Academic program review with five cited lenses per program, the self-study checked against the data, and the options modeled before the Provost decides
Each program review assembled in two days, not six weeks, every figure cited.
See one case, screen by screen ↓Academic program review is the cyclical evaluation of each degree program in higher education, usually every five to seven years, in which the department writes a self-study, a faculty committee reviews it with the program's figures, and the provost decides whether to grow, redesign, merge or teach out the program. Academic Program Review gives each program one cited page: five lenses built by agents from the systems of record, the self-study checked claim by claim, and the options modeled, with the committee approving before the Provost decides.
AI in education: a buyer's guide to agentic AI for colleges, universities and school districts →
Updated
An academic program review is the cyclical look a university takes at each degree program, usually every five to seven years: the department writes a self-study, institutional research pulls enrollment, cost, completion and earnings figures from four or five systems, a committee of faculty reads it all, and the provost decides whether to grow, redesign, merge or teach out the program. In the academic program review process as most campuses run it today, the figures arrive as spreadsheets weeks after the self-study is written, so nobody checks the self-study's claims against the data.
Departments, institutional research and the budget office each count majors and credit hours their own way, and the committee spends its meetings reconciling whose number is right. Options such as a merger or a teach-out are proposed without the students affected, the shared credits, the approvals or the money. Meanwhile the stakes have risen: state laws now close low enrollment programs, and from July 2027 a program whose graduates earn too little can lose federal student loans. Academic Program Review puts eight agents in front of the assembly, checking and modeling, so the committee and the Provost spend their time on program viability, not on the numbers.
Estimated split of one program review's assembly, by hand and with the agents. The 'before' column, about 27 working days, matches the demo university's own estimate of five to six weeks by hand; the 'after' column totals two days, the design goal. Our arithmetic, kept modest.
Five agents build the lenses, a Critic checks the self-study, a Peer and a Scenario agent compare and model; the committee approves and the Provost decides.
The BS in Industrial Technology at Halden State University, CIP 15.0612, was last reviewed in 2019. Its department chair submitted the self-study on September 18, 2026. On Wednesday, October 14, the review is with the Provost, with 6 days left on her clock. Here is what happens to it, screen by screen, in the working solution.
27 programs are in full review this cycle and 3 wait for the Provost, the soonest in 6 days. The queue shows what the agents found on each one: the BA in French confers 3.7 degrees a year, and the MS in Applied Statistics has 38% more majors.
Labor market data has not answered since 6:10 AM, so one job demand strip is held rather than built from old figures.
Enrollment is 184 majors, down 41% since fall 2021 while the university fell 3%. A credit hour costs $612 against the university's $388, six-year graduation is 52% against 64%, graduates earn a median $58,400 against $52,900, and regional postings for related roles rose 22% to 2,245. Each strip names its system of record, an as-of date and a confidence score.
Cost per credit hour climbed from $431 to $612 over fall 2021–2025, while the average section shrank from 15 students to 11 and sections under 10 students rose from 21% to 38%. The method is direct instructional cost by section divided by credit hours, from ERP cost lines joined to HR assignments, as of October 9, confidence 0.97.
Counts only; no student-level record leaves the student information system.
The Critic agent tested 7 claims: 3 supported, 1 partly supported, 2 not supported and 1 that cannot be checked. "Enrollment has stabilized" is not supported: majors fell from 246 to 184 since fall 2023. The two partner letters cannot be checked, and on October 12 the department chair answered that one partner is renewing its letter with an amount.
"Strong job placement" rests on an 18% survey response: 7 of 39 graduates.
Of eight peer universities, five offer the program. From public completions data, Halden's completions fell 36%, from 61 to 39; Peer A rose 7% after adding an automation concentration, and Peer D finished a teach-out in 2024. Peers are labeled, never named.
Peer B merged the same program into engineering technology in 2023, and combined completions rose 19%, from 64 to 76.
