"Walk me through how you got to 3.2%."
It's the second week of the year-end audit. A controller — four years out of a good accounting program, two of them in public accounting — is on a call with the engagement senior. The subject is the allowance for credit losses. The rate moved from 2.6% to 3.2%, and the auditor wants to understand why.
The estimate is defensible. The controller built it carefully: aging data, two quarters of deteriorating collections in one customer segment, a reasonable forward-looking overlay. The work is genuinely good.
What comes out of her mouth is: "So, we looked at the aging, and, I mean, the trend seemed worse than last year, so we thought 3.2 felt more appropriate. We can look at it again if you think that's aggressive."
Nothing she said is false. And she has just converted a well-supported estimate into a soft one, invited a scope expansion nobody needed, and taught the engagement team to approach every other judgment in her financial statements with more doubt than the work warrants.
She was never taught to do this. It's not in the curriculum. It's arguably the single most common conversation in corporate accounting, and no course covers it.
We train prosecutors and then send them to the defense table
Here is the structural oddity of accounting education. Nearly every course that touches judgment teaches students to evaluate someone else's number. Audit courses teach challenge. Fraud units teach suspicion. Case analyses hand students a set of management assumptions and ask what's wrong with them. The entire posture of the curriculum is adversarial toward the preparer.
And then most graduates become the preparer.
The majority of accounting graduates end up in industry, not public practice — and the ones who start in public accounting overwhelmingly move to industry within a few years. Which means the dominant professional conversation of their careers is one the curriculum never rehearsed: being the person whose judgment is being questioned, by someone whose entire job is to question it, in a setting where their credibility across every other estimate is quietly being priced.
Financial reporting judgment is taught as a technical exercise — pick the method, source the inputs, document the conclusion. But judgment that can't be explained under questioning is functionally the same as judgment that wasn't exercised. The explanation isn't a soft skill bolted onto the technical work. It's the last step of the technical work, and it's the step we don't teach.
The two ways it goes wrong
Defending accounting estimates fails in two opposite directions, and students need practice at both edges.
Hardening. The controller reads questions as accusations. She gets clipped. She stops volunteering context, answers only what's literally asked, and starts arguing rather than explaining. The auditor, entirely rationally, reads defensiveness as a signal of risk and digs harder. This is the failure mode of people who are confident in their work and have never been questioned about it by someone with authority to disagree.
Folding. The controller has good support and abandons it at the first sign of doubt — "we can look at it again," "if you think that's aggressive," "we're not married to that number." This is the more common failure among recent graduates, and the more damaging one. An estimate surrendered without a fight tells the engagement team that management's numbers move when pushed. That inference does not stay contained to the one estimate.
Both failures come from the same root: the student has never separated being questioned from being doubted. An auditor asking how you got to 3.2% is not accusing you of anything. They are performing a required procedure. The ability to hear a hard question as a request for reasoning rather than a challenge to your integrity is a learned response — and nobody learns it by reading that they should.
What a good answer actually contains
Effective controller auditor communication on an estimate has four components. Students can be taught to hit all four in ninety seconds, but only through practice.
1. The method, stated plainly and first. "We used a loss-rate approach by aging bucket, with a qualitative overlay for one customer segment." The auditor needs to place your answer in a framework before they can evaluate it. Leading with the conclusion — "3.2%" — or with the narrative — "well, collections got worse" — makes them do the structuring work themselves, and people who are made to do structuring work get suspicious.
2. The inputs and where they came from. "Aging is straight from the subledger as of 12/31. The segment deterioration is two quarters of DSO data, which is in the file." Sourceable, checkable, unglamorous. Students are well trained here and still tend to skip it, because it feels obvious to them — but it is precisely what the auditor is trying to establish.
3. Why not a different number. "At 2.6% we'd be assuming that segment behaves like the rest of the book, and the data doesn't support that. Above 3.5% we'd be extrapolating past what two quarters can support." This is the component almost no student produces unprompted, and it is the one that carries the most credibility, because it demonstrates that the estimate was chosen from a considered range rather than landed on.
4. The honest boundary. "The softest part of this is the forward-looking overlay — that's judgment, and I'd expect you to want to test it." Counterintuitive, and the most powerful move available. Volunteering where your own estimate is weakest signals that you've thought about it adversarially yourself, and it directs audit attention to the place you're most prepared to support. Students find this almost impossible to do at first because every instinct says don't hand them ammunition. Learning why that instinct is wrong is worth an entire class session.
There's also a tell to teach them to avoid: "We've always used 3%." Precedent offered in place of reasoning is the single fastest way to convert a routine walkthrough into an expanded procedure — and it's the answer that comes out under pressure when the reasoning was never articulated.
Why the current curriculum can't produce this
Faculty aren't ignoring this. The obstacles are structural.
