Georgia State University — J. Mack Robinson College of Business PATH — Pathways for AI Training & Hiring CIS 4394 Agentic AI  ·  Fall 2026  ·  Dr. Xinyu Fu
Week 1 · Foundations

Introduction to Agentic AI, and what enterprises actually use.

Everyone uses AI; few profit from it yet. Agents, systems where the model owns the control flow, are the bet the industry is making to close that gap. This week builds your mental model: the loop, the vocabulary, the autonomy dial, and the 2026 landscape.

In the news · 2026

Five numbers that frame the semester

Click each card for the “so what” and the source. Every figure is dated and linked to its source; this field moves fast, so treat the numbers as snapshots.

TAP
88%
of firms use AI regularly, but only ~39% report EBIT impact
Everyone uses AI; few profit yet. Agents are the industry’s bridge from usage to value, and the paradox this course exists to close.
McKinsey, State of AI in 2025 (2025), approx.
TAP
40%
of enterprise apps expected to ship task-specific agents by end of 2026, up from <5% in 2025
You are early. The skill you build this semester is arriving in the software you’ll use at work.
Gartner (2025), projection, approx.
TAP
4+
live agentic-commerce efforts: Visa + OpenAI, ChatGPT Instant Checkout, Amazon “Buy for Me,” Mastercard Agent Pay
Software can now hold a wallet. What goes right, and what goes wrong, when an agent can transact?
Digital Commerce 360 (2026)
TAP
80–90%
of a real cyber-espionage campaign was automated after a state-linked group jailbroke an agentic coding tool (~30 targets)
Autonomy cuts both ways. Power without guardrails is a liability; this previews our security week.
Anthropic / Cybersecurity Dive (Nov 2025)
TAP
−14%
hiring rate for 22–25-year-olds in AI-exposed jobs, yet ~57% of AI use augments rather than automates
It’s not replacing you; it rewards the people who can use it well. That’s the career case for this course.
Anthropic, Labor-Market Impacts / Economic Index (2026), approx.
The tension

The 2026 split: exploding capability, stubborn unreliability

Advancing fast

Task length doubles ~every 88 days

The length of tasks agents can complete autonomously has been doubling roughly every three months (METR, approx.). Demos get dramatically better every semester.

Still unreliable

~20% vs 72%

On long computer-use tasks, top agents score around 20% where humans score ~72% (OSWorld, approx.). Benchmarks lag the demos; reliability is the semester’s recurring villain.

Both of these are true at once. Holding that tension between capability and unreliability is the whole discipline of agentic AI.
See it move · the same request, three ways

Who owns control flow?

One request, “find me a time with Priya next week”, handled by a chatbot, by a workflow, and by an agent. Press Play on each tab and watch where the decisions happen. The difference between the three is not intelligence; it is who decides what happens next.

Try this: run all three, then answer in one sentence each: who chose to call the calendar? who decided to stop? which run would you let act unattended, and why?

This week

Work through the four parts

Objectives

By the end of Week 1 you can…

← BackCourse home