If you are choosing where AI role play enters your institution, start here. Career readiness has everything a beachhead needs: enormous unmet volume (every student needs interview practice; almost none get enough), low pedagogical controversy (nobody argues students should practice interviews less), immediate student pull (they are already anxious about it), and outcomes the institution is publicly measured on (placement rates).
The volume problem, concretely
A typical career center offers each student one mock interview per year, booked weeks out, with a counselor who has thirty minutes and no context on the target role. Meanwhile the student faces six to ten real interviews in a recruiting season, each with a different format: behavioral, case, technical screen, panel. The first real interviews get burned as practice, exactly the failure mode from Chapter 1: learning on attempts that cannot be replayed, with companies that do not call back.
The NIU guide’s example list starts with exactly this: “In preparation for career interviews, students can assume the role of the interviewer and/or the interviewee.” The pedagogy was always right. What was missing was a counterpart who could do it two hundred times per cohort, per week, calibrated to each student’s target industry.
Drill moments, not marathons
The strongest design lesson from sales training, where AI role play matured first, transfers directly: do not simulate the whole interview end to end. Isolate the make-or-break moments and drill them.
A basketball player working on free throws does not scrimmage full games; they stand at the line and shoot two hundred. The interview equivalents:
- The first two minutes. “Tell me about yourself” decides the interviewer’s frame for everything after. It deserves its own drill, repeated until the student can deliver it three ways for three audiences.
- The weakness question and its cousins. Behavioral deflections under pressure.
- The salary conversation. The single highest-ROI drill in the catalog: five minutes of practice moves real money, and almost no student has ever done it once before doing it for keeps.
- The “walk me through your project” probe. Two follow-ups deep, where padded resumes collapse.
- Case openings for consulting-track students; system-design framing for engineers.
Each becomes a five-to-ten-minute scenario a student can run on repeat. Full-length mock interviews still matter, but as certification: the thing you do once the drills are landing, to prove the pieces integrate. The habit-building volume lives in the drills.
The interviewer must push back
An interviewer persona that nods and moves on teaches interview theater, not interviewing. The persona needs the resistance mechanics from Chapter 11: probe two or three layers deep (“you said you led the project — what did you specifically decide?”), hold skepticism until the student earns the concession, and occasionally throw the curveball that real interviewers throw: the time squeeze, the interruption, the “we’ve talked to candidates with more experience; why you?”
Difficulty should be a ladder. The same recruiter persona at beginner level asks and accepts; at intermediate it probes twice; at advanced it is having a bad day, running late, and unconvinced. Students self-select upward as confidence builds, which is precisely the progression a human counselor cannot offer at volume.
From service to curriculum
The more interesting institutional move is happening in courses, not career centers. Business schools already run sales simulations, negotiation exercises, and case-competition prep as graded coursework; several now configure AI scenarios to the curriculum and grade on rubric scores, with unlimited practice attempts and the best attempt submitted. The University of Houston’s sales program reported a 24% improvement in SPIN selling scores across a semester of AI simulation practice; a single university semester on one platform logged over 21,000 simulations. Those numbers should recalibrate what “practice volume” means: not one mock interview per student per year, but hundreds of attempts per student per course.
The grading pattern that works: practice is unlimited and private; the assignment is “submit a session that scores 80+ on the rubric, plus a half-page reflection on what changed between your first and best attempts.” The reflection clause matters; it is the Chapter 13 debrief showing up as assignment design, and it converts a performance task into a metacognition task.
On Tough Tongue: career scenarios are the most re-used pattern on the platform: an interviewer persona configured with the job description and the student’s resume as context, a rubric scored automatically after each session, and deployment as a private link or an LMS embed. Career centers assign the same scenario across a cohort and see attempt counts and score distributions without scheduling a single appointment.
Exercise
Build the “tell me about yourself” drill for your students. Write a one-paragraph interviewer persona (industry, seniority, mood, what impresses them), three follow-up probes it should always ask (“what’s the thread connecting those choices?”, “why this company?”, “you have 30 seconds left — land it”), and a four-line rubric: structure, relevance to role, specificity, and time discipline. Whether you implement it on a platform this week or run it live in class, writing it will take twenty minutes and expose exactly how underspecified your current mock-interview practice is.