What the Org Chart Can't See

Twenty surgeons agreed to be filmed. All were board-certified, all with years of practice behind them. By every record a hospital kept, they were the same surgeon twenty times over.

Then their peers watched the recordings, blind to who was holding the instruments, and rated the surgery itself. Those ratings were linked to the outcomes of 10,343 patients. Among surgeons in the bottom quartile for technical skill, complications occurred in 14.5% of cases. Among those in the top quartile, the rate was 5.2%. Operative mortality ran 0.26% against 0.05%. A patient in the lower-skilled surgeon's operating room was almost three times as likely to suffer a complication. [1]

The study, published in the New England Journal of Medicine, exposes a distinction most organizations blur. A credential tells you that someone cleared a threshold. It does not tell you how the people who cleared it differ from one another.

The follow-up made the lesson sharper. Surgical skill predicted complications, but it didn't predict patients' weight loss or the resolution of their conditions a year later. Those depended more on the surrounding clinical program. [2] Skill decides what skill controls, and no more. Even so, on the outcomes it did control, nothing separating the best surgeons from the rest was visible on a resume.

That's the management problem in miniature. Organizations know who reports to whom, what title a person holds, what training they finished, and whether their calendar looks full. They know far less about what those people actually do unusually well, whether that edge is current, which situations bring it out, what work they care about, and how their contribution changes when it's combined with other people's.

And those are the differences that decide the work that matters.

The Org Chart Is an Administrative Map

An org chart is useful. It shows how authority is distributed, where budgets sit, and who owns a function. A resource plan is useful too. It shows how much capacity appears to be free and where commitments have been made.

Neither was built to answer the harder question:

Who is most likely to make a distinctive contribution to this particular piece of work, under these conditions, at this moment?

In drug development, that question is often answered on a Friday afternoon. A study needs a lead. A functional head opens a resourcing view, scans the names she knows, checks who looks available, and makes the assignment. The decision is reasonable given what she can see. It's also mostly gut feel wearing the costume of a plan.

The title stands in for capability. Years in role stand in for expertise. Utilization stands in for readiness. Familiarity stands in for fit. The most consequential science on earth is staffed roughly the way a shift gets filled.

The managers aren't careless. The organization has simply made each person legible at the wrong resolution. The chart can see a position. The work needs a contribution.

That gap widens as work grows complex. A classic analysis estimated the standard deviation of individual output at about 19% of mean output in unskilled and semiskilled work, 32% in skilled work, and 48% in managerial and professional work. [3] The figure isn't a law, and it doesn't mean any two professionals differ by exactly that much. What it does establish is the direction: the more complex the work, the wider the spread in what people actually produce.

This is where careful matching earns its cost. If almost anyone can do a task to the same standard, elaborate matching is wasted motion. Where judgment, knowledge, pattern recognition, and coordination move the result, choosing the right person becomes part of the work itself.

Contribution Is a Relationship, Not a Trait

Organizations usually store capability as something a person owns: "regulatory strategy," "oncology," "program leadership," "machine learning." The label goes in a profile, often with a proficiency score, and the system treats it as durable.

Contribution does not sit still inside a profile. It emerges from the relationship among a person, a piece of work, a team, and a moment. A strong match rests on five drivers.

Capability: has the person demonstrated what this work requires?

The best evidence of capability lives in closely related work, not in a broad portrait of the person.

A major 2022 reanalysis of personnel-selection research revised several widely cited validity estimates downward. What held their value were the methods closest to the work itself: structured interviews built on real behavior, job-knowledge tests, and work samples. [4] The exact ranking of methods is still debated, but the managerial lesson is durable. What we measure to sort people in the abstract tends to miss what actually sorts them: the specifics of what they've done.

Raw tenure is a weak guide for the same reason. A meta-analysis of 81 samples found that prehire experience, measured mostly as duration, correlated .06 with later job performance and roughly zero with turnover. Even role-relevant experience stayed weak when it was counted in years. [5] That doesn't mean experience is unimportant. It means elapsed time is a poor way to show it. Ten years of repeating one familiar assignment is not ten years of varied conditions, hard decisions, honest feedback, and recovered failures. The question isn't how long someone has worked. It's what they've actually faced, decided, built, and learned.

The top and bottom quartiles averaged the same eleven years. The tapes told a different story.

Currency: is the capability ready to use now?

A person can hold the right knowledge and still be the wrong assignment today.

