This project is the main deliverable of the course and represents your opportunity to design a thoughtful, practical, and innovative use of agentic AI. Your goal is to select a real problem in a setting you understand well, create a functioning AI-powered workflow—implemented at least in part using n8n—and present the idea, the prototype, and the path to impact through a polished pitch video.

The video is not just a technical demo. It is a pitch to real stakeholders (or potential investors) explaining why your idea matters, how it creates value, and how it could realistically be implemented. The technical prototype is an important part of this, but it is only one piece of the broader story.

Although the project is graded individually, the underlying idea may be shared with one or two peers, subject to the collaboration rules in Section 10.

1. Project Overview

Agentic AI refers to systems that do more than generate a one-off answer: they use reasoning, tools, and workflows to support or automate parts of a process. In this project, you will design such a process to address a meaningful task in a company, organization, or personal domain where you have enough insight to understand the real constraints, goals, and opportunities.

The outcome should be both technically sound and managerially grounded. From a technical perspective, you will build a working proof-of-concept in n8n that uses AI in a non-trivial way. From a managerial perspective, you will use the frameworks from Module 3 (“From AI to Impact”) to explain what kind of value your system creates, how it integrates into a broader process, what risks it entails, and how it might actually be deployed.

Because the course is short, you will need to choose a focused, well-defined idea and iterate from there. Choosing a domain you already know makes this significantly easier.

2. Application Examples

To help you imagine possibilities, here are a few examples across different domains. These are meant as inspiration; you are encouraged to pick something that truly resonates with your experience and where you can make a credible case for impact.

1. A system that produces daily summaries of news relevant to an investment firm’s specific portfolio, improving the speed and quality of decision-making.
2. A workflow that listens to transcripts of doctor–patient conversations and helps clinicians update medical records more consistently and accurately.
3. A personal email assistant that digests your inbox every morning, drafts responses for important messages, ignores newsletters, and manages follow-ups.
4. A tool that helps Kellogg students choose classes based on their background, career goals, and course ratings.
5. A workflow that helps consulting teams manage expenses by consolidating receipts, categorizing charges, and flagging potential issues automatically.
6. A system that assists an insurance company by transforming medical records and police reports into professional demand letters.

These examples illustrate the range of acceptable projects: internal tools, personal productivity agents, startup ideas, or process improvements inside organizations.

3. What You Will Submit (High-Level Summary)

Your submission on Canvas will include the following components.

Deliverable 1: A pitch video (7:00–8:15)

You will create a self-contained video that feels like a professional pitch to decision-makers in your chosen context. The video must explain the problem, the AI-enhanced solution, the value and impact using ideas from Module 3, a live demonstration of your workflow, your evaluation approach, and a realistic path to implementation.

Deliverable 2: Your n8n workflow(s)

You will submit at least one exported workflow file (usually a .json file) that corresponds to the workflow demonstrated in your video. If you have multiple workflows—for example, a main workflow and a separate evaluation workflow—you are encouraged to submit them all.

Optional Deliverable 3: Additional supporting artifacts

If you created slides, an interface (e.g., with Lovable), or any other materials that help communicate your idea, you may upload them as well. Slides are not required, but you may include them if you find them helpful.

4. Deliverable 1 · The Pitch Video

The video is the centerpiece of your project. Think of it primarily as a pitch about impact and value, supported by a live demo of your prototype, rather than as a step-by-step technical walkthrough.

4.1. What Your Video Must Contain

Your video should begin with a clear, engaging explanation of the problem you are addressing. Describe the organizational context, the pain point, and why this problem is worth solving. You should make it easy for a non-technical decision-maker in that context to understand what is at stake.

You should then introduce your AI-powered solution as you would in a pitch meeting. Rather than focusing first on the architecture, focus on the benefit: what will this system enable? How will it change the way people work, decide, or experience a process? Connect this to the ideas from Module 3, such as improvements in quality, speed, cost, risk reduction, or user experience.

Once the value is framed, you will explain how the solution works. At this stage, it is appropriate to walk through the key components of your workflow, but you should keep linking the technical details back to the impact story. You are not just saying “I used n8n and an AI node,” but rather “Here is how this workflow helps achieve the outcome I just described.”

Your video must include a live demonstration of the system. During this demo, you should show the n8n workflow, run the AI-related parts in real time, and comment on what is happening. Screenshots alone are not sufficient; we need to see the workflow in action.

You must also include a discussion of your evaluation process, again using an impact mindset rather than purely technical metrics. Explain how you tested your system, how you chose your test cases, and what you learned from them. Show at least a few concrete examples in the video: the inputs you used, the outputs you obtained, and why you deem them satisfactory or not.

