Plenty of students are being told to “learn AI” without being told which hour of which week will count. Handshake’s class of 2026 graduation report, and Inside Higher Ed’s 30 April 2026 account of that survey, put numbers on the gap. The numbers are a course-planning problem. They are not a personality test.
In a survey of 1,248 seniors, 58 percent said they needed a deeper understanding of AI. Twenty-eight percent said AI was a meaningful part of their coursework. Nearly a third of technology postings mentioned AI. The share was more than 7 percent in financial services, and about 3 percent in government, health care, and education postings. More than 10 percent of internships on Handshake mentioned AI skills. Read the industry split before you assume every job ad has become a machine-learning exam. A health-care posting and a software posting are not living in the same sentence.
The 58 and the 28 are the pair to plan from. A majority of these seniors said they needed more than they had. A minority said their classes had made AI a real part of the work. The difference is a curriculum gap you can close with one course, one project, or one job, if you pick the option that leaves evidence.
What the posting share asks you to show
A posting that mentions AI is often asking whether you can use a tool inside a task the employer already has. It is less often asking for a graduate thesis. The Handshake figures do not list the verbs. You still have to read the posting. “Familiarity with AI” beside a marketing role is a different demand from “train a model” beside a research role. Write the verb down. Then pick a course or a project that makes you perform that verb on a dataset, a text, a design, or a process you can explain.
@GergelyOrosz shared a hiring lead’s complaint on 9 October 2026: new graduates were not familiar with AI coding tools, even when they carried an “AI-native” label. Treat that as one person’s report from interviews. The Handshake survey is the rate. The anecdote matches the gap between a label and a skill someone can watch you use. If your only proof is a line that says “AI,” you are the student that complaint describes.
A course counts when it produces an artifact. A notebook with a short analysis, a feature you built with a coding tool and can modify without the tool, a written brief that shows what you asked a model and what you checked by hand. A completion certificate with no file is closer to the 58 percent who feel the need than to the 28 percent who met the skill in class. You want to be able to open the file in an interview and say what you changed.
How to choose the one course
Look at the postings you have pinned. Skip a ranking of “best AI classes.” If your targets are in government, health care, or education, about 3 percent of those postings mentioned AI in the Handshake account. You may need one literate project. A new major is a larger step than that share requires. If your targets are in technology, nearly a third of postings mentioned it, and the project should sit near the center of the portfolio. If your targets are in financial services, the share was above 7 percent. Scale the time to the share, then stop. A second course that repeats the same tool is how a semester disappears.
Ask three questions before you register.
Will you leave with a file you can put in front of a stranger? If the syllabus is only exams about vocabulary, it will not close the gap the survey describes.
Can you explain the output without the tool in the room? Interviewers ask this. A course that forbids you from looking at what the tool did is a poor rehearsal. A course that never makes you check the tool is also a poor rehearsal.
Does the department already require something you could bend toward this? A methods class, a studio, or a writing class can hold the artifact. You do not need a course with AI in the title if the work is real. You do need the work.
If nothing on the timetable qualifies, build the artifact beside a required course and keep the posting’s verb in the filename. The survey did not say the learning has to be branded. It said most seniors did not meet the skill in the courses they already had.
A small loop for the term
Week one: choose the posting and circle the AI-related line, if there is one. Week two: choose the course or the project. Weeks three through ten: do the work the syllabus or the brief asks for, and save the intermediate files. The last week: write five sentences on what you asked the tool to do, what was wrong, and what you fixed. Those five sentences are the interview. The file is the proof.
Do not add a list of model names to a resume because the survey made you nervous. The 28 percent figure is about coursework that included the skill in a meaningful way. A meaningful way is something you can walk through. A keyword is something a screen can see and a person can dismiss in a minute.
If your field’s postings rarely mention AI, do the small version anyway if you have time, and do not let it crowd out the skill those postings do name. A 3 percent mention rate is a reason for proportion. It is not a reason for contempt, and it is not a reason to rebuild a major.
How Flightpath helps
Pin the role first. Flightpath lists the gaps against facts you have confirmed, so an AI skill the posting actually names can show up as a gap, and a skill the posting does not name does not take over the plan. The portal plan by year is where you put the course or the project, in the term you can finish it. When the artifact is real, you confirm it. A line stays off the resume until you do.
A tailored resume for that posting is built from the confirmed facts, and you review the changes. The weekly check-in is the reminder to look at the gap again. You approve anything before it is sent. Handshake’s percentages stay in the class of 2026 report. The plan holds one course and one file.