Cornell University ILR School Canvas Student Center
Illustrated collage of the ILR School building in red and cream
ILRHR 5635 · Spring 2027

AI and HR

When should managers trust, use, and govern AI to create value without creating harm?

3 credits | Tuesday & Thursday, 75 minutes | 28 sessions | Ives Hall, Room TBD
Course Description

People decisions, made with machines

Guiding Question

When should managers trust, use, and govern AI to create value without creating harm?

Course Sequence

Understand the machine → understand the humans → make it work at work.

HR leaders and managers increasingly shape how organizations adopt and govern artificial intelligence. This course examines how modern AI works, including large language models, retrieval systems, and agents. It explores how these technologies are integrated into real organizational workflows. Topics include AI capabilities and failure modes, human judgment in working with AI, and the implications of AI for job design, teams, decision making, and organizational change.

Through cases, hands-on exercises, and an applied project, students evaluate AI use cases and design AI-enabled workflows for Human Resources, with attention to bias, privacy, compliance, governance, and worker impacts.

Prerequisites
MILR Core
Enrollment
MILR
Grading Option
Letter grades only
Final Deliverable
Final project
Instructor

Dr. Scott Seidenberger

Postdoctoral Associate and Future of Work Fellow
Human Resource Studies, ILR School
Cornell AI Experimenters Faculty Fellow

I am an ILR ’16 alum and came back to Cornell by way of an Air Force jet, a startup exit, and a Ph.D. in data science. My research is on how people understand, direct, and rely on AI at work, and what that does to judgment, opportunity, and the design of jobs.

My recent work examines how language models narrow the set of career options they present to the people relying on them for guidance, how AI portrays employer brands, and how enterprises measure AI adoption and redesign work around it. I run field studies inside HR organizations through Cornell’s CAHRS network.

Office Hours
Day & time TBD · Office TBD
or by appointment
Contact
scotty@cornell.edu
Course site on Canvas
Learning Outcomes

By the end of the semester, students will be able to

01

Explain core AI concepts such as large language models, training, inference, and retrieval-augmented generation at a managerial level.

02

Diagnose common capability limits like hallucination, bias, and privacy to propose mitigations.

03

Assess the human factors that determine whether AI deployments succeed.

04

Analyze organizational implications for work design, labor relations, and HR compliance.

05

Design and experiment with AI pilot planning that implements risk controls and governance.

06

Build a cross-functional change plan and communicate ROI, controls, and policy alignment to both executives and worker representatives.

Course Format

AI-enabled course design

The course meets twice weekly for 75 minutes across 28 sessions, organized into three parts.

Survey course

This is a survey course on rapidly evolving emerging technology, meant to teach fundamental principles and concepts about AI — with a focus on generative AI — as they apply to the field of HR. It is not about teaching prompt engineering or any specific model, tool, system, or commercial application. We cover the impacts of generative AI on about a dozen core HR domains in order to drive further student exploration. Students are encouraged to bring their work experience and skills developed in other courses.

Interactive lectures

Core conceptual content taught by lecture, with class participation in discussions.

Hands-on work with generative AI tools

Direct work with real generative AI systems, including a dedicated agent studio lab.

Business cases

Case studies complement the lectures and serve as lecture reinforcement.

Guest practitioners

Senior HR leaders and industry professionals complement the lecture material and are core to the course.

Guest Speakers

Global Head of HR
Financial data and media platform · 20,000+ employees
AI across the talent lifecycle at enterprise scale
Investor and former operator
Global venture capital firm · $20B+ under management
The judgment premium, from the operator's chair
Principal, People Advisory
Global professional services firm · 300,000+ employees
Building the change plan
VP, People Operations
AI-native insurance company · under 1,000 employees
People operations in an AI-native company
Global Head of People Analytics
Freight and logistics operator · 30,000+ employees
Measuring work to find AI value
Civil rights enforcement attorney
State-level government enforcement office
The enforcement view of algorithmic management
Head of enterprise AI adoption
Enterprise software company · 6,000+ employees
The AI Academy as corporate L&D at scale
Names and affiliations are released as each guest confirms. The schedule is tentative; updates are posted on Canvas.
Course Progression

Three parts

Part I
Understand the Machine
Jan 26 – Feb 25 · 9 sessions · Modules 01–04
Part II
Understand the Humans
Mar 2 – Mar 16 · 5 sessions · Modules 05–06
Part III
Make It Work at Work
Mar 18 – May 11 · 14 sessions · Modules 07–10
Part I Module 01

Framing & Trust

Why trust is the through-line, and what a survey course on a moving technology can promise.

