The market has more information about people than ever. It understands them less.
AI is raising the floor of professional performance while weakening many of the signals historically used to distinguish exceptional people. We think that changes how talent needs to be understood.
Thesis in brief
AI is improving professional output faster than the market is improving its understanding of people.
As familiar signals become easier to optimise, richer evidence, context and trusted judgment become more valuable.
Academic research, substantial datasets and large institutional surveys. This carries the greatest evidentiary weight.
What credible practitioners report seeing. These perspectives bring practical experience to the thesis and can identify mechanisms that empirical research may not yet fully capture.
Our synthesis of what the evidence may imply. Hypotheses are identified as hypotheses, not presented as settled fact.
AI is raising the floor.
Empirical evidenceGenerative AI can materially improve performance across many knowledge-work tasks. It can help professionals write, research, code, structure analysis, synthesise information and produce credible output faster than before.
In a field study covering 5,179 customer-support agents, researchers found that access to a generative-AI assistant increased productivity by nearly 14% on average. Gains were largest among novice and lower-skilled workers, consistent with AI helping diffuse the practices of stronger performers.1
Customer-support agents using a generative-AI assistant.
NBER field research1Consultants working on tasks inside the model’s capability frontier.
HBS / BCG experiment2Human-rated performance in the HBS field experiment’s AI-suited tasks.
HBS / BCG experiment2The BCG experiment reached a similarly consequential conclusion. Across 758 consultants, GPT-4 users completed more tasks, worked faster and produced higher-rated work on tasks that sat within the technology’s capabilities. But performance deteriorated on a task outside that frontier, where plausible AI analysis obscured a critical insight.2
Another controlled experiment found that ChatGPT reduced the time professionals spent on mid-level writing tasks by 40% and increased output quality by 18%. The distribution of outcomes compressed: weaker participants benefited more.3
AI materially increases the capabilities available to an average professional. The evidence does not show that exceptional performers necessarily receive the largest percentage gain. It suggests something subtler: baseline execution can converge while already capable people gain more surface area across which to deploy scarce judgment.
AI raises the floor. It does not necessarily eliminate the ceiling.
At the same time, the signal is deteriorating.
A labour market cannot observe capability directly. It relies on signals: the CV, employer, title, qualification, cover letter, stated skills, written communication and interview performance.
These signals remain informative. But generative AI dramatically reduces the cost of optimising many of them. A mediocre candidate can produce an excellent-looking application. An exceptional candidate can produce an excellent-looking application. The visible distinction between them becomes weaker.
Korn Ferry describes this directly: AI has “commoditized résumés, cover letters, and even interview answers,” producing repeated language and similar claims across candidates.7 Its 2026 talent research also found that 73% of surveyed talent-acquisition leaders placed critical thinking and problem-solving above AI skills as their most important recruiting priority.8
This does not make a CV useless. It changes what can reasonably be inferred from polish, tailoring and fluency alone.
Visible competence is converging faster than underlying capability.
When applications become cheap, volume increases.
Empirical evidenceA considered application historically required meaningful effort. Candidates researched the organisation, reworked a CV, drafted a letter and made a deliberate decision to submit. AI and automation reduce that marginal cost.
Candidates can identify more opportunities, tailor applications faster, optimise keywords, automate outreach and increasingly automate submission. That mechanism predicts more top-of-funnel activity. The available market data are consistent with it, although AI is not the only cause: labour-market conditions, remote work and application design also matter.
Ashby’s analysis of more than 54 million applications and 93,000 jobs similarly describes a high-volume baseline, with applications per hire having tripled between 2021 and 2024.5 In the UK graduate market, 65% of employers reported increased application volumes in 2025.6
“There’s too much noise, and it’s harder to find the right signal.”
Sphere CEO & CFO Nicholas Rudder has described how AI-lengthened cover letters, poorly personalised emails and outreach sent at automated scale can create substantial employer-side noise. His perspective gives practical expression to the mechanism visible in the broader volume data: when outreach becomes cheaper to produce, identifying the right signal can become harder.15
More noise encourages simpler filters.
If organisations receive more applications without proportionally increasing human-review capacity, they must process them somehow. ATS filtering, keyword matching, AI screening, title and industry matching, credential filters, recruiter heuristics and rapid CV review are rational operational responses.
Greenhouse reports that applications per role increased while recruiter headcount per organisation fell between 2022 and 2025.4 Korn Ferry’s 2025 research found that 40% of talent specialists worried that excessive use of AI in recruitment could make the process impersonal and cause organisations to miss strong candidates.9
A proposed mechanism, not an empirically proven universal sequence.
