botminds.aiRequest a demo
WorkflowAdmissions & Enrollment Management

Holistic Admissions Review

The reader scores first on every holistic admissions review; the agent's score opens beside it, cited to the page.

The reader scores first; the agent's score opens beside it, cited to the page.

See the demoBook a demo
See one case, screen by screen ↓
demo23citationspieces of evidence pinned to a page and a line, for one applicant file
demo127,340documentssplit from 41,206 applicant files this admissions cycle
target6mindesign goal for a first read, against 15–45 minutes reported by hand
target100%of two-point score gaps reach a second reader before any decision
In short

What is Holistic Admissions Review?

Holistic admissions review is how a selective office judges an application file against its own multi-row rubric — curiosity, leadership, contribution, persistence and fit, each read against the opportunities the applicant's own school offered, not test scores or grades alone. Holistic Admissions Review puts five agents in front of that reading: they assemble the file, pin every score to its page and line, read the school's own profile, flag any sentence naming a protected trait, and score all seven rubric rows into a slot sealed until the reader scores first.

AI in education: a buyer's guide to agentic AI for colleges, universities and school districts →

Updated

The problem

Why holistic admissions review takes eight to fifteen minutes a file, and proves nothing afterward

Holistic admissions review is how a selective office judges an application file against its own multi-row rubric, not a test score or a grade point alone — and a public research university reads tens of thousands of first-year applications in a ten-week season, most of them in the ten days after early action closes. The Common App carried 10.79 million applications from 1.53 million applicants in 2025-26, up 6%, at 7.06 applications per applicant, so every office reads more files per enrolled student than it did five years ago.

A first read gets eight to fifteen minutes at most selective offices, and the first three of those minutes are usually spent finding things — which page is the school profile, how many AP courses the school offered, what a teacher actually said. Nothing a reader writes points back at the page it came from, so a second reader starts from zero and readers drift from each other as a long season goes on. Since the 2023 SFFA ruling, an office must be able to show race played no part in its scoring.

typical15–45mina manual first read, by vendor report — directional, not measured here
typical10.79millionCommon App applications carried in 2025-26, up 6% year over year
typical19–27$/hrpublished pay for seasonal admissions readers trained for three weeks
Where a first read's minutes go, by hand and with the agentsestimated
By hand12 minutes
With the solution6 minutes
  • Finding which page in the packet has the evidence3 → 0.2 min
  • Reading the application, the essay and the letters5 → 3 min
  • Checking the school profile for rigor context2 → 0.3 min
  • Scoring and citing all seven rubric rows2 → 2.5 min

Estimated split of a typical first read's minutes; the 'after' column totals about six minutes, the design goal — our arithmetic, kept modest.

How it works

How one applicant file is read

Five agents read, cite and check every file before the reader opens it; a sixth measures the readers each night.

What comes in
One file in6 documents · application, essay, two letters, school profile, transcript
Agents at work
File intake agentsplit and number every page
Then
Context agentwhat the school offered
Reader agent7 rows, every score pinned
Then
Compliance agentprotected traits flagged, with the rule
Then
Sealed packetheld closed until the reader scores
A person decides
The reader scores firstthen the seal opens beside it
What comes out
Seven cited rows
Every gap to a second reader
Chair records the decision
One case, step by step

One file, through holistic admissions review to a signed decision

Applicant HSU-27-104882, Brenmoor High School, applying to the College of Engineering, early action. Here is what happens to the file, screen by screen, in the working solution.

  1. 016:12 AM

    The packet built before the office opens

    Five agents

    Before Talia Ferreira signs in, the agents have split the file's six documents into ten numbered pages — application, essay, two recommendations, the Brenmoor High School profile and the transcript — and pinned 23 pieces of evidence to the lines they came from. All seven rubric rows are scored already, into a slot that stays sealed until she scores.

    2 minutes 41 seconds to assemble the packet; 23 citations pinned.

