A study of how selective employers and universities choose

Same candidate, different system.

Every one of the six markets sampled sets a test or assessment and an interview. After that they diverge: a writing task in three, a case or simulation in two.

The purpose of this site: to set out what selective employers and universities publish about how they choose, market by market, and to make every figure here checkable against the source it came from.

Why now: AI makes a polished application easy to produce. Where sampled institutions publish AI rules, they protect ownership of the work rather than ban AI outright.

See the evidence Download the reports Press and interviews

What each market's published selection includes

UKGCCINUSSGHKof 6

Prior record5

Test or assessment6

Writing task3

Case or simulation2

Interview6

Panel or committee4

Verified in the official source Not observed in the sources reviewed

Full matrix, definitions and sources

WRITING AND COMMENTARY PUBLISHED IN

Hassan Akram's own writing and commentary, named for context. Neither publication reviewed or commissioned this research.

What this is

A research programme published by Elite Careers Strategy. It reads how selective employers and universities say they choose people, in their own official words, and compares six markets.

Why it exists

The rules of elite selection are not secret, but they are scattered across institutional pages and unevenly understood. This programme collects them, codes them against a published method, and publishes every source so that any figure can be checked rather than taken on trust.

Who it is for

Journalists, researchers and institutions examining how selection works, and anyone trying to understand it. Elite Careers Strategy is an advisory firm working in this field; that interest is disclosed, and no client evidence is used.

What was studied

107 coded selection signals, drawn from 69 official employer and university pages, across 6 markets: the UK, the Gulf (GCC), India, the United States, Singapore and Hong Kong. Two reports, one public source register.

What a selection signal is

One thing an institution publishes about how it chooses: a requirement, a stage, an assessment, a statement the institution makes about selection, or a material change to the process. One official page can carry several.

How to use this site

Two minutesRead the headline finding and the four findings below.Start here
Ten minutesRead a report: Report 02 compares six markets; Report 01 covers AI and the written application.The reports
Checking a figureEvery coded signal carries a source ID, the official page and a review date, published in full in the register.Source register
If you are a journalistCleared quotes, biography, charts and what the research must not be reported as saying.Press kit

The headline finding

Two checks appear in every market sampled. The rest of the system is local.

How to read itEach dot is a verified passage on the institution's own page. An open ring means not observed in the sources reviewed, not absent.

26 of 36 cells verified
Checked 31 August 2026, re-verified 8 September 2026

Read Report 02

Exhibit 3What the sampled sources actually showEach cell was checked against the official source page. Two of the six evidence types are verified in all six markets: a test or structured assessment, and an interview. A prior academic record is verified in five. The assessment the institution sets itself is where the sampled systems differ most.

Evidence type

UKUK

GCCGCC

IndiaIN

USUS

SingaporeSG

Hong KongHK

VerifiedOf 6

Documentary evidence of prior attainment

Prior record or academic threshold

Assessment set by the institution

Test or structured assessment

Writing assessment

Case, task or simulation

Interaction and decision

Interview

Panel assessment or committee review

Verified present

Scroll to compare all six markets

Verified in the official source Not observed in the reviewed source set

Boundary

This does not establish prevalence in any country or sector. The counts describe the sampled institutions only.

This does not establish that any market is harder, fairer or more predictive than another. The matrix compares what the sampled sources publish; it does not rank.

This does not establish absence. "Not observed" means the feature was not found in the sources reviewed. It does not mean the institution does not use it.

This does not establish national practice. Two to three institutions were reviewed per market, and Freshfields appears in both the UK and the Hong Kong sample, so the two columns are not independent.

Every cell in this matrix is verified against a passage on the institution's own page. The full mapping is published as the cell audit (CSV). It carries the quoted passage, source URL, retrieval method and date, verification status and publication ruling for every cell.

Source Global Elite Selection Monitor, Report 02 purposive sample: 50 signals, 21 official sources, two to three institutions per market, coded 31 August 2026 and re-verified against the official source pages on 8 September 2026. Binary and conservative: marked present only where the cited source explicitly describes it. Two of the nine Hong Kong signals are coded jointly with the UK.

About this sample and the cell definitions

This is a purposive six-market sample: eight coded signals per market at first publication, with a ninth coded for the United States and for Hong Kong on 9 September 2026, two to three institutions each. The allocation is therefore no longer equal, and those two additions were found while looking for support for specific cells rather than sampled neutrally. The institutions are not matched by type, and the pathways are not equivalent. Indian management-school admissions, US undergraduate admissions and employer recruitment are different processes compared side by side, not the same process in six places.