- 01AI lowers application cost
- 02Application volume increases
- 03Signal-to-noise deteriorates
- 04Screening becomes more automated or heuristic
- 05Exact-match signals become more valuable
- 06Non-linear candidates become harder to interpret
- 07Potential human alpha is missed
As the information problem becomes harder, the economically rational response may be to simplify the decision. But the easiest candidate to identify then becomes the candidate who most closely resembles the expected candidate.
Pattern matching identifies familiarity extremely well.
That is not necessarily the same as identifying exceptional capability.
Recruiters use patterns for defensible reasons: limited time, incomplete information, mandate requirements, reputational risk, hiring-manager expectations and the need to produce a coherent shortlist.
Common patterns include the same title, industry, employer type, qualification and career trajectory. Each can carry information. The structural risk lies in treating the pattern as the capability rather than as one imperfect proxy for it.
The market frequently asks people with limited time and incomplete information to infer high-resolution human capability from low-resolution professional data.
The most sophisticated search firms already know this.
The methodologies published by major executive-search firms do not treat a résumé as a complete account of a leader. They combine experience with capability, potential, context, judgment and structured assessment. That validates the information problem without implying that any firm has solved it perfectly.
Experience is important—and incomplete.
Its assessment work distinguishes present capability, relevant context and future potential. Its CEO research also cautions against assuming that prior CEO experience will transfer mechanically to a different mandate.1011
AI amplifies the importance of judgment.
Its leadership work argues that AI can accelerate aspects of hiring but cannot determine potential, assess character under pressure or understand career inflection points on its own.12
Polished presentation is becoming less discriminating.
Its 2026 research describes commoditised application materials while its talent survey places critical thinking and problem-solving at the centre of AI-era capability.78
The mandate must be mapped to outcomes.
Its published search approach begins with business outcomes, then uses structured evaluation to assess candidates against the role rather than a generic job description.13
Looking good on paper is not the same as thriving.
Its Leadership Portrait combines relevant experience and competencies with a multi-dimensional view of readiness and potential in a specific environment.14
Executive search already knows that the résumé isn’t enough.
Our question is why deep understanding of a person should begin only when someone opens a search mandate.
A title tells you what someone was called. Context tells you what they actually experienced.
Market observationExecutive-search practitioner Jack Syred has helped articulate why “PE-backed experience” is an incomplete signal unless the underlying environment is examined. His perspective draws attention to the materially different mandates, pressures and forms of learning that can sit beneath an identical label—and provides a practical bridge from the market thesis to Sandery’s emphasis on career context.17
Consider two people whose CVs both say CFO — PE-backed business.
- First institutional investment
- Professionalisation
- Rapid buy-and-build
- Exit preparation
- Covenant pressure
- Turnaround
- Management change
- Liquidity preservation
The titles look identical. The accumulated experience is not. Understanding transferability requires situational experience, operating environment, complexity, responsibility, decisions, transitions, outcomes, relationships and what the person demonstrably learned to do.
What if the system is becoming better at identifying conformity?
Repeated reliance on familiar professional signals can produce homogeneous candidate pools—not because recruiters are indifferent to difference, but because familiarity is easier to recognise, explain and defend.
Investor and adviser Lee McCabe has been a deliberately provocative voice on the tendency of private equity to recruit from similar educational and professional pipelines. He argues that repeated reliance on familiar profiles can reinforce cognitive similarity and narrow the field of view. His perspective has helped sharpen an important question within the Sandery thesis: when recognisability becomes a proxy for quality, which forms of differentiated capability does the market fail to see?16
Pattern recognition is very effective at finding people who resemble previous successful candidates. But what happens to people whose capability is real and transferable, but whose trajectory is different?
Human Alpha
Human alpha is the differentiated value created by an individual’s particular combination of judgment, experience, adaptability, relationships, pattern recognition, capability and decision quality beyond baseline professional competence.
Human alpha does not mean that a person can—or should—be reduced to a single score. It is a conceptual framework for thinking about differentiated human capability.
AI increasingly commoditises certain forms of competent execution. It does not necessarily commoditise judgment, problem selection, contextual understanding, leadership, commercial instinct, relationships, adaptability, accountability or accumulated experience.
Exceptional professionals can potentially use AI to deploy those scarce capabilities across a larger amount of work. AI can therefore make average talent better while simultaneously increasing the economic importance of finding exceptional talent.
