JFSC AI Initiative

What this was

The JFSC AI Initiative was a two-year effort to bring AI into the Joint Forces Staff College, taught as a practice rather than handed down as a policy. It ran on two tracks at the same time.

The first track was education and adoption. Before anyone could use AI well, they had to understand what it was. This track produced the college's first AI literacy course, GenAI 101, along with faculty workshops, scenario-based labs that gave faculty hands-on practice, and steady one-on-one help as people worked through their own questions. The aim was to move people away from treating AI as either magic or a threat and toward seeing it as a tool with real strengths and real limits that they could learn to handle responsibly.

The second track was the tools. Alongside the education work, I designed and built custom AI tools for specific teaching problems. These were not general-purpose GPTs with a brief. They went after particular friction points in the classroom, like reflective writing that came back flat or case studies that asked nothing of the student.

Each tool did one job. The Reflections GPTs pushed student writing to be sharper and more analytical. The Spruance and Desert Storm case studies made students commit to a position before they saw how history turned out. The Rubric-Guided Grading Assistant backed up faculty judgment instead of replacing it. Those are a few of them. The full set grew past 30 custom tools over the two years and reached nearly every subject in the curriculum.

Why the skepticism helped

JFSC does not adopt new things easily, and that is on purpose. Senior officers are trained to question sources, challenge assumptions, and push back on anything pitched before the problem is clear. That made AI adoption harder and also more honest. Anything that survived had to prove it was worth using, and any claim I made about what AI could do had to hold up, because an inflated one would get caught fast. Some of the work landed, measurably, and it did so by earning its way in rather than being required.

Testing it in my own seminar

The clearest evidence came from my own seminar, which I ran as a live test bed across the full 40-week curriculum. I kept it out in the open, and the students knew exactly what we were doing and why. The rules were simple. They could experiment with AI and push on how far it could go in planning and analysis, as long as it never came at the expense of learning. I wanted them to challenge how things had always been done and to be honest about what the tools really produced.

The pattern held. Used with care in the planning process, AI made the work faster, and the planning products got better, subjectively but consistently. Students valued it. What came out of that year was more than a proof of concept. It was a repeatable way to bring AI into a serious professional setting without forcing it on anyone or overselling what it could do.

The numbers

Several tools logged 100 or more sessions, which counts for something in a setting where courses run in fixed cohorts. The Reflections GPTs produced student writing four to ten times longer than earlier cohorts, and noticeably deeper, because the Socratic format made students engage instead of padding. The faculty labs reached roughly 75% of JFSC faculty over the life of the initiative. These are not enterprise-scale numbers, since JFSC is a small institution, but they are real, and they came from people choosing to use the tools rather than being told to.

Why it carries over

Most of this transfers. Almost none of it is specific to a military college, and the same problems show up in any larger organization trying to adopt AI. You build trust before you mandate anything, and you watch what is changing rather than what looks like adoption. The seminar test bed is the most portable piece of all. A 40-week experiment in AI-augmented planning, run in the open and honest about its limits, is the same approach that works in industry. The setting changes and the playbook does not.