  2. 029:04 AM

    She scores first, then the seal opens

    Talia Ferreira · Senior Admissions Reader

    Talia scores Intellectual curiosity a 3 and locks the row. Only then does the agent's score appear: a 4, at confidence 0.88, citing the physics teacher's letter, paragraph 3, and the essay, page 2, line 14 — "I rebuilt the school's broken weather station from parts the shop teacher had written off."

  3. 03One click

    A score joined to the page it came from, both ways

    Talia

    Clicking the citation scrolls the file to line 14 and draws a thin line to the pin; clicking the pin instead lights up the rubric row it fed. A second reader inherits the same lines rather than starting again.

  4. 04Context

    Rigor read against what the school actually offered

    Context agent

    Brenmoor High School offers two Advanced Placement courses and no International Baccalaureate program; this applicant took both AP courses and a dual-enrollment course beyond the school. The Context line states that under the rubric, so rigor is a cited row, not a reader's private adjustment.

  5. 059:18 AM

    She raises her score, with a reason on the record

    Talia

    The evidence changes her mind. She raises Intellectual curiosity from 3 to 4, picks the reason code "Evidence I had not reached," and initials the row — her first score stays on the record, and the change is a second entry, not an overwrite.

  6. 06On Leadership and initiative

    Two points apart, so the file stops

    The rubric

    Talia scores Leadership and initiative a 2 against the activities list; the agent scores a 4, citing the counselor's letter, paragraph 4, and the essay, page 2, line 22. A two-point gap on any single row is a hard stop — the file cannot reach the committee until another reader settles it.

  7. 079:41 AM

    A second reader settles the row — never the first reader

    Marcus Ebel · Second reader

    The file goes to Marcus with both reads and both sets of citations attached; the six rows that agreed stay closed. He settles Leadership and initiative at 3, with the reason code "Second read: evidence beyond the activities list."

  8. 08Paragraph 2

    A protected trait, flagged with its rule — the reader decides

    Compliance agent

    One sentence in the counselor's letter names the applicant's national origin. The agent keeps it out of its own score and scores the two rows it touched again without it; the sentence itself is untouched, shown to Talia with the rule beside it. She decides it is the counselor's account, not the applicant's own, and weighs it as individual experience — either answer would have gone on the record.

  9. 094:40 PM

    Only the chair records the decision

    Dr. Priya Venkatesan · Admissions Committee Chair

    With every row resolved and cited — a final total of 27 of 35 — the file reaches the chair's list. She records Admit, College of Engineering, early action, with the reason code "Meets the early action band (26+)," e-signed. A reader can see the committee list and change nothing on it.

    Open to a decision in under two working days, every one of the seven rows cited to a page and a line.

  10. 10Audit-ready

    Every reveal and decision, exportable in one pack

    Carla Whitlock · Internal Audit

    Every read, reveal, score, change, flag and decision on this file — by a person or an agent — sits on a trail that cannot be edited. Carla's evidence pack exports the six documents, all 23 citations, the flag and Talia's decision, the rubric version, the second read and the chair's signature, in one click.

Who it’s for

Built for everyone who reads, resolves or signs a file.

The same applicant file, carried by four people — what their day looked like, and what it looks like now.

TF
Talia FerreiraSenior Admissions Reader · 31 files this week
Senior Admissions Reader
Before
Reads each file in eight to fifteen minutes, the first three spent finding which page has the evidence.
Now
Scores first, then sees the agent's cited score beside her own and raises it with a reason, in seconds.
ME
Marcus EbelSecond reader · settles every disputed row
Senior Admissions Reader · second reads
Before
Settles a disputed row from a free-text note, starting from zero because nothing points back at a page.
Now
Opens the file with both reads and both sets of citations already attached, and settles the row in minutes.
NC
Dr. Naomi CastellanosDean of Admissions · owns the rubric for each college
Dean of Admissions
Before
Learns a reader has drifted only at a norming session, or not at all.
Now
Sees every reader's drift against the pool and against a constant agent, row by row, the morning it appears.
CW
Carla WhitlockInternal Audit · reads everything, changes nothing
Internal Audit
Before
Rebuilds a file's evidence from several systems when an auditor or an investigator asks.
Now
Exports one evidence pack per file — every document, citation, score, flag and signature, as it was the day it was decided.
Built on the engine

6 agents. Each with one job, and hard limits.

