Course syllabus

Vibe Coding for Business Leaders: From Business Problem to Working AI Product · MBA USM elective · 2 SCT-Chile

Format follows the USM MBA Programa de Asignatura template. A Word version in the same template is available here: Vibe_Coding_MBA_Elective_Syllabus.docx.

1. Course identification

Course Elective: Vibe Coding for Business Leaders. From Business Problem to Working AI Product
Department Ingeniería Comercial
Code MGE 4xx (to be assigned)
Prerequisites Those indicated in the admission profile. No programming experience required.
SCT-Chile credits 2
Exam No
Direct teaching hours 20 (15 lecture/workshop + 5 assessment)
Autonomous work hours 25
Total chronological dedication 45 hours
Language English
Knowledge area (OECD) Economics and Business

2. Course description

This course belongs to the elective area of the MBA. In it, students analyze business problems that can be solved with software and develop the ability to build, deploy and govern working AI-powered applications by directing AI coding assistants in natural language, without prior programming experience.

Vibe coding, the practice of building software by describing intent to an AI and iterating on what it produces, has collapsed the cost of a working prototype from weeks of engineering to hours of a manager’s time. This changes which ideas are worth testing, how managers specify what they need, how they work with technology teams, and which risks they now own. Over five intensive workshop sessions, students take one real business problem from framing to a deployed web application with an integrated language-model capability, evaluate the security, privacy, cost and maintainability risks of AI-generated software, and present a working product and business case to decision makers.

The course is deliberately built around build first, then reflect. Students ship a first application in the first session; strategy, user experience, governance and organizational questions are then examined on something they have made themselves. Students leave with a deployed product, a tested decision framework for build / buy / vibe, and a first-hand understanding of what has changed in the economics of building software.

3. Entry requirements / prior learning

Those indicated in the MBA admission profile. In addition:

  • No programming or software development experience is required or assumed.
  • Working reading and listening comprehension of English; deliverables may be written in English or Spanish.
  • A laptop (any OS) with a modern browser and administrator rights to install one application.
  • Accounts on the course toolkit (GitHub, an AI coding assistant, a managed database and a hosting platform). Estimated out-of-pocket cost for the course: USD 20 to 40 per student, confirmed before enrollment.

4. Graduate-profile competencies the course contributes to

Generic transversal competencies

  • CGT1. Act with high standards of professional, social and ethical responsibility in business management for the sustainable development of the organization and the community.
  • CGT2. Lead work teams in the organization to achieve personal and collective objectives.
  • CGT3. Communicate ideas effectively in written, oral and visual form, incorporating information technologies in professional performance, to contribute to the development of the organization.

Specific competencies

  • CE1. Analyze the organization and its environment to formulate plans that ensure the organizational Mission and Vision.
  • CE2. Solve business problems in changing environments that affect the different functional areas, integrating acquired knowledge and generating value for the organization.
  • CE4. Apply computing and technological tools (ICT) to management, as support for decision-making within the organization.
  • CE3 (secondary). Design strategic plans in complex scenarios for the sustainable development of companies or organizations.

5. Learning outcomes

From the generic transversal competencies

  • LO-G1 (CGT1, indicators 1.1, 1.2, 1.3). Evaluates the ethical, privacy and security implications of AI-generated software and of processing personal and company data in a prototype, and declares transparently the role AI played in their own work.
  • LO-G2 (CGT2, indicators 2.1, 2.2, 2.4). Leads a small team through an iterative build cycle with explicit roles, milestones and critical peer review of their own and others’ work.
  • LO-G3 (CGT3, indicators 3.1, 3.2, 3.3). Communicates a working product and its business case to decision makers through a live demonstration, a written memo and a visual presentation, adapting the message to a technical and a non-technical audience.

From the specific competencies

  • LO-S1 (CE1). Analyzes a business process and its context to identify, prioritize and frame an opportunity where an AI-powered application creates measurable value, specifying users, value hypothesis and success metric.
  • LO-S2 (CE2). Solves a business problem by specifying, building and deploying a functional web application with AI coding assistants, integrating a language-model capability with appropriate guardrails and an evaluation set.
  • LO-S3 (CE4). Applies AI coding assistants, language-model APIs, version control and managed deployment platforms as management tools, and decides between building, buying, continuing to prototype and handing off to engineering using explicit criteria.
  • LO-S4 (CE3). Designs a roadmap from prototype to production that identifies the strategic, organizational and technical conditions under which the solution should be scaled, and what remains a defensible advantage when building software is cheap.

6. Contents

Unit 1. The new economics of building software: vibe coding from hour one

  • What changed: from writing code to expressing intent. Vibe coding versus AI-assisted engineering; what each is good for.
  • The toolkit: conversational app builders, agentic coding assistants, version control, managed databases and hosting. What each costs and where the data goes.
  • Problem Pull versus AI Push. Selecting a business problem worth building for: users, pain, value hypothesis, success metric.
  • First build: from a one-paragraph brief to a deployed web application within the session.

Unit 2. From specification to product

  • Prompting as specification: product briefs, user stories and acceptance criteria written for an AI.
  • Iterative building: small steps, testing by using, reading what the assistant did without reading code, recovering from a bad change.
  • Data: forms, tables and a managed database; authentication for real users.
  • Version control for managers: commits as checkpoints, branches as experiments.
  • User experience for AI products: designing for non-deterministic output, trust, error states, loading and streaming.

