A master’s in AI, rewritten for the age of AI agents.
The MSAI curriculum was designed for a field that has since changed underneath it. This working draft pairs all 18 Concordia University Wisconsin graduate Computer Science courses with a proposed revision for agentic practice, gives each a fresh set of course learning outcomes, and maps every outcome to the MSAI program learning outcomes and, through them, to Concordia's six university learning outcomes.
Teaching AI is not the same as teaching AI the way it is now built.
Most of these courses already are AI courses, so the question is not whether AI belongs in them but what agentic practice changes inside them: models are now assembled into tool-using systems, evaluated with harnesses rather than intuition, and operated under cost, latency and safety budgets. The graduate catalog is also shared with the MSCS program, so eight of these courses serve systems students; those keep their own disciplinary identity and change only where practice has genuinely moved.
What the revisions add
- Agents, tool-calling and retrieval as the default system shape
- Evaluation harnesses, LLM-as-judge and measured failure modes
- Cost, latency and energy treated as design constraints
- Prompt injection, tool supply chain and model provenance as security concerns
- Post-deployment monitoring, drift and accountability for autonomous behaviour
How AI weight is distributed
In a graduate AI program most courses carry AI as content. The interesting variation is in the courses shared with the MSCS program, where AI arrives as practice rather than subject.
Six university outcomes, twelve program outcomes
Concordia University Wisconsin states six university learning outcomes for every graduate. The MSAI proposal renders each as two program learning outcomes — twelve in all.
Christian Faith
“Our graduates are grounded in the Christian faith while also recognizing other major worldviews and how they differ from a Christian understanding of the world.”
Service and Global Citizenship
“Our graduates are globally-minded citizens.”
Integrated Disciplinary Knowledge
“Our graduates integrate insights from a wide range of disciplines.”
Critical Thinking / Creative Problem Solving
“Our graduates think rationally, critically, and creatively.”
Communicative Fluency
“Our graduates communicate effectively.”
Analytical Fluency
“Our graduates work with data effectively.”
Which outcomes the revised curriculum actually reaches
Each bar counts the courses whose proposed outcomes provide evidence for that program learning outcome. An outcome carried by many courses is well served; one carried by few is a question for the department, not necessarily a defect.
Table view
| PLO | University outcome | Courses | CLOs |
|---|---|---|---|
| PLO 1.1 Ethical and Legal Integrity in AI | ULO1 Christian Faith | 18 | 24 |
| PLO 1.2 Vocation and Christian Worldview | ULO1 Christian Faith | 6 | 7 |
| PLO 2.1 Data Acquisition for Real Problems | ULO2 Service and Global Citizenship | 12 | 16 |
| PLO 2.2 Domain-Specific AI Agents | ULO2 Service and Global Citizenship | 6 | 9 |
| PLO 3.1 AI Theory and Mathematics | ULO3 Integrated Disciplinary Knowledge | 7 | 12 |
| PLO 3.2 Ethical and Legal Integrity in AI | ULO3 Integrated Disciplinary Knowledge | 18 | 24 |
| PLO 4.1 Learning-Based Systems and Agents | ULO4 Critical Thinking / Creative Problem Solving | 5 | 6 |
| PLO 4.2 Prompt Engineering for LLMs | ULO4 Critical Thinking / Creative Problem Solving | 7 | 8 |
| PLO 5.1 Communicating AI to Any Audience | ULO5 Communicative Fluency | 17 | 20 |
| PLO 5.2 Defining and Evaluating AI | ULO5 Communicative Fluency | 5 | 5 |
| PLO 6.1 LLMs in Industrial Settings | ULO6 Analytical Fluency | 12 | 18 |
| PLO 6.2 Model Evaluation and Refinement | ULO6 Analytical Fluency | 15 | 18 |
Start anywhere
Every course page shows the current catalog description, the proposed revision, what changed and why, and the proposed course learning outcomes with their mapping.
AI Ethics and Vocation
Agentic AI shifts the ethical question from what a model outputs to what a system does on someone's behalf, so a practitioner's accountability now extends to delegated action, machine-produced deliverables, and the data and labor supply chain behind a model. The course keeps its vocation-grounded, case-study identity and adds delegated agency - escalation, guardrails, audit, provenance, and disclosure - as the setting in which that moral reasoning is practiced.
Applied Artificial Intelligence
Applied AI is no longer a survey of separate perception techniques: practitioners now assemble vision, speech, and language capability out of pretrained and hosted models, wire them together with tool-calling agents and retrieval, and are judged on evaluation, cost, and accountability rather than on hand-built pipelines. The course therefore moves from demonstrating techniques to specifying, orchestrating, evaluating, and taking responsibility for working AI systems.
Data Security and Information Assurance
Enterprise defense now has to cover systems that are themselves built and operated by AI: prompt injection, tool and model supply-chain compromise, over-scoped agent credentials, and machine-generated code and runbooks reaching production are ordinary incident causes, while detection authoring, log correlation, and triage are increasingly AI-assisted. The course therefore treats the model, retrieval, and agent-tool layer as a first-class attack surface and holds students accountable for evaluating and bounding machine-produced security work.
Mobile Computer Architecture
On-device inference has made neural accelerator provisioning, model quantization, memory bandwidth, and energy per inference first-order mobile architecture constraints rather than specialist concerns, and AI coding agents now produce much of the profiling and kernel-level optimization work architects once wrote by hand. The course therefore budgets silicon for inference workloads and requires agent-produced analysis to be verified against measured device telemetry before it informs a design decision.
Applied Computer Networking
Network engineering and operations are now largely machine-assisted: telemetry is triaged by learned detectors, configuration and remediation are drafted by AI agents, and the enterprise network has become the delivery and control path for AI services. The course keeps its networking identity and adds the evaluation, validation, and accountability skills a practitioner needs to supervise that automation, including the rollout and rollback discipline a machine-proposed change must clear before production.
Database Administration
Database administration has become a control point for AI systems: production stores now hold the embeddings and governed records that retrieval-augmented applications and model training depend on, while schemas, queries, and migrations are increasingly drafted by AI coding agents and executed by automated service accounts. The course therefore adds vector and hybrid indexing, benchmarking of AI-assisted and learned query optimization, provenance for training data, and least-privilege access and auditing for agent traffic, without displacing core relational design and operations.