Skills decay when they go unused, though the rate varies with the type of skill, the depth of original mastery, the interval, and the chances to practice. Meta-analytic evidence shows meaningful loss after stretches of nonuse, and cognitively demanding, accuracy-dependent skills are often the most vulnerable. [6] [7] A regulatory strategist who handled several hard agency interactions five years ago still carries judgment worth having, but she may not be as ready for this assignment as someone who closed a closely analogous negotiation last quarter. The credential stays frozen in the system. Readiness does not.

Currency also covers load, recovery, and attention. Research on within-person performance warns against treating anyone as a fixed score. Performance varies meaningfully in the same person over time, and patterns seen across people don't always hold inside one person from week to week. [8] [9] Availability, then, isn't an empty calendar slot. It's the capacity to bring the relevant capability to bear now.

Context: does the situation bring the strength out?

A strength doesn't act on its own. Trait Activation Theory holds that a characteristic becomes behavior only when the situation supplies the cues, demands, or openings that make it relevant. [10] The same meticulous scientist is invaluable when the work rewards catching subtle inconsistencies and a liability when it demands fast experimentation under ambiguity.

The predictive power of vocational interests makes the point. Matched to the specific type of work, interest measures related to performance far more strongly than when they were applied across every job the same way. [11] The person did not change. The precision of the question did.

Context is the domain, the stakes, the ambiguity, the technology, the stakeholders, the cost of error. "Regulatory writer" is a role. Preparing a first-in-class submission after conflicting agency feedback is a situation. The second is where fit becomes visible.

Commitment: does the person want this particular work?

Capability decides whether someone can perform. Commitment shapes the attention and the effort they choose to give it.

A 40-year meta-analysis found intrinsic motivation more strongly associated with the quality of performance, and external incentives more strongly associated with its quantity. [12] That difference matters most in work whose value comes from noticing what nobody asked for, challenging an assumption everyone else accepted, or staying with an anomaly long enough to understand it.

Commitment is task-specific. A scientist may be lit up by one molecule and indifferent to the next. A senior expert can be bored by work she performs flawlessly while an adjacent learner is hungry to master it. None of this makes desire a substitute for competence. It makes desire useful evidence among the people who can credibly do the work, and a strong signal when the assignment can be built as a supervised stretch. Recorded as a permanent verdict, motivated or not, it's crude and unfair. Asked about a specific assignment, do you want this, and why, it becomes something a staffing decision can use.

Configuration: does the person strengthen this team?

Even perfect knowledge of individuals wouldn't finish the job, because the work is rarely done alone.

Team-composition research shows the strongest set of individuals does not automatically make the strongest team. Performance turns on coverage of the essential capabilities, complementary styles, minimum levels on a few interpersonal traits, and the outsized weight of certain roles. [13] One of the most useful ideas here is transactive memory: a shared sense of who knows what, whose expertise to trust, and how to route it. In a study of 121 trauma resuscitations, a one-standard-deviation increase in the strength of that shared map was associated with about 3.3 fewer hospital days and 1.9 fewer days in intensive care, after accounting for injury severity and team size. [14]

The lesson is operational. Expertise that can't be found, trusted, and routed is, for practical purposes, expertise the organization doesn't have. So contribution is less the highest individual score than the value a person creates in a particular configuration.

Why Organizations Learned to Manage Proxies

None of these findings is new. The real story is why organizations couldn't act on them.

To make a better assignment, a large enterprise would need to know, across thousands of people: what each has actually done, the complexity and context of that work, which outcomes their contribution plausibly moved, how recent the evidence is, what they want next, who trusts their judgment, and how they work alongside others. Most of that has lived in scattered work products, meeting histories, manager memory, project records, and the heads of colleagues. Capturing and maintaining it by hand would cost more than most assignments could justify.

So organizations managed the variables they could count. Headcount. Role. Level. Utilization. Location. Tenure. These aren't useless; they're administrative stand-ins. Over time the stand-in quietly became the reality. A better planning system refined the proxies. A reorganization redrew their boxes. Neither made the underlying contribution much easier to see.

This is why the problem gets filed under talent management when it belongs elsewhere. It's closer to an operating-system problem: an organization can't route capability it can't perceive.

What AI Changes, and What It Does Not

AI changes the economics of perception.

Work already leaves traces: decisions, analyses, designs, code, experiments, reviews, exceptions, handoffs, outcomes, and the people others keep consulting. Modern systems can help extract and connect those traces, reconcile different language for related capabilities, find analogous experience, attach recency, and make scattered evidence searchable. Instead of asking only who holds the title, an organization can begin to ask who has solved a closely analogous problem, under what conditions they did it well, how current the evidence is, who else must be in the room for the contribution to work, which credible candidates actually want the assignment, and what support would make a developmental choice viable.

That's a real advance. It is not omniscience.