Cell definitions Prior record or academic threshold: a stated academic record, transcript or grade threshold used as an input to the decision. Test or structured assessment: a test or structured assessment set or required by the institution; timing and supervision are recorded only where the source states them, and were stated in one market only. Writing assessment: a writing task set and assessed by the institution during selection. It excludes a previously written paper submitted with the application and excludes the application itself. Where the source does not state where or under what conditions the writing is produced, the cell records a writing assessment and nothing about supervision or location. Case, task or simulation: a case, exercise, simulation or practical task performed during selection. Interview: a scheduled interview. Panel assessment or committee review: either an interview conducted by more than one assessor, or a named committee that reviews or decides applications. Subtypes behind each positive cell in that row: India, both (ISB panel interview of faculty, alumni and senior staff; ISB Admissions Committee review). US, committee review only (Harvard states that the Admissions Committee considers many other factors). Singapore, panel assessment only (DBS panel interview involving HR and senior management).

What else we found

Four findings, each published with its limit

How to read these findingsEvery finding is published with the claim it does not support. If you quote the finding, quote the limit.

The programme's argumentAI has made polish free. It has not made judgement free.

All findings

01
Selection includes assessment the candidate does not write at leisure.
Verified examples, one per institution: a timed cognitive assessment at KPMG; an assessment centre with a written and a group exercise at PwC Middle East; a case discussion at Strategy& Middle East; essays written on the interview day at the Indian School of Business; an Admissions Written Test before interview at NUS Law; an in-person Case Challenge at DBS. No proportion is claimed: the inclusion rule and row-level review are not complete.
This does not establish that live assessment is new, increasing or superior, and says nothing about prevalence beyond the observations coded.
02
The AI rules observed concern ownership, not blanket prohibition.
Six observations were coded newly explicit about AI. They come from three institutions: DLA Piper, UBS and Cambridge, which permit forms of preparation while protecting independent assessment and truthful representation.
This does not establish that AI caused institutions to introduce live assessment. Fifty of 57 observations were coded as existing processes at the review date.
03
Every sampled market checks something beyond the written application. The instruments differ.
Every market in the sample includes a published mechanism that checks more than a self-authored application, but the instrument is local: supervised retesting, on-the-spot writing, authenticated school work, case challenges, language tests.
This does not establish prevalence in any country or sector. Two to three institutions were reviewed per market.
04
Access to the live gate is itself designed.
Shortlisting, programme preference, prior tests and invitation-only events determine who is allowed to demonstrate judgement at all.
This does not establish intent, fairness or outcome effects. The tension is identified, not measured.

The evidence behind the findings

In the institutions' own words

Coded sample57 signals in Report 01, coded from official employer and university pages only. No private client evidence, no modelling, no estimates.

Percentages describe the coded sample only.

Exhibit 1Assessment set by the institution, in the institutions' own wordsSix verified examples of assessment beyond the written application. The programme does not publish a proportion for this, because the inclusion rule and row-level review are not complete.

Market

Institution

What the source states

Source

KPMG

A cognitive assessment of numerical and logical reasoning. The page states it is timed and gives 36 minutes.

SRC-051

GCC

PwC Middle East

An assessment centre combining reasoning tests, a written exercise, a group exercise and a partner or director interview.

SRC-054

GCC

Strategy& Middle East

Three to four interviews, including a real-world case discussion.

SRC-056

India

Indian School of Business

A panel interview of faculty, alumni and senior staff, with essays written on the interview day.

SRC-057

Singapore

National University of Singapore

From 2026, Law applicants must pass an Admissions Written Test before they are invited to interview.

SRC-065

Singapore

DBS

An in-person Case Challenge, followed by a panel interview involving HR and senior management.

SRC-067

Each row was read on the institution's own page and is quoted at the level the page supports. These are examples, not a count: no proportion is claimed from them.

Boundary

This does not establish that live assessment is new, increasing or superior. Fifty of the 57 observations were coded as existing processes at the review date, which records what the page said, not when the practice began.

This does not establish prevalence. The count describes these 57 observations and nothing wider.

Source Global Elite Selection Monitor, Report 01: 57 signals drawn from 48 official sources, held in a 50-entry source register, across 23 institutions or systems. Checked 30 August 2026. Counted as beyond prose: live judgement, simulation, case or work simulation, multimodal or technical assessment, problem solving, gamified assessment, whole-person evaluation.

Exhibit 2: what the coded observations record about AI, and what they cannot show

Exhibit 2What the coded observations record about AI and selectionFifty of 57 observations were coded as existing processes at the review date. Six observations were coded newly explicit about AI. They come from three institutions.

Coded existing

Processes coded as already existing

50 of 57 observationsobservations coded as existing processes at the review date

Tests, cases, interviews and panels appear in the published processes reviewed. No Report 01 observation records when any of them was introduced.

Undated

No introduction dates are recorded

No Report 01 observation carries a date of introduction, so no sequence can be established for those 57. Report 02 does contain dated changes: NUS dates its Law test-before-interview rule to 2026 and its Dentistry sequencing to 2021.

This is a limitation of the register, not a finding.

Coded newly explicit about AI

Three institutions publish explicit AI boundaries

6 of 6observations coded newly explicit about AI, from three institutions

DLA Piper, UBS and Cambridge permit forms of preparation while protecting independent assessment and truthful representation.