S01 Course Intro: Should You Trust Me? Framing, grading, and the trust question.
Part I Module 02

How the Machine Works

Tokens to transformers to agents: enough of the mechanism to reason about it.

S02 How LLMs Work I Tokens, training, transformers.
S03 How LLMs Work II Inference, context, retrieval → agents.
Part I Module 03

The Stack & the Frontier

What is buildable today, who sells it, and what to build, buy, or wrap.

S04 Are We at AGI? Single-session Socratic debate.
S05 The AI Stack Layers, vendors, build vs. buy vs. wrap.
S06 The Agent TA Meeting Hopper.
Part I Module 04

Failure Modes & Bias

Where systems break, how bias enters, and what enterprise deployment looks like.

February break · Feb 13–16
S07 Failure Modes Hallucination, robustness, verification — taught by stress-testing AI-drafted performance reviews.
S08 Bias Data → models → deployment.
S09 Guest: Global Head of HR, financial data and media platform AI across the talent lifecycle at enterprise scale.
Part II Module 05

Judgment, Trust & Cognition

Trust calibration, automation bias, and what offloading your thinking costs.

S10 Trust, Verifiability, and Taste Trust calibration, automation bias; curation as the new moat.
S11 Cognitive Offloading From Plato to autopilot; cognitive load and healthy AI use at work.
S12 Guest: Investor and former operator, global venture capital firm The judgment premium, from the operator's chair.
Part II Module 06

Manifesting Intent

Intent to outcome: semantic representations, context, harnesses, and taste.

S13 How to Manifest: The Agent Maestro Semantic representations of the world. Words mean things.
S14 How to Manifest II Context, harnesses, and taste.
Part III Module 07

Org Design & Information Flows

Whether AI solves the information-routing problem the org chart was built for.

S15 The Org Chart & Information Flows Does AI solve the information-routing problem?
S16 Guest: Principal, people advisory practice Building the change plan.
Part III Module 08

The Economics of AI Work

Budgeting AI as labor, then finding the work streams worth piloting.

S17 Should We Treat AI Costs as Labor Costs? Budgeting and headcount when AI does the work.
Spring break · Mar 27 – Apr 4
S18 Guest: VP People Operations, AI-native insurance company People operations in an AI-native company.
S19 Identifying Work Streams That Benefit from AI Task analysis and measurement before you pilot anything.
S20 Guest: Global Head of People Analytics, logistics operator Measuring work to find AI value.
Part III Module 09

Jobs, Selection & Employment Relations

Displacement and augmentation, hiring funnels, and who captures the gains.

S21 What's It Going to Do with Jobs? Displacement, augmentation, and the four skills that remain.
S22 AI in Selection & Assessment Signal collapse, automated interviews, auditing the funnel.
S23 AI & Employment Relations Power, progress, and who captures the gains.
S24 Guest: Civil rights enforcement attorney, state government office The enforcement view of algorithmic management.
Part III Module 10

Pilots, Governance & Adoption

Designing pilots with risk controls and evaluation, and taking adoption to scale.

S25 Designing AI Pilots Risk controls, governance, and evaluation design.
S26 Guest: Head of enterprise AI adoption, enterprise software company The AI Academy as corporate L&D at scale.
S27 Designing AI Pilots II Evaluation design, continued.
S28 Final Session Reserved. Final report due in the University exam period, May 15–22.
Module 01 · Framing & Trust
Agentic TA Experiment

A course about deploying AI at work should include piloting AI at work

As part of the Cornell AI Experimenters Faculty Fellows program, this semester will have an AI teaching assistant built specifically for ILRHR 5635. The agent supplements office hours and other course functions. It does not replace the instructor, and it is not the authority on grades or policy.

What you will do

01

Interact with the agent TA over the semester for course questions, readings, assignment coaching, and FAQ.

02

Complete short surveys about your experience at a few points in the term. Each counts toward out-of-class participation.

03

Report failures. Wrong answers, confident nonsense, and dead ends are the most useful data, and diagnosing them is a graded skill.

Data and privacy

Conversation logs and survey responses are collected to improve the agent during the semester and to report on the pilot. Anything reported externally is aggregated and de-identified. No student is named or identifiable, and nothing you say to the agent affects your grade. Do not enter sensitive personal information about yourself or others.

