Your courses were designed for a world without AI in it. Which ones still work?
Most colleges and universities can answer that for a handful of courses somebody happens to feel strongly about. Almost none can answer it for the catalog. The dean or the provost sponsors the work. Faculty lead it. Nobody’s curriculum authority moves.
Course level, not institution level
Scored from your syllabi, not just the catalog
Faculty keep curriculum authority
Your records and decisions, exportable any time
Curriculum decisions are being made one course at a time, by whoever cares most.
A faculty member redesigns an assignment because a student handed in something a model clearly wrote. A department adds an AI policy. A dean asks for a plan. Each of those is reasonable. Together they are not a strategy, and nobody can say which courses are most exposed or which should move first.
Ask what fraction of your catalog anyone has actually assessed for this. The number is usually close to none.
Meanwhile the assumption underneath a degree — that the capability it builds stays scarce long enough to be worth the tuition — is exactly the assumption AI is testing. That is a strategic question about the institution’s position, not an instructional-technology question, and it does not get answered by adding a chatbot to the library.
The dean or the provost sponsors. Faculty lead. Nobody’s curriculum authority moves.
Curriculum modernization run at faculty rather than with them reliably produces resistance, and it deserves to. Faculty senates hold authority over curriculum, subject matter and methods of instruction, and that is not a procedural obstacle to route around — it is the reason a change that does get adopted actually holds.
So the roles are separated on purpose. Someone sponsors — cover, resources, and the urgency that makes the work real — without mandating a single course change. Across an institution that is the provost; inside one college it is the dean, and the job is identical either way. Faculty decide what happens to any course. Sagebridge supplies the evidence and runs the process, and has no opinion that a committee cannot overrule.
Practically: the audit arrives as a faculty intelligence tool, not an administrative directive. Deans and chairs are briefed college by college. Departments opt in. What comes back is a scored view faculty can argue with, which is the point — an argument about evidence is a better argument than one about whether the administration should be doing this at all.
Inside one college the unit is smaller and nothing else changes: that college’s curriculum committee decides, and its faculty hold exactly the authority the senate holds university-wide.
Seven questions, asked of every course in the same words.
Four ask where AI puts pressure on the course. Three ask where the course builds what AI does not replace. They are scored separately and never netted against each other, because a course can be high on both.
The four that combine into a displacement risk score.
Task Automation Exposure
Can AI perform, speed up or simulate the tasks students repeatedly do in this course?
Content Commoditization
Is the course’s value mainly in explaining information that AI now explains on demand?
Assessment Vulnerability
Could a student complete the assessed work with undisclosed AI help?
Labor Market Displacement
How exposed are the occupations this major leads to?
The three that combine into a relevance opportunity score.
Human Capability Durability
Does the course build judgment, ethics, collaboration or hands-on capability — and assess it?
AI Augmentation Potential
Could well-designed AI use improve learning, feedback or practice here?
Experiential and Applied Intensity
Do students do consequential work under real conditions — labs, field work, placements, clients?
Each is scored 1 to 5 against fixed written descriptions of every level, rolled up to department and college. The two composites are shown side by side, never as a single number: the useful finding is usually a course that is high on both.
Read a finished audit before you commission your own.
Most of what a vendor says about a scoring instrument cannot be checked until you have bought it. This one can be read in full, right now, for a university we invented so that no real institution’s courses are on display.
Every score shows its evidence, or says it is a guess
A score either rests on something the source actually says, or it is marked inferred. At catalog tier, assessment is always inferred — catalogs do not describe how students are graded — and the report says so on every affected course rather than in a footnote.
Thin evidence produces no course-level result at all
Below moderate confidence a course still counts toward department and college figures, but its own scores are withheld. A curriculum committee is right to refuse a number it cannot interrogate, so the instrument does not offer one.
It stops at review, redesign and investment
The audit never recommends closing a course or a program, and it is not an instrument for evaluating faculty — which we ask you to put in writing before any faculty member sees a score.
It starts on your public catalog, so the first read needs nothing from your registrar and nobody has to approve it first. What it can see improves as your syllabi arrive, department by department — and the report is explicit about which findings rest on which.
You can run it yourself, or we can run it with you.
Same instrument either way. What changes is who does the work of getting evidence out of departments and faculty into the room — which is the part that decides whether an audit becomes a decision.
Your people in the seats
The audit runs in a hosted service on your own schedule. Your administrator manages the roster and the evidence. Your faculty see the scores and the evidence behind each one, and can correct, dispute or accept. Your dean or provost confirms what becomes an institutional finding. Sagebridge operates the service and stays out of the decisions. Sized to one department or college to begin with.
Guided First Cycle
For a whole college, where the constraint is not the scoring. It is getting a thousand syllabi out of departments and faculty into the room. Six months, phased department by department, with the service included and Sagebridge doing the chasing and running the faculty sessions. At the end the guidance steps down and the college carries the next cycle itself.
Which one fits turns on how much of your syllabus evidence is already held centrally, and whether you have someone who can chase the rest. That is most of what a first conversation is for.
One college is the right size to start, and usually the faster one.
The work is scoped to a college for a practical reason. Good scoring runs on syllabi, and a dean can get them — a request to chairs you already talk to, department by department over a semester. The same request made across every college of a university is itself a project, and it is the reason institution-wide efforts stall before the scoring gets good.
A college also has what a whole catalog does not: one budget holder, one curriculum committee, and chairs who know each other. That is a unit where evidence can actually be collected and faculty can actually be convened. No cabinet decision has to be arranged first.
One thing worth doing early. Tell the provost’s office the work is happening. Not permission — visibility. Shared governance runs on nobody being surprised, and a college-level engagement the provost’s office hears about second-hand starts on the wrong footing for reasons that have nothing to do with the work itself.
It also tends to be how a wider conversation starts. A college that has done this has something a proposal does not: findings from courses the rest of the institution recognizes.
The catalog is where this starts. It is not where it should end.
Course descriptions are frequently years out of date, written for accreditation and recruitment rather than to describe what is actually taught, and pitched at a generality that hides the very thing AI displaces. The syllabus has the assignments, the assessment scheme, the tools students are told to use, and the week the topic actually appears.
That difference is not cosmetic. Of the seven dimensions scored, one is decisive:
From the catalog, assessment vulnerability is inferred
We are reading a paragraph of description and estimating what the assignments probably look like. Useful, and honest about being an estimate.
From the syllabus, it is read off the page
The assessment scheme is right there, with its weightings. Whether a course can be completed by a model without the student learning anything stops being a guess.
Every score reports which of the two it came from, and every audit reports the mix. You will always be able to tell which findings rest on evidence and which on inference.
Ten years of this was spent inside a university, not consulting to one.
Daren co-founded the Nittany AI Challenge at Penn State in 2017 and built it into the Nittany AI Alliance, which he ran as Executive Director of Innovation until March 2026, with IBM, Google and AWS among its partners.
Which means the parts that are specific to your world are not new to him: shared governance, a faculty senate that can and should say no, deans with eleven different views, the gap between what a provost can announce and what a department will adopt, and the fact that a chief academic officer turns over faster than any other senior role on campus while a curriculum change takes longer than one of them lasts.
Let’s run it on your catalog.
Name the college or department you would start with, and what you are trying to decide. The first read is built from your public catalog, so it needs nothing from your registrar and commits your institution to nothing.