A merger with Engineering Technology brings a credit hour to $419 and instructional cost down an estimated $1,310,000 a year; all 184 current majors finish under their catalog, and two retirements are not refilled, with no layoffs. A teach-out saves more by 2031 but leaves 2,245 regional postings with no local program. The merge rests on 23 shared credits, and the Provost can change that assumption and watch the figures move.
The Scenario agent never recommends; the person picks.
Joan picks the merger in the recommendation box. The Scenario agent drafts it with every figure linked to its source: an Industrial Technology concentration within the BS in Engineering Technology from fall 2028, 23 of 41 required credits shared. She sends it to the academic program review committee, and its 30-day clock starts.
Leah approves the review with reason CR-02, approve with the department's response attached. Until then the Provost's decision stays locked, and the screen tells Joan the Provost cannot approve for the committee. A committee reader from the department under review is recused and cannot vote. Joan then decides with reason PR-04: merge, from fall 2028.
Daniel releases version 4 of the record to Priya Venkataraman, Faculty Senate President: the summary, the five lenses, the self-study and its check, the options modeled, the approvals and the decision, and the department's response. Every count under 10 is suppressed.
The send, the committee's approval with one member recused, the decision and the release sit on a trail that names who, what, when, why and the evidence, for people and agents alike. Entries are kept 10 years after the decision, longer under legal hold, and the record pack exports for the board, the accreditor or a state reviewer.
The same review, seen by the five people who write it, check it, decide it and receive it: what their cycle looked like, and what it looks like now.
Five agents build the lenses, a Critic checks the self-study, a Peer and a Scenario agent compare and model; the committee approves and the Provost decides.
Builds the enrollment strip: fall-census majors, new majors and credit hours for five years, against the university.
Builds cost per credit hour from instructor salary and benefits by section, joining HR assignments to ERP cost lines.
Builds six-year graduation and degrees conferred for the program's entering cohorts.
Builds graduates' median earnings four years after completion and runs the federal earnings test, with the small-cohort roll-up the rule requires.
Builds regional postings for the occupations linked to the program, and names the region.
Finds every factual claim in the self-study and tests it against the five lenses and the attached evidence, with a verdict and the figure that settles it.
Compares the program with the board-approved peer universities from public completions data, and notes what each peer did.
Models grow, redesign, merge and teach out to five years, with the students, faculty, money and approvals each affects, and drafts the text for the option a person picks.
Decides after the committee chair approves. The agents propose; a named person decides.
The assistant answers in plain words, cites the source of every figure, and opens the screen it is talking about.
Is the job placement claim supported?
Not by the data. The self-study says virtually every graduate works in the field, but the placement survey had an 18% response rate: 7 of 39 graduates of 2024–25. Six of the seven work in the field. State wage records find 31 of 39 employed in the state, with no occupation recorded. The department can answer the claim before the committee reads it.
What happens if we merge with Engineering Technology?
With 23 of 41 credits shared, a credit hour falls from $612 to $419 and instructional cost by an estimated $1,310,000 a year by year three. All 184 current majors finish under their catalog, and two retirements are not refilled, with no layoffs. It needs the curriculum committee, the senate, the Board of Trustees and notices to the state board and the accreditor.
Which programs fail the earnings test?
Three of 190: the BFA in Musical Theatre ($31,900) and the BA in Studio Art ($33,400) against the $36,200 benchmark, and the MFA in Creative Writing ($41,200) against the $47,600 graduate benchmark. Five more are within 10%. A first failure means warning current students within 30 days; two failures in three years cost Direct Loan eligibility.
Prepare the record for the faculty senate
The record is assembled: five lenses with their sources, the seven claims and their verdicts, the four options, the signatures and the department's response. Counts under 10 are suppressed, so the placement survey reads "fewer than 10 of 39". It stays a draft until the Provost decides and the Vice Provost releases it.
13 screens from the working solution, on its sample data. Pick one to see it large.
