IDSC 4590 · Carlson School of Management · University of Minnesota
Agentic AI for Business Applications
A 7-week undergraduate course on generative and agentic AI for business students. No programming background required.
2 credits Fall B-term Mon & Wed, 9:55-11:35 No coding required Prerequisite: IDSC 3001 or consent
About the course
Generative AI has moved from novelty to operational infrastructure in under three years, and agentic AI, systems that plan, act, and coordinate autonomously, is accelerating that shift. Most business graduates enter the workforce able to use a chatbot, but without the conceptual vocabulary to evaluate AI claims, the workflow intuition to spot automation opportunities, or the grounding to lead AI initiatives. This course addresses that gap.
The course is designed for business majors who will manage AI tools, commission AI projects, and make adoption decisions, rather than build AI from scratch. The goal is informed fluency: what AI can and cannot do, how to translate business problems into AI solutions, and how to lead responsibly in AI-augmented organizations. It is open to all Carlson concentrations and requires no prior technical background.
Learning objectives
By the end of the course, students can:
01
Explain
How large language models work, and why they err, in plain business language.
02
Distinguish
Rule-based, predictive, generative and agentic AI, and match the right one to a business problem.
03
Build
Simple no-code AI workflows that connect AI outputs to real business processes.
04
Evaluate
How agentic systems reason and fail, and how much autonomy they should have.
05
Assess
The strategic, ethical and governance implications of AI adoption, including risk, regulation and vendor choice.
06
Pitch
A scoped, evidence-backed AI solution to a real business problem, live at the AI Innovation Showcase.
Course design
Each week has two 100-minute sessions. The first introduces concepts through short lectures, live demonstrations and case discussion; the second is hands-on, with students building AI workflows and agent designs using no-code tools (Gemini, Google AI Studio, Make, and n8n running locally in Docker).
Three build sprints anchor the applied work: students ship a workflow automation, build a first agent, then red-team a peer's agent and harden their own. Assessment also includes four reflection posts, a midterm and final (each split into a closed-book part and an open-AI part), weekly check-ins, and a team capstone worth 25% of the grade, presented at the AI Innovation Showcase.
Learning by doing, every session
Every session includes at least one structured activity. Students audit, argue, build, break and pitch; lecture segments stay short.
Structured in-class activities
Personal task audits, agent blueprint cards, case analysis (a help-desk agent that closed everything), vendor-question drills, pitch rehearsals.
Debates with a position
Does AI multiply a team or replace it? Students take a side, defend it, and confront the strongest counterargument.
Live demos, real failure
Models vary, hallucinate and break on screen. Students learn verification by watching it go wrong, not by being told.
Three build sprints
A workflow automation, a first agent, then a red-team round: attack a peer's build, harden your own, document what you found.
Authentic assessment
The capstone is pitched live to external guests at the AI Innovation Showcase and scored like a real pitch, not a class presentation.
Reflection as a habit
Four graded posts on what you did, how you checked it, and what you would change, building verification into the routine.
Weekly outline
| Week | Theme | Session 1 (concepts) | Session 2 (hands-on) |
|---|---|---|---|
| 1 | GenAI Foundations | How AI Works and How to Work With It | Beyond Chat: Workspaces, Connectors, Agents, and Enterprise RAG |
| 2 | Workflow Automation | Workflow Thinking: What Gets Automated | Build Sprint 1: Your First Automation |
| 3 | Agents and Capstone Kickoff | What Is an AI Agent? | Capstone Launch and Team Formation |
| 4 | Guardrails and Agents | Failure Modes, Guardrails, and Checking at Scale | Agents in the Wild and Build Start |
| 5 | Build and Responsible AI | Midterm Part A, then Build Sprint 2: Your First Agent | Responsible AI |
| 6 | Red Team and Enterprise AI | Build Sprint 3: Red Team and Fix | AI in the Enterprise and Idea Pitching |
| 7 | Showcase and Strategy | AI Innovation Showcase | Strategy and the Long View |
A sample of Session 1 from Week 1, with the full slide deck, is on the sample session page.
Capstone event · Week 7
AI Innovation Showcase
The course ends with a trade-show-style showcase rather than a final presentation day. Cross-concentration student teams staff stations while practitioners, faculty and guests rotate through, spending several minutes with each team: the problem, the design, the business value, what could go wrong, and what comes next.
External audience
Guests include practitioners who make real AI decisions. Questions are unscripted, and honest uncertainty scores better than bluffing.
Judged like a pitch
Guests and peers score each team on problem and business value, solution design, responsible AI, and the quality of pitch and answers.
Show the experience
Every team walks guests through the user experience with a mock, screens or a storyboard. A working no-code demo is a bonus, not a requirement.
Capstone projects from the Fall 2026 cohort will be featured here after the Showcase on December 14, 2026.
This course runs on an AI-native workflow. Everything students see, from the syllabus to slide decks and Canvas pages, starts as plain text in a git repository, produced by a coding agent under instructor control. Read about it: AI-Native Teaching: My New Course Development and Maintenance Workflow
Instructor

Dr. De Liu is the Xian Dong Eric Jing Professor of Information and Decision Sciences at the Carlson School of Management, University of Minnesota. He studies and teaches how AI, analytics and digital platforms reshape business.