Insegnamento a.a. 2026-2027

21119 - AGENTIC AI FOR DATA-DRIVEN INSIGHTS

Department of Marketing


Course taught in English
Go to class group/s: 31
ACME (6 credits - II sem. - OP  |  SECS-P/08) - AFM (6 credits - II sem. - OP  |  SECS-P/08) - AI (6 credits - II sem. - OP  |  SECS-P/08) - CLMG (6 credits - II sem. - OP  |  SECS-P/08) - DSBA (6 credits - II sem. - OP  |  SECS-P/08) - EMIT (6 credits - II sem. - OP  |  SECS-P/08) - ESS (6 credits - II sem. - OP  |  ECON-07/A  |  SECS-P/08) - FIN (6 credits - II sem. - OP  |  SECS-P/08) - GIO (6 credits - II sem. - OP  |  SECS-P/08) - IM (6 credits - II sem. - OP  |  SECS-P/08) - MM (6 credits - II sem. - OP  |  SECS-P/08) - PPA (6 credits - II sem. - OP  |  SECS-P/08)
Course Director:
KAI ZHU

Classes: 31 (II sem.)
Instructors:
Class 31: KAI ZHU


Mission & Content Summary

MISSION

Agentic AI systems — language models that can plan, use tools, write and execute code, query APIs, and iterate on their own outputs — are rapidly becoming a central layer of the data-driven decision-making stack. For marketing analysts, strategists, and researchers, the consequence is profound: a single professional can now design workflows that autonomously assemble data from heterogeneous digital sources, analyse it at scale, and produce decision-ready outputs, provided they know how to direct and validate such systems. This shift redefines what constitutes a marketable analytical skillset, and it is moving faster than traditional curricula can absorb. This course addresses that gap by treating agentic AI not as a black box or a marketing buzzword, but as a methodology that must be designed, instrumented, and critically evaluated. Within the Master programme in Marketing, the course complements existing courses on marketing analytics, consumer behaviour, and digital marketing by giving students hands-on competence in building end-to-end AI-augmented pipelines — from data acquisition to interpretation — and by training the methodological judgement required to defend the resulting insights to interdisciplinary business and policy audiences.

CONTENT SUMMARY

The course is organised into three connected macro-topics that mirror the agentic AI lifecycle: configuring the environment, building autonomous data pipelines, and packaging reusable analytical capabilities.

 

1. Foundations of agentic AI workflows. The capabilities and limits of current agentic systems; setting up a professional AI workspace (Claude Code, VS Code, Markdown); the Read–Think–Act workflow for delegating analytical tasks; verification habits and the explore–plan–code discipline; context management (how context windows work, when and how to compact); persistent project configuration through structured instruction files.

 

2. Autonomous data acquisition and analysis pipelines. Building pipelines that ingest and clean structured data from APIs, marketplaces, and platform exports; aggregation, comparison, and visualisation of large-scale platform metrics; AI-powered analysis of unstructured text including automated labelling, sentiment detection, and thematic extraction; large-scale interpretation of user-generated content; agentic web research and source-grounded fact-checking; competitive scans across heterogeneous web sources. 

 

3. Reusable agentic capabilities and end-to-end integration. Discovery, evaluation, and adaptation of community-built AI Skills; designing custom Skills that encode domain-specific analytical workflows; composing multiple Skills into end-to-end pipelines that integrate acquisition, analysis, iteration, and interpretation; methodological trade-offs between traditional analytics and AI-augmented systems (validity, bias, robustness); ethical use, validation discipline, and translation of technical outputs into actionable business and policy recommendations. 


Intended Learning Outcomes (ILO)

KNOWLEDGE AND UNDERSTANDING

At the end of the course student will be able to...

At the end of the course, students will be able to:

 

- Describe the architecture and operational logic of modern agentic AI systems, including their tool-use, planning, and iteration capabilities.
- Identify the failure modes of agentic AI — hallucination, brittle planning, context-window limits, silent errors in autonomous loops — and the verification practices that mitigate them.
- Distinguish the analytical roles best suited to AI augmentation from those that still require human judgement or traditional statistical methods.
- Explain how reusable agentic capabilities (Skills, structured project instructions, multi-tool pipelines) are designed and orchestrated for autonomous workflows.
- Outline the methodological criteria — validity, bias, robustness, reproducibility — used to evaluate AI-generated outputs in a marketing or policy context.
 

APPLYING KNOWLEDGE AND UNDERSTANDING

At the end of the course student will be able to...

At the end of the course, students will be able to:

 

- Design agentic AI workflows that autonomously collect, process, and analyse large-scale digital data from APIs, social media platforms, online marketplaces, and web sources.
- Build end-to-end computational pipelines that integrate multiple AI tools for data acquisition, analysis, iteration, and interpretation.
- Develop custom Skills that encode reusable, domain-specific analytical procedures and can be shared with collaborators.
- Compare AI-augmented analytical approaches with traditional methods and justify the chosen approach on methodological grounds.

- Apply state-of-the-art AI techniques to extract structured insights from quantitative datasets and unstructured text (automated labelling, sentiment, thematic analysis).


Teaching methods

  • Lectures
  • Individual works / Assignments
  • Collaborative Works / Assignments

DETAILS


☑ Face-to-face lectures


☑ Practical exercises / Lab work

 

☑ Case studies

 

☑ Individual assignments

 

☑ Interactive class activities
 


Assessment methods

  Continuous assessment Partial exams General exam
  • Written individual exam (traditional/online)
     x
  • Individual Works/ Assignment (report, exercise, presentation, project work etc.)
 x    
  • Collaborative Works / Assignment (report, exercise, presentation, project work etc.)
 x    

ATTENDING STUDENTS

  • Participation: 30%

    • Engagement in class, completion of in-class exercises.

  • Assignments and Project: 30%

    • A set of individual or group assignments throughout the course.

  • Final Exam: 40%

    • Written exam on both conceptual knowledge and applied skills developed in the course.

 

(Students need to more than 75% all sessions in person to be qualifed as attending students)


NOT ATTENDING STUDENTS

  • Assignments and Project: 50%
    Final individual project

  • Final Exam: 50%
    Written exam covering both conceptual knowledge and applied skills developed in the course.


Teaching materials


ATTENDING AND NOT ATTENDING STUDENTS

  • Lecture slides and Jupyter notebooks prepared by the instructor.

  • Selected academic papers, reports, and practitioner articles on AI and market research.

  • Links and documentation for LLM / AI tools and APIs used in class.

Last change 07/07/2026 09:03