AI cannot reliably decide who caused a team outcome by reading the record. It can't infer private motivation safely from behavior alone. And it cannot fully capture tacit judgment, tell visibility from value on its own, or settle a person's future potential. The system should surface hypotheses about fit, not verdicts about people. Its job isn't to replace a manager's judgment but to improve the evidence that judgment runs on, widen the field past the names one manager happens to know, and make the assumptions explicit enough to argue with.

This faculty has a name: Contribution Intelligence, the capacity to perceive the distinctive contribution each person can make and the conditions under which it is most likely to appear. It is the first faculty of Organizational Intelligence.

Organizational Intelligence is the larger capacity for an organization to make itself visible to itself and to move as one. It has three faculties. The first is the one this piece has described: perceiving who can contribute what, where, and now. The other two coordinate that contribution across functions, so the enterprise acts on one reality, and preserve the reasoning behind its decisions, so judgment compounds instead of leaving with the person who held it. Perceive, coordinate, learn. Together they let an organization adapt as its world moves, which is the whole point of building them.

Building Contribution Intelligence

A leader doesn't need to model every person and task at maximum resolution. That builds bureaucracy before it builds intelligence. Start where fit has consequence: work that is complex, costly, hard to reverse, dependent on scarce judgment, or likely to bottleneck the portfolio. Five practices make the idea real.

Describe the work at the resolution where fit matters. "Program lead" is too broad; "send the follow-up email" is too narrow. A useful description names the outcome, the judgment required, the context, the deadline, the cost of error, and the other people involved. The right level is the point at which a different person would change the result.

Build evidence from work, not profiles alone. Use self-declared skills and aspirations, then go past them, to work products, decisions owned, problems solved, analogous situations handled, peer observation, outcomes, and improvement over time. Every capability claim should show where it comes from: what supports it, in what context, how recently, and with how much certainty. "Excellent at regulatory strategy" is a label. "Led four recent FDA response strategies for late-stage biologics involving ambiguous clinical feedback, strongest in written response design and cross-functional preparation, limited recent EMA evidence" is matchable.

Recompute the match as people and work change. A staffing decision is a hypothesis, not a permanent truth. The task shifts after a regulatory signal, a competitor move, a protocol amendment, a supply disruption. The person gains experience, loses currency, becomes overloaded, or develops a new interest. A match made six months ago can be wrong while every box on the plan stays green. Revisit the big assignments when the work meaningfully changes, not just at annual planning.

Evaluate the configuration, not just the individual. Ask which capabilities must be present somewhere on the team, which styles complement one another, where continuity is worth protecting, and which relationships carry the information and the trust. The aim isn't a roster of stars. It's a configuration that can produce the outcome.

Learn from the assignment without pretending to prove causality. Afterward, capture what happened: whether the outcome met the standard, what the person contributed, what helped or constrained them, what new capability showed itself, which inference was wrong, and whether the evidence should strengthen, expire, or stay uncertain. That closes the loop between perceiving people and learning as an organization, and it keeps the capability map from hardening into a set of permanent reputations.

Seeing People Without Reducing Them

The strongest objection to Contribution Intelligence is also the reason to design it carefully.

A system that keeps routing important work to people who already have evidence gives them still more chances to generate it. People who haven't been chosen look less capable, because they've been denied the very work through which capability becomes visible. Perception can quietly become caste. The scarcest resource in the organization is often not talent. It is the chance to prove it.

So a responsible system carries two budgets. An execution budget, spent when failure is costly and the strongest evidence should win. And an exploration budget, spent on bounded, reversible work to test adjacent capability, grow the next experts, and surface the people the record has overlooked. It also needs plain safeguards: people can see and correct material claims about themselves, evidence is time-stamped and allowed to expire, inference is kept distinct from observed fact, private and irrelevant data stays out, developmental evidence isn't quietly repurposed to punish, the high-stakes assignments get human review, and uncertainty stays visible.

No method delivers certainty. Even the strongest assessments leave much of performance unexplained. The goal isn't to calculate a person or stamp a permanent score. It's to be systematically less wrong than title, proximity, familiarity, and whoever happens to be free. The last era of management made work legible by standardizing the worker. This does the opposite. It looks for where a person's particular experience, judgment, interests, and relationships make a difference, without mistaking the map for the person.

The First Faculty of an Intelligent Organization

For most of organizational history, one of the forces that shaped outcomes the most, who was actually doing the work, was also one of the least visible. So the organization guessed, and called the guessing a plan.

It doesn't have to guess to the same degree anymore. An intelligent organization does not know its people merely because it employs them. It knows them when it can recognize the distinctive contribution each is capable of making, under the conditions in which that contribution matters.