Extra credit for improving the agent

Structured bug reports, evaluation cases the agent gets wrong, prompt or knowledge-base contributions, or a short memo proposing a design change. Full criteria and the submission form are on Canvas. Extra credit is capped, is credited only for substantive contributions, and cannot substitute for missed required work.

Required Materials

Many different sources

All readings are articles, working papers, industry reports, short videos, and instructor-provided materials, distributed through Canvas or Cornell Library, or publicly available at no cost.

AI tool access

Several assignments require working directly with generative AI systems. Students work with the instructor to make sure they have the tool access needed for any assignment core to the course.

Foundations

CS50 for Business for the technology from the ground up, Getting Good at Claude for working with generative AI, plus 3Blue1Brown on large language models and Wolfram on what ChatGPT is doing.

Research & reports

The 2026 AI Index Report from Stanford HAI, The Enterprise AI Playbook from the Stanford Digital Economy Lab, working papers on AI and work, and industry adoption reports.

Cases

Two written cases. First on the HR analytics of an AI workforce transformation. Second on org design in an AI-native firm. 

Assessment & Grading

Half the grade is class citizenship

Being a good citizen of this course is your obligation throughout: show up, come prepared, contribute, and keep the conversation going between sessions. Two absences are automatically given; further absences follow University guidelines. Midterm feedback on participation keeps you on course for full points.

Class citizenship 50%
Attendance10% · ongoing
In-class participation10% · ongoing
Out-of-class participation (online)10% · ongoing
Guest speaker reflections20% · 48h before & after
In-class participation is graded on the quality of contribution, not volume. Online participation happens between sessions on the course site, using Perusall for collaborative annotation of readings. Speaker briefs are due 48 hours ahead so questions can be forwarded to the guest, with a short reflection 48 hours after.
Cases 20%
Case I10%
Case II10%
Course project 30%
Presentation15% · delivered virtually
Report15% · final exam slot
Teams apply the concepts from this course to an AI adoption or deployment challenge faced by an organization. Details are posted on Canvas in the opening weeks. The report is due at the University-scheduled final examination time, in the May 15–22, 2027 examination period.
Course Policies

The ground rules

Use of AI in this course

AI use is permitted and expected on every out-of-class deliverable. Disclosure is required: each submission carries a short process appendix stating which tools were used, the substantive prompts, and what you accepted, what you rejected, and why. The appendix is graded as part of the work, because trust calibration is a learning outcome and the appendix is where it becomes visible. Undisclosed AI use is an academic integrity violation.

The analog classroom

In lectures, discussions, and guest lectures, laptops and phones are put away and notes are taken by hand. Research shows handwriting outperforms typing for retention and forces the compression that makes ideas stick. Summary notes are generated by the instructor and all lecture materials are posted on Canvas.

The Chatham House rule

Guest speakers discuss actual organizational scenarios and decisions, so sessions run under the Chatham House rule: students may use what they hear but may not attribute it to the speaker or their employer outside the classroom. No recording of any kind. Summary notes are provided by the instructor.

Late work

Late submissions lose a letter grade per 24 hours, up to 72 hours, after which work is not accepted. Speaker briefs are the exception and are not accepted late at all, because their purpose is to reach the speaker before class. Extensions are granted for documented circumstances; ask before the deadline rather than after.

Scope of the course

Disagreement about ideas is expected and welcome; disrespect toward people is not. This course is a broad survey of AI as applied to HR, not an AI ethics and society course, so questions of generative AI ethics outside HR functions are out of scope here.

FAQ

Common questions

Do I need a technical background?

No. Core AI concepts — large language models, training, inference, retrieval — are taught at a managerial level. The goal is the foundational understanding and critical thinking to know what to learn next as the technology keeps changing.

How much coding is involved?

None. You will work hands-on with real generative AI systems, but this is not a course on prompt engineering or on any specific model, tool, or commercial application. It teaches the fundamental principles that outlast the current tools.

Can I use AI on the assignments?

Yes — AI use is permitted and expected on every out-of-class deliverable, with a disclosure appendix that is graded as part of the work. Undisclosed use, or submitting output you cannot explain or verify, is an academic integrity violation.

Who can enroll?

Enrollment is set for MILR students, with no prerequisites, and the course is offered for letter grades only. Cross-registration questions go to the instructor.

Will guest speaker sessions be recorded?

No. Guest sessions run under the Chatham House rule with no recording of any kind. The instructor provides summary notes as part of the course materials.

What if I have to miss class?

Because life happens, two absences are automatically given. Further absences follow University guidelines. Attendance credit is for presence in the room under the analog classroom conditions, with personal devices put away.