The approvals, thresholds, clocks and privacy rules every review runs on are settings the university owns, each change versioned.
| Setting | Default | Choose from |
|---|---|---|
| Who approves the review | Program Review Committee Chair | Nobody approves their own step, always on |
| Full review cycle | Every 7 years | Stepped, versioned |
| Check-in flag: degrees a year | Fewer than 5, averaged over three years | The effect this year is shown: 5 programs |
| Other check-in flags | Enrollment down 25%, cost 1.40× the university's, earnings within 10% of the benchmark | Each a stepper |
| Clocks | Self-study 60 days, committee 30, Provost 21, check-in response 45 | Escalation to the Vice Provost 5 days before a clock runs out |
| Minimum cell size | 10 | Counts below it read "fewer than 10"; redaction on export |
| Record retention | 10 years after the decision | Legal hold |
| Agent confidence floor | 0.90 | Below it, a person confirms the figure; versions pinned per cycle |
Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.
Sources: One Big Beautiful Bill Act earnings accountability, HEA §454 as amended (final rule July 1, 2026; effective July 1, 2027) · Ohio SB 1, ORC 3345.454 (fewer than five degrees a year over three years) · Kentucky Council on Postsecondary Education program review process · "target" = design goal, measured in the live solution · "demo" = seen in the working solution, on its sample data · "typical" = published rules · "estimated" = our arithmetic. Halden State University, its programs, its peers and the people named on this page are fictional: characters and sample data in the working solution.
Academic program review is the cyclical evaluation of each degree program in higher education, usually every five to seven years, in which the department writes a self-study, a faculty committee reviews it with the program's figures, and the provost decides whether to grow, redesign, merge or teach out the program. Academic Program Review gives each program one cited page: five lenses built by agents from the systems of record, the self-study checked claim by claim, and the options modeled, with the committee approving before the Provost decides.
The department writes a self-study; the figures for enrollment, cost, completion and earnings are assembled; the academic program review committee reads both and recommends; the provost decides; and the faculty senate receives the record. Here the agents assemble the five lenses overnight and check the self-study, the department answers the check, the Provost sends an option to the committee, the chair approves with a reason code, and only then can the Provost decide.
At the demo university, assembling one review's figures by hand takes an estimated five to six weeks, so they arrive after the self-study is written. The design goal for Academic Program Review is two days to assemble a review with every figure cited, measured in the live solution.
The Critic agent finds every factual claim in the self-study and tests it against the five lenses and the attached evidence, marking it supported, partly supported, not supported or cannot be checked, with the figure that settles it. The department chair sees the check and can answer it before the committee reads the review, and a person can set any verdict aside with a reason code.
Yes. Between full reviews, the agents rebuild the five lenses for every program each September and test them against the university's own thresholds for degrees a year, enrollment, cost and earnings; on the sample data, 14 of 190 programs are flagged. The Vice Provost asks the flagged departments in bulk or moves a program into next year's full reviews.
Under the earnings test added to the Higher Education Act in 2025, graduates' median earnings four years after completion must beat a benchmark. After a first failure the institution must warn current students within 30 days and prospective students before they enroll, and two failures in three years cost the program its Direct Loan eligibility, from July 1, 2027. The Earnings agent runs the test on every program a year early, including the small-cohort roll-up.
Yes. The agents build, check and model; they never decide, recommend on their own, send, release, or bring a student-level record into the review. The committee chair approves the review with a reason code, the Provost decides with a reason code only after that, and the Vice Provost releases the record to the senate. Nobody approves their own step.
The Agentic Solution Engine builds and deploys it from your requirements and documents, such as your review policy, past self-studies and the systems your figures come from, and it goes live once every quality gate has passed. We will walk you through it on one of your own programs first.
No uncited sentence reaches the accreditor, and likely findings are answered first.
1-min demoIPEDS ReportingEvery survey cell rebuilt from source, every variance explained before lock.
1-min demoFaculty CredentialingEvery section cleared before the term starts, with every credential tied to its proof.
1-min demoWe'll run Academic Program Review on one of your programs, from your own systems and self-study.