Finally, your pitch should address implementation and change management. Using the language and concepts from Module 3, explain how this AI-driven process could be integrated into a broader system, what changes people or organizations would need to make, what risks or failure modes you anticipate, and how you might mitigate them. A strong pitch makes it clear not only that something is technically possible, but that it is worth doing and realistically implementable.

4.2. Minimum Expectations for the Video

At a minimum, your video must:

• Clearly state the problem, the context, and the stakeholders.
• Explain what your system does and why that matters, using value-oriented thinking from Module 3.
• Show a live demo of your n8n workflow, including the AI component running.
• Demonstrate your evaluation process with concrete examples.
• Discuss how the solution could be implemented and how it would create value in practice.

Slides are optional. If you choose to use slides, they should support your story rather than replace the demo or the explanation. It is completely acceptable to give your pitch using only your face, your voice, and your screen showing n8n or other tools.

4.3. Technical Requirements and Timing

Your video should feel like a professional, coherent pitch. You must appear on camera at least part of the time, and your screen must be visible during the demo. The audio should be clear enough for us to follow your reasoning and explanations.

The target length is around eight minutes. For full credit, the video must be between 7:00 and 8:15. Videos that are significantly shorter or longer may receive reduced credit because it becomes harder to fit all the required elements into a coherent pitch.

4.4. Recording Your Video with Zoom (Recommended)

While you may use any recording tool, Zoom is often the easiest option for Northwestern students. Here is a suggested workflow:

1. Open Zoom and start a new meeting.
2. Turn on your camera and microphone, and quickly check that lighting and sound are acceptable.
3. Share your screen, either by selecting a specific window (such as slides or the n8n interface) or your entire desktop.
4. Position your video thumbnail so that your face is visible without blocking important content on the screen.
5. Click Record → Record to this Computer to start recording.
6. Deliver your pitch as if you were talking to decision-makers: first explain the problem and the value, then walk through the solution, then show the live demo and evaluation, and finally discuss implementation and risks.
7. Stop the recording when you are finished.
8. End the meeting. Zoom will automatically save the recording to a folder on your computer (typically under Documents/Zoom).

You may lightly edit your recording if you wish, but this is optional. A single, well-prepared take is perfectly acceptable.

5. Deliverable 2 · Your Workflows and Evaluation Artifacts

5.1. Minimum Expectations for Your Workflow

You must submit at least one workflow created in n8n that uses an AI node. This AI node is often an AI Agent node, but you may use any AI-related node that fits your design. The AI portion of the workflow must be runnable so that the teaching team can execute it and see the behavior you demonstrated in your video.

Your workflow does not need to be a fully polished production system. It is entirely acceptable to have parts that are partial or illustrative, such as placeholder steps or incomplete integrations. What matters is that:

• The core AI-driven part of the process runs, and
• The overall structure of the workflow clearly reflects the process you described in your pitch.

5.2. Evaluation Artifacts and Expectations

Your evaluation can be implemented in several ways. One option is to build a separate evaluation workflow using n8n’s evaluation nodes. Another option is to design a manual evaluation with a small but meaningful set of test cases. In either case, you must describe and demonstrate this evaluation in your video.

A minimal but acceptable evaluation includes:

• A set of representative test inputs,
• The corresponding outputs produced by your workflow, and
• Your explanation of why these outputs are good, problematic, or mixed, and what you would change.

If you built an explicit evaluation workflow in n8n, you are encouraged to export and submit it alongside your main workflow, but this is not strictly required.

5.3. Exporting Workflows from n8n

Exporting a workflow from n8n works as follows:

1. Open the workflow you want to export.
2. Click the three dots (…) in the top-right corner of the n8n interface.
3. Select Download.
4. This will download a file, typically in .json format, containing your workflow.
5. Upload this file to Canvas without modifying its format.

If you have more than one workflow that you use in your demo—for example, a main workflow plus an evaluation workflow—you may export each of them and upload all of the resulting files.

6. Submitting Your Project on Canvas

Canvas allows you to upload multiple files. Your submission should include:

• Your video file, in a standard format such as .mp4 or .mov.
• At least one exported n8n workflow file (.json).
• Any optional supporting materials such as slides, images, or documentation, if you believe they help tell your story.

The video should be playable without special software, and the workflow files should remain in their original JSON format so that we can import them into n8n if needed.

7. Grading Rubric

Your grade will be determined using the following rubric. Notice that impact and pitch quality are just as important as technical implementation.