Unit 3. Adding intelligence

  • Language-model APIs for managers: what a call is, what it costs, latency, context windows, choosing a model.
  • Integration patterns: classification, extraction with structured output, summarization, conversational assistant.
  • Retrieval over company documents (RAG): when it is needed, how it fails, how to test it.
  • Simple agents and automations: tool calling, when autonomy is worth its risk.
  • Guardrails and evaluation: building a test set, measuring hallucination, keeping a human in the loop.

Unit 4. Governing AI-generated software

  • Security: secrets and keys, authentication and authorization, prompt injection, the OWASP Top 10 for LLM applications.
  • Data protection: personal data under Chile’s Ley 21.719 and GDPR basics, data residency, vendor terms and training on customer data.
  • Cost control: token economics, rate limits, budgets, caching.
  • Maintainability: the code nobody understands, documentation, tests, technical debt in prototypes that became products.
  • Intellectual property and licensing of generated code; vendor lock-in.
  • Decision frameworks: build / buy / vibe; prototype versus production; when to stop vibe coding and hand off to engineering; the questions to ask an engineering team.

Unit 5. Demo Day and strategy revisited

  • Demo Day: live demonstration of the product and presentation of the business case to an invited panel.
  • What remains a moat when building is cheap: proprietary data, distribution, workflow integration, trust.
  • From prototype to production: roadmap, organizational implications, citizen developers and governance.
  • Reflection: what the AI did, what I decided, what I would do differently.

7. Methodology (teaching and learning strategies)

  • Workshop format, build first then reflect. Each 4-hour session follows the same rhythm: a 45-minute framework block, a 2-hour guided build in teams working on their own project, a 45-minute critique and reflection block, and a 30-minute milestone review.
  • Live demonstrations by the instructor of every technique before students apply it, including failures and recovery.
  • One cumulative team project (teams of 2 to 3) chosen by the students from their own organizations, carried from problem brief to deployed product to Demo Day.
  • Short pre-readings and a pre-work task before each session (60 to 90 minutes), so class time is spent building.
  • Individual build log: each student keeps a running record of prompts, decisions, failures and what the AI did, used for assessment and reflection.
  • Peer review at every milestone with a shared rubric.
  • Course toolkit (subject to update each term): an agentic AI coding assistant as the primary tool, a conversational UI builder, GitHub for version control, a managed database and a managed hosting platform. Principles are taught tool-agnostically so students can switch tools.
  • Asynchronous support through a course channel between sessions.

8. Assessment of learning outcomes

Component Weight Type Learning outcomes Due
M1. Problem brief and product specification 15% Individual LO-S1, LO-G3 Before session 2
M2. Working prototype v1, deployed, with build log 20% Team LO-S2, LO-S3, LO-G2 Before session 3
M3. AI feature with evaluation set and risk memo 25% Team LO-S2, LO-G1 Before session 5
Demo Day: live demo and business case (10 min + Q&A) 25% Team, individually adjusted LO-S4, LO-G3, LO-G2 Session 5
Individual reflection memo and participation 15% Individual LO-G1, LO-S3 End of session 5

Approval and grading. Grades on the 0 to 100 scale used by the MBA; the course is passed with a weighted average of 55 or higher and attendance of at least 75% of sessions, in accordance with MBA regulations. There is no final exam. Team grades are adjusted individually by up to ±10 points based on the build log and peer assessment.

Rubrics. Each component is graded on four criteria with a published rubric: problem and value (is the problem real, the user named, the metric measurable), functionality (does the product work for its user end to end), AI quality and safety (is the model capability useful, evaluated and guarded), and communication (is the product and decision explained clearly to its audience).

Policy on AI use. Using AI tools is mandatory in this course and must be declared. The build log is the record of that use. Presenting AI output as one’s own reasoning without the log, or fabricating a log, is academic dishonesty under MBA regulations.

9. Learning resources

Basic bibliography

  • Mollick, E. (2024). Co-Intelligence: Living and Working with AI. Portfolio / Penguin.
  • Willison, S. (2025). “Not all AI-assisted programming is vibe coding (but vibe coding rocks)” and “Vibe engineering”. simonwillison.net.
  • Karpathy, A. (2025). Post introducing the term “vibe coding”, 2 February 2025.
  • Anthropic (2024). Building effective agents. anthropic.com/engineering.
  • OWASP Foundation (2025). OWASP Top 10 for LLM Applications 2025.
  • Iansiti, M., & Lakhani, K. R. (2020). Competing in the Age of AI. Harvard Business Review Press.

Recommended bibliography

  • Huyen, C. (2025). AI Engineering: Building Applications with Foundation Models. O’Reilly. Selected chapters.
  • Agrawal, A., Gans, J., & Goldfarb, A. (2022). Power and Prediction: The Disruptive Economics of Artificial Intelligence. Harvard Business Review Press.
  • Ries, E. (2011). The Lean Startup. Crown Business.
  • Thomke, S. (2020). Experimentation Works. Harvard Business Review Press.
  • Ley N° 21.719 (2024). Regula la protección y el tratamiento de los datos personales y crea la Agencia de Protección de Datos Personales. Chile.
  • NIST (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0).
  • Official documentation of the course toolkit (AI coding assistant, UI builder, GitHub, managed database, hosting platform), updated each term.

Dedication hours (SCT-Chile)

Activity Hours
In class
Lectures / workshop 15
Assessment (milestone reviews and Demo Day) 5
Autonomous
Mandatory assignments (milestones M1 to M3) 8
Preparation of presentations (Demo Day) 7
Personal study (pre-readings, build log, reflection) 10
Total (clock hours) 45
Transferable credits 2 SCT-Chile

Course summary

Prepared by: Sebastián Azócar M., Ingeniero Matemático, Magíster en Data Science. Department approval, Graduate Office approval and CPP approval: pending (2026).