The paradox is that AI may make exceptional talent more valuable at exactly the moment traditional signals make it harder to identify.
More data is not the answer.
The market already has enormous amounts of professional data: LinkedIn profiles, CV databases, ATS records, recruitment CRMs, HRIS systems, skills taxonomies and assessment results.
The problem is not simply absence. It is fragmentation, shallowness, limited context, uncertain provenance, static representation, transaction-specific collection and insufficient longitudinal understanding.
tells you where somebody worked.
helps explain what they actually did, what they demonstrated and where that capability might transfer.
Financial markets built intelligence infrastructure around assets.
We think people deserve something better too.
Financial markets progressively moved from fragmented documents and raw information towards structured data, longitudinal histories, comparable information, research, analytics, contextual intelligence and decision-support systems.
Professional markets remain comparatively dependent on fragmented representations of people. The analogy is directional, not literal: better information can improve the quality of judgment without pretending to replace it.
Sandery’s ambition is to make professional information more useful by placing the individual—and the substance of their experience—at the centre of the analysis.
The Sandery Thesis
Research note 01Better decisions about people require better information about people.
AI is making competent professional output cheaper. It is making candidate presentation easier to optimise. It is increasing the amount of information flowing through talent markets while potentially reducing the discriminatory value of many traditional signals.
The natural response is greater automation and greater reliance on recognisable patterns. But people are not patterns.
A career is a longitudinal collection of experiences, decisions, achievements, relationships, environments and accumulated judgment. As AI raises baseline capability, those differentiated human qualities may become more economically important. We call that difference human alpha. The challenge is identifying it.
Sandery is being built around a deeper understanding of professional experience—not simply more applications, more profiles or better keywords.
The market has spent decades building sophisticated intelligence infrastructure around companies and financial assets. We think there is an opportunity to build something analogous around people.
Continue the thesis
Explore what Sandery is building.
References
Sources and evidentiary notes
External claims link to the original source wherever available. Practitioner perspectives are presented separately so their contribution to developing and illustrating the thesis remains clear.
Research & empirical evidence
- Generative AI at WorkErik Brynjolfsson, Danielle Li and Lindsey RaymondNBER Working Paper 31161, 2023 (revised 2025) ↗
- Navigating the Jagged Technological FrontierFabrizio Dell’Acqua et al.Harvard Business School Working Paper 24-013, 2023 ↗
- Experimental Evidence on the Productivity Effects of Generative Artificial IntelligenceShakked Noy and Whitney ZhangScience, Vol. 381, 2023 ↗
Industry research
- The Hire Standard: 2026 Hiring Benchmark ReportGreenhouseApplications per role, recruiter capacity and time-to-fill benchmarks ↗
- Recruiting Operations BenchmarksAshbyAnalysis of more than 54 million applications and 93,000 jobs, 2026 ↗
- The Application ExplosionInstitute of Student EmployersEmployer survey on graduate application volumes, 2026 ↗
- When All Candidates Sound the SameKorn FerryBriefings Magazine, March 2026 ↗
- Talent Acquisition Trends 2026: Human–AI Power CoupleKorn FerrySurvey of more than 1,600 talent leaders ↗
- Talent Acquisition Trends Report 2025Korn FerryTalent-acquisition survey and practitioner research ↗
- Why Effective Executive Development Begins with AssessmentSpencer StuartCapabilities, context, potential and multi-method assessment ↗
- Predicting CEO Success: When Potential Outperforms ExperienceSpencer StuartResearch on first-time and experienced CEOs ↗
- Human Judgment: Powering Leadership in the Age of AIEgon ZehnderLeadership and talent perspective, 2026 ↗
- Executive Search: Technology & AI OfficersHeidrick & StrugglesMandate-to-outcomes and structured-evaluation methodology ↗
- Executive Assessment and BenchmarkingRussell Reynolds AssociatesMulti-dimensional assessment of experience, readiness and potential ↗
Practitioner observations
- Employer observation on AI-enabled application and outreach noiseNicholas RudderCEO & CFO, Sphere — LinkedIn, August 2026 ↗
- Practitioner observations on cognitive similarity in private equityLee McCabeFounder, Claymore Partners — LinkedIn, July 2026 ↗
- Beyond ‘PE-backed experience’: what the environment demandedJack SyredPartner, Wilton & Bain — LinkedIn, April 2026 ↗
This thesis presents Sandery’s interpretation of the evidence and practitioner perspectives cited. It is a framework for understanding structural changes in professional markets, not a claim that every mechanism operates uniformly across roles, organisations or labour markets.