What Teaching Looks Like When AI Is Ambient

The governance work is done, or at least begun. Now the harder conversation: what happens to teaching itself when the tool is everywhere and no longer optional to think about.

A professor I would describe as thoughtful, not resistant, put it this way: “I redesigned my course around discussion because the essays stopped telling me anything. Then I watched a student consult a chatbot, in class, to prepare for the discussion. I am not angry. I just no longer know where the student ends and the tool begins.”

That sentence is the real state of teaching right now, and it is worth taking seriously precisely because it is not a complaint about cheating. It is a question about what the work of learning consists of when an always-available assistant sits inside every assignment, every draft, every study session. For nine issues this series has been in the governance wing of the institution. This one walks into the classroom, where the questions get harder and the stakes get closer to the mission.

A quick frame for this part of the series, because it deserves one. The Educate group of the map, the academic mission, asks how the academic core changes: teaching, learning, assessment, and scholarship, which is to say the reason the institution exists. The temptation is to reduce all of it to the integrity conversation, the one that starts and ends with cheating. That conversation matters, and it gets its own issue next. But it is the doorway, not the room. The room is bigger: what instructors are for, what students should learn, and what a course is, when AI is ambient in all of it.

The institutional question

What is the instructor’s role when every student has an always-on academic assistant?

Not in the panicked version of that question, where the answer is supposed to be “obsolete.” In the practical version. A student can now get an explanation, a worked example, a critique of their draft, and a study plan at 2 a.m. without the institution being involved at all. Some of that is what we used to call teaching. So what is the part that was never really about delivering explanations, and are we organized to do more of it?

What this looks like in practice

Start with the role shift, because everything else follows from it. When explanation is abundant, the instructor’s scarce contributions become the things AI does poorly: knowing this student, in this course, at this institution. Judging whether understanding is real. Designing the sequence of struggle a novice actually needs, because a chatbot will helpfully remove the struggle if asked, and the struggle was the point. The shorthand I keep coming back to: AI made content delivery cheap, which makes the genuinely human parts of teaching more valuable, not less. That is an encouraging sentence that implies an enormous amount of redesign work, and both halves are true.

Then AI literacy, which institutions keep treating as one thing when it is at least three. There is operating fluency, knowing how to use the tools well. There is critical fluency, knowing when the tools are wrong, biased, or confidently making things up, which is the part employers quietly value most. And there is disciplinary fluency, understanding what AI means inside the student’s field, because what it means in nursing, in accounting, in graphic design, and in philosophy are four different answers. A single required “AI module” bolted onto the general education core does not get there. The institutions doing this well are asking each program to answer a specific question: what does a graduate of this program need to be able to do with, and against, AI?

That phrase “each program” is doing real work, because the discipline differences are not a footnote. In computing courses the tool writes the code, and the faculty argument is about what assessment even measures now. In clinical programs the hands still have to know what to do, and AI mostly threatens the paperwork around the learning rather than the learning. In writing-centered disciplines the tool sits directly on top of the thing being taught, which is why the humanities are having the most existential version of this conversation. An institutional AI teaching policy that ignores those differences will be simultaneously too strict for some programs and too loose for others. The center can set the floor. The disciplines have to build the rooms.

Course design is where the shift becomes concrete, and it is also where AI enters as help rather than threat. Faculty are already using it to draft practice problems, generate examples, build rubrics, and produce the third version of the syllabus that the semester always demands. That is real capacity returned to people who had none to spare. It comes with one caveat the enthusiasm tends to skip: AI-generated course material still needs a subject-matter expert to review it, because it is fluent in exactly the way that makes errors hard to spot. The time saved drafting is partly reinvested in checking. The trade is still good. It is just not free.

Modality shapes all of this more than most policy conversations admit. In a face-to-face seminar, the instructor can see the learning happen. In an asynchronous online course, nearly every artifact a student submits can be produced without the student learning anything, which means online programs are not facing a variant of the assessment problem. They are facing its purest form. Institutions with large online enrollments need to be honest that their exposure here is structural, not marginal, and that the redesign conversation is correspondingly more urgent. That thread runs straight into the next two issues.

And then the people question, which the tool conversation keeps stepping over. Some faculty are experimenting enthusiastically. Many are exhausted, skeptical, or quietly afraid that this is one more transformation they are expected to absorb without time, support, or a say. Both groups are rational. The worst institutional response is to split the difference with a memo. The better response is structural: time that is actually protected for redesign, examples from their own discipline rather than a generic workshop, and an explicit institutional statement that experimenting and declining to experiment are both, for now, professionally safe choices. Faculty who feel cornered do not innovate. They comply, minimally, and wait for the initiative to pass. Anyone who has been in higher ed longer than one strategic plan knows that posture well.

The Atlas connection

In Atlas, our AI operating map, this is the teaching and learning domain, and it opens the Mission group — Educate — deliberately. Assessment and integrity, coming next, are downstream of the pedagogical choices this domain describes: you cannot decide how to measure learning until you have decided what learning consists of now. Faculty development, later in this part of the series, is how any of this becomes real rather than aspirational. And the governance stretch of the map just concluded — Decide and Protect — is the quiet prerequisite, because a campus without decision rights and data rules will find its teaching conversation repeatedly derailed by exactly the fights the first nine issues were about preventing.

Questions worth putting on the agenda

Better with provosts, deans, and faculty leaders in the same room, since none of these belong to one office.

• What is our answer, program by program, to “what should a graduate be able to do with, and against, AI”?
• Where in our curriculum do students still get productive struggle, now that a tool will remove it on request?
• Are our AI teaching expectations set at the right level, an institutional floor with disciplinary rooms, or one rule pretending to fit everyone?
• What protected time and discipline-specific support have we actually funded for course redesign, as opposed to announced?
• For our online programs specifically: which assessments still measure learning, and which now measure tool use?

The bottom line

The professor who no longer knows where the student ends and the tool begins is not describing a discipline problem. He is describing the new baseline condition of teaching, and the institutions that thrive will be the ones that redesign for it on purpose rather than police it into a corner.

AI made explaining cheap. It made teaching, the human part, more valuable than ever. The work now is building courses, and institutions, that act like both of those are true.