Five agents read, cite and check every file before the reader opens it; a sixth measures the readers each night.

File intake agent

Pulls each packet from the admissions CRM and the application portal, splits it into application, essay, recommendations, school profile and transcript, and numbers every page and line so a citation can name it.

  • Never alters an original document
  • Never marks a file complete when a document is missing
Context agent

Reads the high school's own profile and the published school data file, so the rubric's "rigor against what was offered" row has a stated basis on the screen.

  • Uses school-level facts only
  • Never infers anything about a person, and never uses a protected trait
Reader agent

Scores all seven rubric rows and pins each score to the page, line and sentence it came from, with a confidence — sealed until the reader locks the row.

  • Never shows a score before the reader locks the row
  • Abstains below the confidence floor of 0.70, and says so
  • Never records a decision
Compliance agent

Flags any sentence naming race, color, national origin, religion or sex, keeps it out of the agent's own score, scores the affected rows again without it, and shows the rule beside the sentence.

  • Never alters or removes a sentence
  • Never decides what a flagged sentence means — the reader does
  • Every flag is shown and logged, never silent
Assignment agent

Routes each complete file to a reader by deadline, workload and declared conflict of interest, escalates a file within a day of its deadline, and routes a two-point gap to a second reader.

  • Never assigns a file to a reader who declared a conflict
  • Never moves a file past an unresolved gap
Calibration agent

Every evening, compares each reader's scores with the pool and with the agent's scores on the same files, row by row, and reports drift to the Dean of Admissions.

  • Never changes a score
  • Reports drift; never acts on it
Reader, second reader, chair

Score, settle, decide. The agents propose; a named person decides.

Every screen

The working solution, as it ships.

13 screens from the working solution, on its sample data. Pick one to see it large.

One applicant file, and the rubric beside itApplicant HSU-27-104882 from Brenmoor High School, applying to the College of Engineering. The file's ten pages run down the left; Halden State's seven-row reading rubric runs down the right. Each row has two score slots: the reader's own, and the agent's — sealed until the reader locks the row.
The reader scores first, then the seal opensTalia Ferreira scores intellectual curiosity a 3 and locks the row. Only then does the agent's score appear beside hers: a 4, at confidence 0.88, with the three places it came from. Her read is independent, and the reveal is a recorded event.
Every score joined to the page it came fromClicking a citation scrolls the file to the exact line and draws a thin line from the agent's score to the pin in the text. Essay page 2, line 14 — the sentence that earned the 4. Click a pin in the document and the rubric row it fed lights up instead.
Rigor read against what the school actually offeredA Context agent reads the high school's own profile, so course rigor is judged against the opportunities the applicant had. Brenmoor offers two Advanced Placement courses and no International Baccalaureate program; this applicant took both AP courses and a dual-enrollment course beyond the school.
Changing a score after the reveal takes a reason codeThe evidence changes Talia's mind, so she raises intellectual curiosity from 3 to 4. Her first score stays on the record; the change is a second entry with its reason code, her initials and the time. Nothing is overwritten.
Two points apart, so the file stopsOn leadership and initiative the reader scores a 2 against the activities list and the agent scores a 4, citing the counselor's letter and the essay. A two-point gap on any single row is a hard stop: the file cannot reach the committee until another reader settles it.
A second reader settles the row — never the first readerThe file goes to Marcus Ebel with both reads and both sets of citations attached, and the six rows that agreed stay closed. Talia can see it, but she cannot settle it: a second read is always another reader.
A sentence naming a protected trait, flagged with its rule — the reader decidesOne sentence in the counselor's letter names the applicant's national origin. The agent kept it out of its own score and scored the two rows that touched it again without it. The sentence itself is untouched: it is shown to the reader with the rule beside it, and the reader decides what it means. Either answer goes on the record.
Every file in one queue, with the deadline on eachThe reader's own files, sorted by deadline, each with a packet already cited. Three are due within a day, and the assignment agent escalates a file a day before it is due — and will not offer one to a reader who has declared a conflict with that school.
Only the committee chair records a decisionFiles reach the chair's list only when every row is resolved and every score is cited. A reader can see the list and change nothing on it; a file with an open second read is held on a separate list and cannot be decided at all.
Reader drift measured every night, not at the end of the seasonBecause the same agent scores every file, each reader can be measured against the pool and against a constant, row by row, the day drift appears. One seasonal reader is nine tenths of a point low on leadership, so the Dean can call a norming session in week two rather than in March.
The rubric belongs to the Dean, and every version is datedOne rubric per college, versioned for each cycle, with the anchor for every score written in plain words. The rules the agents follow — the gap that forces a second read, the confidence floor below which the agent abstains, who may record a decision — are settings here, not code.
One evidence pack per file, for the auditorEvery read, reveal, score, change, flag, flag decision, assignment and decision — by a person or an agent — in one stream that is never edited. An auditor exports the file exactly as it was when it was decided, with identifiers redacted unless they are named on the hold.
Governance