That is Contribution Intelligence, the first of the three faculties an intelligent organization needs: to perceive its people. Coordinating them and learning from them come next. In life sciences the promise is concrete: better odds on each program, months of avoidable delay taken out, and medicines that reach patients sooner.

‍

References

[1] Birkmeyer, J. D., et al. "Surgical Skill and Complication Rates after Bariatric Surgery." New England Journal of Medicine 369, no. 15 (2013): 1434–1442. https://doi.org/10.1056/NEJMsa1300625

[2] Scally, C. P., et al. "Video Ratings of Surgical Skill and Late Outcomes of Bariatric Surgery." JAMA Surgery 151, no. 6 (2016): e160428. https://doi.org/10.1001/jamasurg.2016.0428

[3] Hunter, J. E., Schmidt, F. L., and Judiesch, M. K. "Individual Differences in Output Variability as a Function of Job Complexity." Journal of Applied Psychology 75, no. 1 (1990): 28–42. https://doi.org/10.1037/0021-9010.75.1.28

[4] Sackett, P. R., Zhang, C., Berry, C. M., and Lievens, F. "Revisiting Meta-Analytic Estimates of Validity in Personnel Selection: Addressing Systematic Overcorrection for Restriction of Range." Journal of Applied Psychology 107, no. 11 (2022): 2040–2068. https://doi.org/10.1037/apl0000994

[5] Van Iddekinge, C. H., et al. "A Meta-Analysis of the Criterion-Related Validity of Prehire Work Experience." Personnel Psychology 72, no. 4 (2019): 571–598. https://doi.org/10.1111/peps.12335

[6] Arthur, W., Bennett, W., Stanush, P. L., and McNelly, T. L. "Factors That Influence Skill Decay and Retention: A Quantitative Review and Analysis." Human Performance 11, no. 1 (1998): 57–101. https://doi.org/10.1207/s15327043hup1101_3

[7] Tatel, C. E., et al. "Procedural Skill Retention and Decay: A Meta-Analytic Review." Psychological Bulletin (2025). https://doi.org/10.1037/bul0000481

[8] Dalal, R. S., Bhave, D. P., and Fiset, J. "Within-Person Variability in Job Performance: A Theoretical Review and Research Agenda." Journal of Management 40, no. 5 (2014): 1396–1436. https://doi.org/10.1177/0149206314532691

[9] Dalal, R. S., Alaybek, B., and Lievens, F. "Within-Person Job Performance Variability over Short Timeframes: Theory, Empirical Research, and Practice." Annual Review of Organizational Psychology and Organizational Behavior 7 (2020): 421–449. https://doi.org/10.1146/annurev-orgpsych-012119-045350

[10] Tett, R. P., Simonet, D. V., Walser, B., and Brown, C. "Trait Activation Theory: A Review of the Literature and Applications to Five Lines of Personality Dynamics Research." Annual Review of Organizational Psychology and Organizational Behavior 8 (2021): 199–233. https://doi.org/10.1146/annurev-orgpsych-012420-062228

[11] Nye, C. D., Su, R., Rounds, J., and Drasgow, F. "Vocational Interests and Performance: A Quantitative Summary of Over 60 Years of Research." Perspectives on Psychological Science 7, no. 4 (2012): 384–403. https://doi.org/10.1177/1745691612449021

[12] Cerasoli, C. P., Nicklin, J. M., and Ford, M. T. "Intrinsic Motivation and Extrinsic Incentives Jointly Predict Performance: A 40-Year Meta-Analysis." Psychological Bulletin 140, no. 4 (2014): 980–1008. https://doi.org/10.1037/a0035661

[13] Bell, S. T., Brown, S. G., Outland, N. B., and Abben, D. R. Critical Team Composition Issues for Long-Distance and Long-Duration Space Exploration. NASA/TM-2015-218568 (2015). https://ntrs.nasa.gov/citations/20140016953

[14] Argote, L., et al. "Transactive Memory Systems and Hospital Trauma Team Performance." Organization Science (2025). https://doi.org/10.1287/orsc.2024.19022

Further Reading

• Organizational Intelligence: The Enduring Advantage When AI Is Everywhere. The manifesto for the larger thesis: perceive, coordinate, learn.
• Headcount Is Not Capability. Why adding people does not add capability.
• People Are Not Your Greatest Asset. How allocation breaks when fit gives way to availability.
• Estimating Is Not Calculating. The difference between role-level estimation and task-level calculation.
• Capability Tells You Who Can. Desire Tells You Who Cares. Why task-specific desire changes how work gets done.

‍