Boundary

This does not establish that AI caused institutions to introduce live, interactive or whole-person assessment. This review does not establish when the Report 01 processes began or whether AI caused changes to them. A small number of Report 02 observations do carry dates, and they are stated where they appear.

Source Global Elite Selection Monitor, Report 01: 57 signals, checked 30 August 2026.

How it was built

Official sources, manual audit, published limits

107 coded selection signals69 official sources6 markets: UK, Gulf (GCC), India, US, Singapore, Hong Kong101 observations published after manual audit, 6 withheld

Research lead Hassan Akram, Founder and Principal Advisor, Elite Careers Strategy. Official public sources only. Evidence dated 30 and 31 August 2026, re-verified 9 September 2026. Full methodology

AccountabilityCoded by Hassan Akram against official primary sources. Source pages were retrieved and compared with the register with machine assistance, and every resulting change was reviewed before publication.

Full methodology

Official sources only

Every signal traces to one official employer or university page, with source ID, URL, grade and review date. No private client evidence, no modelling, no estimates.

Two reports, one register

Report 01 codes 57 signals. Report 02 codes 50 signals in a purposive six-market sample. Together: 107 signals citing 69 official sources, held in a 71-entry register.

Every observation audited

All 107 observations were read against their source on 9 September 2026. 80 are published as safe with source. 21 are published as context only, and barred from affirmative claims. 6 are held and excluded, because the source could not be independently retrieved.

What this research does not claim

It does not measure prevalence in any country or sector. Counts describe the sampled institutions only.
It does not establish that AI caused institutions to introduce live, interactive or whole-person assessment, or when any of these processes began.
It does not establish that any market is harder, fairer or more predictive than another. It compares; it does not rank.
It says nothing about selectivity, acceptance odds or candidate outcomes.
It does not imply affiliation with, or endorsement by, the institutions named. Invited student-community sessions are not institutional relationships.

Exhibit 4: sample composition and limits

Sample limits. The sample is authority-led and cross-sectional. It is weighted toward UK and global process pages, law and banking. Percentages describe the coded sample only.

Exhibit 4Law and banking account for 37 of the 57 coded signalsSignal composition by sector, Report 01, n = 57. The sample is authority-led, not representative.

Law21

Banking and finance16

Professional services8

University admissions6

Consulting5

Education and access1

Sector and market coded independently; one of each per signal.

Methodology and limits Source register Corrections

Reports and data

Read online, download, or check a figure

Start with Report 02, the six-market comparison. Below it: Report 01 on AI and the written application, the university companion, the full source register, and the raw evidence files.

Report 02
31 August 2026

Same Candidate, Different System

How published selection architecture differs across six markets. A purposive six-market sample of 50 signals, allocated UK 8, GCC 8, India 8, United States 9, Singapore 8, Hong Kong 9, from 21 official pages.

Read online Download PDF

50 signals
21 sources
6 markets

Report 01
Revised 31 August 2026

The Application as an Auditable Claim

What 57 elite-selection signals reveal about ownership, judgement and authenticity in the age of AI. Full source register: 48 cited sources in a 50-entry register.

Read online Download PDF Journalist briefing (PDF)

57 signals
48 cited sources
23 institutions

Companion

Elite University Selectivity

What thirty of the world's most highly ranked universities publish about admissions.

Read online Download PDF

30 institutions

Evidence

Global Elite Selection Monitor

Every coded signal and every official source, with source ID, URL, grade and review date. Published in full so that a figure can be checked rather than taken on trust.

Open the register Source register (PDF)

107 coded signals
69 cited sources
71 register entries

Data

Evidence files

The files behind the tables. The source evidence record; the cell audit, with the quoted passage for every matrix cell; the publication ruling for every observation; and the manifest tracing each published sentence to its observations.

Source evidence record (CSV)Cell audit (CSV)Publication rulings (CSV)Claim dependency manifest (CSV)Matrix chart (PNG)Matrix chart (SVG)

Open data

For journalists

Speak to the research lead

Mediahassan@accessecs.com
Interview or comment within three working days. Data checks same day.

Press kit

Hassan Akram

Founder and Principal Advisor, Elite Careers Strategy

Research lead for Who Gets In?. He is accountable for the coding decisions, publication rulings, source grading and corrections across the Global Elite Selection Monitor, and will take a data challenge directly.

Before founding ECS he worked on the recruitment side across corporate law, investment banking, private capital and consulting, and has reviewed more than 10,000 applications and candidate materials. He studied History at University College London.

Approved biography and quotations

No affiliation with or endorsement by the firms or institutions named is implied.

Approved quotations

Attribute to Hassan Akram, Founder and Principal Advisor, Elite Careers Strategy. Fresh quotes on request.

"AI is making polish abundant. Judgement is not."
On AI and the graduate application
"The evidence does not show that AI caused live assessment. It shows that live assessment was already there, and that AI has made its purpose harder to ignore."
On causation
"The information around elite professions is not secret, just unevenly distributed."
On access and information asymmetry