Built for the question a 2023 ruling made every office answerable for.

Six roles, separated by dutyA reader scores and cannot decide; a second reader resolves a gap and never their own file; the Admissions Committee Chair alone records admit, deny, waitlist or defer; the Dean of Admissions owns the rubric and scores no individual file; an administrator configures the system with no file access; Internal Audit reads everything and changes nothing.
A two-point gap forces a second readerAny rubric row where the reader and the agent differ by two points or more is a hard stop: the file cannot reach the committee until a different reader resolves that row with a reason code. The committee's list is assembled from resolved reads only.
Every score cites its page and lineA rubric score is never a number alone — it is joined to the page, the paragraph or line, and the sentence it came from, in both directions. A second reader, the chair and an auditor all see the same lines.
Protected traits are never a scoring inputThe Compliance agent keeps any sentence naming race, color, national origin, religion or sex out of its own score, scores the affected rows again without it, and shows the sentence and the rule to the reader, who decides what it means — on the record either way.
A complete, exportable audit trailEvery read, reveal, score, change, flag, flag decision, assignment, rubric publication and decision — by a person or an agent — is on the record with who, what, when, why and the evidence, exportable as one pack per file, or per cycle for the flag register.
FERPA-aware access and retentionAccess to a file is logged per reader, records are kept three years after the cycle before deletion, and identifiers are redacted in an export unless an auditor is named on a hold. The product says what it does; it makes no compliance promise.
Configuration

Your rubric and your rules, not ours

The thresholds the agents follow are settings, versioned and on the audit trail — not code.

SettingDefaultChoose from
Second-read gap2 pointsThe per-row difference that forces a second reader
Agent confidence floor0.70Below it, the agent abstains on a row and says so
First-read SLA5 working daysFrom file complete to first read
Escalation1 day before the SLAThe file joins the escalation strip
Decision band26 of 35 (Engineering)Set per college, per cycle
Who may record a decisionAdmissions Committee ChairNever a reader or a second reader
Who may resolve a gapSenior Admissions ReaderNever the file's first reader
Connections

Works with the systems an admissions office already runs on

Admissions CRMapplications, plans and reader assignments
Application portalessays, recommendations and activities lists arrive here
Document managementtranscripts and school profiles, kept for the record
Student information systemdecisions written back after the chair signs
Identity providerrole-based sign-in for readers, the chair, the Dean and audit
What it changes

The difference, in numbers.