8. Timeline and Milestones

9. Getting Help

You are not expected to do this in isolation. You have several resources:

• Recitations will introduce n8n, AI nodes, and evaluation tools with concrete examples.
• Office hours with the professor and the TA are available every week.
• Kai, the course’s AI assistant, can help you brainstorm ideas, refine prompts, and think about evaluation.
• Slack is available for asking questions, sharing ideas, and getting feedback from peers.
• Homework assignments will serve as project checkpoints so that you do not fall behind.

10. Advice and Collaboration Rules

10.1. Advice for a Strong Project

Strong projects usually start from a simple but sharp problem. When you understand a domain deeply, you can explain the value of your idea in a way that feels credible and specific. Use the lenses from Module 3—such as efficiency, quality, risk, user experience, and organizational change—to structure your thinking. Ask yourself: if I were a decision-maker in this context, would this pitch convince me?

Avoid the temptation to chase complexity for its own sake. A simple workflow that clearly creates value and is thoughtfully evaluated is much more impressive than a sprawling system that is hard to explain or justify.

Finally, start early and iterate. Your first idea will almost never be your best; your goal is to converge to something you are genuinely proud to pitch.

10.2. Collaborating with Others

You may collaborate conceptually on the same project idea with up to three students in total. However, grading remains strictly individual, and each student must submit their own deliverables.

In particular:

• Each student must record their own pitch video.
• Each student must create and submit their own workflow(s), focusing on the parts they personally developed.
• Each video should highlight a different angle or part of the system so that there is no simple duplication.
• The total amount of work for a shared project idea should scale appropriately with the number of participants.

11. Submission Checklist

Before Submitting Your Project, Make Sure You Have…

Your Pitch Video

• A clear explanation of the problem, context, and stakeholders.
• A value-focused story answering the question: why would your project be useful? Get inspiration from what we did with the Proxima case.
• A live demonstration of your workflow, showing the AI component running in real time.
• A demonstration and explanation of how you evaluated your process.
• A discussion of implementation, risks, and change management.
• Your face visible at least part of the time and your screen visible during the demo.
• Clear, understandable audio.
• A total length between 7:00 and 8:15.
• Your video uploaded in a standard format (for example, .mp4 or .mov).

Other files

• At least one n8n workflow (.json file) containing a runnable AI component (the entire workflow does not necessarily have to “work”).
• Optional: Any other thing you want to attach if they help tell your story.

12. Appendix and Resources

12.1. “Getting Started with the Project” Video

If you would like to revisit the minimum expectations or see a guided overview of how a simple n8n workflow can be constructed and evaluated, you may refer to the following recording:

This video provides a helpful example of what a minimal but complete project could look like.

12.2. AI Agent Starter Workflow (n8n Template)

Below is a simple template to help you begin experimenting with n8n’s AI Agent node. You may paste it directly into your n8n workspace.

{
  "nodes": [
    {
      "parameters": {
        "promptType": "define",
        "text": "Data to be transformed here",
        "options": {
          "systemMessage": "Instructions on how to transform data here"
        }
      },
      "type": "@n8n/n8n-nodes-langchain.agent",
      "typeVersion": 2.2,
      "position": [
        208,
        0
      ],
      "id": "5c5e0d72-fb02-4704-9f38-b8d33ef9cc55",
      "name": "AI Agent"
    },
    {
      "parameters": {
        "model": {
          "__rl": true,
          "mode": "list",
          "value": "gpt-4.1-mini"
        },
        "options": {}
      },
      "type": "@n8n/n8n-nodes-langchain.lmChatOpenAi",
      "typeVersion": 1.2,
      "position": [
        160,
        192
      ],
      "id": "1f76eeb9-340f-4921-a124-140d33248d74",
      "name": "OpenAI Chat Model"
    },
    {
      "parameters": {},
      "type": "n8n-nodes-base.manualTrigger",
      "typeVersion": 1,
      "position": [
        0,
        0
      ],
      "id": "ccdaaba0-bdb1-4a28-b3b2-cd2cc2b92efc",
      "name": "When clicking ‘Execute workflow’"
    }
  ],
  "connections": {
    "OpenAI Chat Model": {
      "ai_languageModel": [
        [
          {
            "node": "AI Agent",
            "type": "ai_languageModel",
            "index": 0
          }
        ]
      ]
    },
    "When clicking ‘Execute workflow’": {
      "main": [
        [
          {
            "node": "AI Agent",
            "type": "main",
            "index": 0
          }
        ]
      ]
    }
  }
}

This workflow will allow you to start experimenting with system messages, user data, and simple transformations before expanding into a more complete solution.

Enjoy building your project. Treat this as a chance to practice exactly what Module 3 is about: going from “AI is possible” to “AI is valuable, implementable, and worth pitching.”