The computation is what's assessable. Estimate coverage in intermediate and advanced courses concentrates on building the number — allowance models, impairment tests, fair value hierarchies, warranty and return reserves. That work is gradeable at scale. The conversation about the number isn't, so it isn't graded, so it isn't practiced.
Written cases have no interrogator. A student writing a memo defending an estimate is composing at their own pace, with the ability to revise, and nobody interrupts on sentence two with "where did that assumption come from?" The memo tests whether they can construct a defense. It says nothing about whether they can deliver one.
We never assign the role. Think about how rarely a student is asked to play the preparer being questioned. Four years of being trained to challenge, then a career of being challenged. The role reversal alone — just experiencing the auditor's questions from the other chair — changes how students think about documentation, and most graduates encounter it for the first time on the job.
Practitioner-led practice doesn't scale. An alum who plays the auditor for a handful of volunteers gives everyone else a spectator experience, once, with no second attempt — and the second attempt is where the learning is.
What simulation changes
AI role play for higher education removes the binding constraint: a credible, patient, endlessly available auditor who asks real questions and doesn't get tired at student twenty.
For this scenario, the counterparty variation is the pedagogy. The same estimate can be defended against:
- The methodical senior, who works through the model systematically and rewards structure.
- The time-pressured manager, who wants the short version and will accept a weak answer if it’s delivered confidently — which teaches students that clarity under time pressure is a distinct skill from thoroughness.
- The auditor who found an inconsistency, holding last year’s memo that describes the method differently. This is the highest-value rep in the set, because it forces the student to reconcile their own documentation in real time without either dismissing the discrepancy or panicking.
- The naive questioner, a new team member who doesn’t understand the business and asks simple questions that are genuinely hard to answer well. Students consistently underestimate this one.
Every attempt produces a transcript and a recording, which matters more here than in most scenarios: the failure modes — hardening and folding — are both instantly visible when a student watches themselves. "We can look at it again if you think that's aggressive" is a sentence nobody defends after hearing themselves say it.
Designing the module
Three passes across a financial reporting or advanced accounting course:
Pass one — the clean walkthrough. A cooperative auditor asks the student to explain a prepared estimate. Score the four components: method, inputs, why-not-another-number, honest boundary. Most students will produce two of four. Showing them that gap is the whole point of the first pass.
Pass two — the proposed alternative. The auditor suggests a different assumption that would move the number materially. The student must engage with the substance, concede where the alternative has merit, and hold where their support is stronger. Score for both failure modes: did they harden, and did they fold?
Pass three — the inconsistency. The auditor surfaces a conflict between the current estimate and the prior-year documentation. The student must acknowledge it accurately, explain the change if it's justified, and commit to running it down if it isn't. Score honesty and composure — this pass doubles as an ethics assessment, because the tempting response is to smooth over a discrepancy the student can't immediately explain.
Build the rubric on observable behavior: Did they lead with method? Did they source inputs specifically? Did they articulate a range? Did they volunteer a weakness? Did they concede anything they shouldn't have?
This is, in effect, audit readiness training — the same preparation corporate finance teams run before every year-end, delivered four years earlier and for a fraction of the cost.
Why this belongs on a dean's list
It serves where your graduates actually go. Programs measure placement into public accounting because it's countable, but the career is mostly spent in industry. A course component that rehearses the dominant conversation of corporate accounting is a curricular differentiator that maps directly onto alumni outcomes rather than first-job statistics.
It produces direct assurance-of-learning evidence. Communication and professional judgment appear in nearly every program's learning goals and are typically measured through written proxies. A scored behavioral simulation yields a direct, rubric-anchored measure of an observed behavior, applied consistently across sections. For AACSB assurance of learning documentation, that is a materially stronger artifact than a reflection essay.
It's experiential learning without the logistics. The constraint on experiential learning business school programming is coordination — partners, scheduling, supervision, uneven quality. Simulation removes all of it and adds what no placement offers: the freedom to fail and immediately retry.
The asset is reusable. Scenario libraries built for the degree program transfer straight into executive education and corporate finance training, where the same gap is being closed at considerably higher price points.
The short version
Accounting programs produce graduates who can build a defensible estimate and can't defend it. Those are different skills, and only one of them is being taught. Accounting communication skills aren't the polish applied after the technical work — under questioning, they are the technical work, because an estimate that can't be explained is treated by everyone downstream as an estimate that wasn't reasoned.
The chair is the same. The questions are the same. The only variable is whether your graduates have sat in it before it counted.
Foretell AI lets faculty build conversational simulations — including controller–auditor estimate walkthroughs like the one above — with configurable counterparties, transcripts, recordings, and rubric-based evaluation. If you're mapping professional judgment and communication outcomes to assurance-of-learning goals, we're happy to walk through how other departments have structured it.