Every figure is labelled: a target the solution is built to, an estimate, a typical published result, or a proven one.

target
7 of 7
rubric rows cited to a page and line before the reader scores
Every row cited
or the agent abstains and says so
target
6min
per first read, against 15–45 minutes reported by hand
By hand15–45 min
With agents≈ 6 min
target
100%
of two-point score gaps reach a second reader before any decision
Enforced by the rubric, not by memory

Sources: Common App, 10.79 million applications from 1.53 million applicants in 2025-26, up 6% (acceptancerate.com, August 2026) · The Chronicle of Higher Education, "Reading an Application in Under 10 Minutes? Way Too Fast, One Admissions Dean Says" · published seasonal-reader postings, $19–27/hr · the 2023 Supreme Court admissions ruling (SFFA), June 29, 2023, and the Department of Education's Dear Colleague Letter · FERPA, 34 CFR Part 99 · College Board, Understanding Holistic Review in Higher Education Admissions · "target" = design goal, measured in the live solution · "demo" = seen in the working solution, on its sample data · "typical" = published figures. Halden State University and the people named on this page are fictional — characters and sample data in the working solution.

Questions

What admissions offices ask us.

What is holistic admissions review?

Holistic admissions review is how a selective office judges an application file against its own multi-row rubric — curiosity, leadership, contribution, persistence and fit, each read against the opportunities the applicant's own school offered, not test scores or grades alone. Holistic Admissions Review puts five agents in front of that reading: they assemble the file, pin every score to its page and line, read the school's own profile, flag any sentence naming a protected trait, and score all seven rubric rows into a slot sealed until the reader scores first.

How long does holistic admissions review take?

Published figures put a first read at most selective offices at eight to fifteen minutes, with vendors reporting 15 to 45 minutes by hand. The design goal for Holistic Admissions Review, measured in the live solution, is about six minutes once every rubric row is already cited — the first three minutes usually spent finding the evidence are mostly gone.

Is AI used in college admissions review, and can it be biased?

An agent scores all seven rubric rows, but it never sees the reader's score first and never records a decision. Since the 2023 Supreme Court admissions ruling (SFFA), a protected trait is never an agent's scoring input: the Compliance agent flags any sentence naming race, color, national origin, religion or sex, keeps it out of its own score, and leaves what the sentence means to the reader, on the record either way.

Does the agent decide, or read a student's record on its own?

No agent in Holistic Admissions Review decides, sends anything or records an outcome. The Reader agent scores and cites; the Compliance agent flags and leaves the call to a person; the Assignment agent routes a file but never moves it past an unresolved gap; only the Admissions Committee Chair can record an admit, deny, waitlist or defer, with a reason code and an e-signature.

What happens when two readers' scores are two points apart?

A two-point gap on any single rubric row is a hard stop: the file cannot reach the committee until a second reader — never the first reader — resolves that row with a reason code, with both sets of citations attached. In the sample cycle, 420 gaps were raised, 291 resolved and 129 still open, each one listed apart from the chair's ready-to-decide list.

Do people stay in control of the decision?

Yes. The reader always scores before the agent's score is shown, any change after the reveal needs a reason code, and only the Admissions Committee Chair can record admit, deny, waitlist or defer. Internal Audit can read and export everything, and can change nothing.

How long does it take to go live?

The Agentic Solution Engine builds and deploys Holistic Admissions Review from your own rubric, your systems and a sample of past files, and it goes live once every quality gate has passed. We'll walk your office through it on a sample of your own applications first.

Admissions & Enrollment Management

More in this division.

All 8 solutions →
Transfer Credit Evaluation

Credit decided in a day, not a week, with every outcome threaded to the syllabus.

1-min demoTranscript Processing Automation

Every course line read twice, reconciled to the grade key, and posted the same day.

1-min demoCredit for Prior Learning Evaluation

Credit earned before day one, with every credit tied to the evidence it rests on.

1-min demo

See it on
your own rubric.

We'll run Holistic Admissions Review on a sample of your own files and your own rubric.

Book a demoBrowse solutions