next step. That means agents can support tasks that are not fully predictable. Students often get this wrong by treating agents as just another workflow step. They are not the same. A workflow follows a fixed structure; an agent can use more flexible decision making inside that structure or instead of it. Can you see why a changing task often needs more than simple rules?
To review, see:
[5.1 What Are AI Agents?]
[1.2 Types of Automations: Rule-Based, AI-Enhanced, and Agents]
5b. Build simple AI agents with free tools and prompting techniques.
What basic parts do you need to build a simple AI agent with free tools?
How does a clear prompt help the agent understand its job?
What step-by-step setup would you use to make the agent carry out a simple task?
If the agent gives an unclear result, which part of the setup would you change first?
To build a simple agent, you start with a free tool and a clear prompt. A prompt is the instruction you give the AI system so it knows what to do. The course materials on free tools and prompt writing show that good setup begins with a clear task, a defined output, and simple instructions. For a basic agent, you do not need a complex system. You need a goal, a tool that can run the agent, and directions that tell it how to respond.
A simple way to build the agent is to state the task, describe the role it should play, and specify the output you want. For example, if the goal is to summarize incoming messages, you would tell the agent what to look for and how to format the result. The important point is that the prompt should reduce confusion. If the output is unclear, you usually improve the prompt first, because weak instructions often cause weak results. Students often think the tool is the main issue, but the instructions are usually the first place to check. To build well, you must match the task to the tool and keep the prompt direct.
To review, see:
[1.3 Exploring Free No-Code Tools]
[4.4 Writing Effective Prompts for Automation]
[5.2 Building Your First Simple AI Agent]
5c. Design agent behaviors that support decision-making and multi-step task execution.
What behaviors help an agent make useful decisions instead of random ones?
How can you design an agent to complete more than one step in order?
Which behavior should the agent use when one result leads to the next step?
How would you decide whether an agent needs simple instructions or a more structured plan?
To design helpful agent behavior, you give the agent clear rules for choosing what to do next. The course materials on agents and conditional logic show that decision making depends on well-defined conditions, possible paths, and clear task goals. An ai agent can support decision-making when you tell it how to choose among options, what information matters, and when to continue or stop. Without that guidance, the agent may respond in ways that do not fit the task.
For multi-step work, the agent needs an ordered process. It should know the goal of each step, what information to gather, and what output should move to the next step. This is similar to the way a workflow uses conditions to branch, but here the agent may help choose the next move inside a broader task. For example, in a task that first reviews input, then selects a response, and then prepares an output, the agent needs a sequence that matches the task order. The main challenge is not adding many steps; it is making each step useful for the next one. If the task has several possible paths, you need enough structure to keep the agent from drifting.
To review, see:
[3.3 Adding Conditional Logic]
[5.1 What Are AI Agents?]
[5.3 Adding Decision-Making to Agents]
5d. Combine AI agents with no-code workflows to handle complex scenarios.
When should you place an agent inside a no-code workflow instead of using only one or the other?
How do workflow steps and agent steps work together in a complex process?
What part of the process should the workflow control, and what part should the agent control?
If the task has both fixed steps and uncertain steps, how would you divide the work between them?
You combine an agent with a no-code workflow when part of the task is fixed and part needs flexible judgment. A no-code workflow is still useful for stable steps such as moving data, checking conditions, or starting a process. An ai-enhanced workflow uses AI inside a workflow, while an ai agent can take on more flexible task decisions. In a complex scenario, the workflow can handle the reliable structure, and the agent can handle the part that needs interpretation or selection.
This division matters because it keeps the process easier to manage. The workflow can route information, start the agent at the right time, and pass the agent's output to the next step. The agent can then make a decision, choose among options, or complete a task that is not fully predictable. A practical design might use a workflow to collect input, send it to an agent for analysis, and then return the result to a later step. Students often try to make the agent do everything, but that can make the process harder to control. A better design is to let each part do what it does best: the workflow provides structure, and the agent provides flexibility.
To review, see:
[4.3 Creating Responsive AI-Enhanced Workflows]
[5.1 What Are AI Agents?]
[5.4 Combining Agents with No-Code Workflows]
Unit 5 Vocabulary
This vocabulary list includes terms you will need to know to successfully complete the final exam.
decision-making

Last modified: Thursday, 18 June 2026, 1:29 PM
Callback before_footer in local_aigrade component should be migrated to new hook callback for core\hook\output\before_footer_html_generation
  • line 7225 of /lib/moodlelib.php: call to debugging()
  • line 7292 of /lib/moodlelib.php: call to {closure}()
  • line 71 of /lib/classes/hook/output/before_footer_html_generation.php: call to get_plugins_with_function()
  • line 987 of /lib/classes/output/core_renderer.php: call to core\hook\output\before_footer_html_generation->process_legacy_callbacks()
  • line 94 of /mod/page/view.php: call to core\output\core_renderer->footer()