The Proxy Problem

When indicators become substitutes for evidence.

Who should lead the project?

Most leadership teams can answer within minutes. One candidate has led something similar before. Another has the strongest track record. A third has the right professional background and the confidence of senior colleagues. The discussion may be serious and the choice contested, but the evidence is familiar. Experience, reputation, past performance, credentials, and managerial judgment are assembled until one person appears to be the safest choice.

The same pattern governs decisions throughout the organization. Who should be promoted? Who is ready to work with artificial intelligence? Which function needs development? Where does succession risk sit? Even when the questions have strategic consequences, answers arrive with surprising speed.

It is worth pausing over that speed. What, precisely, has been established?

The candidate has led a project before. That is evidence of experience. The employee received a high rating. That is evidence of assessed performance. The professional holds a qualification. That is evidence that a defined standard was met. The workforce completed the program. That is evidence of participation and, perhaps, learning.

All are useful. None directly establishes that the person or organization can perform the work now under consideration.

Management depends on signals of this kind. No executive can personally observe every consequential judgment made across an enterprise. Organizations must compress complexity into information that can travel: titles, credentials, ratings, talent designations, course records, tenure, and career history. Without such representations, management at scale would become impossible.

The trouble begins when the representation and the reality acquire the same name.

A qualification becomes capability. Experience becomes readiness. Performance becomes potential. Completion becomes development. A competency framework becomes an account of the work. The substitution rarely feels reckless because each proxy has some relationship to what leaders want to know. Often the relationship is strong. Sometimes it is strong enough to support a sound decision.

But correlation is not identity. The distance between the two is where management risk accumulates.

Organizations have mistaken capability proxies for capability evidence.

The question each proxy answers

A proxy is not a falsehood. It is an indirect measure used when the thing that matters is difficult to observe. Its value depends on the stability of the relationship between the signal and the condition it is taken to represent.

Qualifications illustrate the point. A degree, license, or professional credential can provide essential assurance. It may establish that someone has learned a body of knowledge, passed a demanding examination, demonstrated practice under specified conditions, or satisfied a regulator. In many forms of work, no responsible organization would ignore it.

Yet every qualification has boundaries. It reflects a particular standard, assessed in a particular way, at a particular time. It cannot automatically establish how someone will apply knowledge when evidence is incomplete, priorities conflict, a novel exception appears, or the work has changed since the standard was designed. The qualification may be excellent evidence for the claim it was built to support. The mistake is quietly expanding that claim until it covers work the qualification never examined.

Experience has greater intuitive force because it was acquired in practice. Years in a role can develop pattern recognition, contextual knowledge, and judgment that formal instruction cannot reproduce. Experience often does predict performance, especially when the new work resembles the old.

But experience measures exposure before it measures learning. Five years may contain expanding responsibility, varied conditions, honest feedback, and repeated refinement of judgment. It may also contain one narrow year repeated five times. The duration alone cannot tell leaders which occurred.

More importantly, experience is historical. Its predictive value depends on continuity between the conditions that formed it and the conditions in which it will be used. When regulation alters the standard, technology redistributes the task, or strategy moves the organization into unfamiliar territory, experience remains real while its relevance changes. The people with the longest record may still be the right people to trust. Their tenure does not settle the question.

Performance appears closer to capability because it concerns results. Strong results should matter. An account of capability that disregarded performance would deserve little executive attention.

Results, however, are produced by more than capability. They also reflect the quality of systems, clarity of authority, availability of expert support, difficulty of assignments, market conditions, resources, incentives, and luck. A team may meet every target because one highly capable employee quietly catches the exceptions. Another may struggle because capable people are working through a badly designed process. The first result can conceal fragility. The second can conceal strength.

Performance ratings add further interpretation. They combine outcomes, behavior, expectations, opportunity, and managerial judgment inside an administrative process. The rating may be entirely defensible in its original context. Once aggregated into succession plans or workforce analytics, much of that context disappears. Numbers that share a scale begin to appear comparable even when the work, standards, and raters were not.

Completed learning sits farther away still. It shows that an intervention occurred. Depending on the design, it may also show recall, understanding, or performance in a simulation. It does not show that the work now permits application, that managers reinforce the practice, that decision rights support it, or that performance has become more reliable. Completion is evidence about learning activity. It becomes evidence of capability only when something connects it to changed performance in the work.

Competency frameworks can organize these signals and create a common language. Their weakness appears when broad labels detach from specific work. “Strategic thinking,” “AI literacy,” or “collaboration” may sound precise because the organization has defined levels beneath them. Yet a level cannot establish readiness until it is related to the actual judgments, standards, and conditions the work demands. Greater taxonomic detail does not repair a missing connection.

Organizations therefore possess a great deal of evidence about people and surprisingly little evidence of capability for work.

When the relationship breaks

In stable settings, this distinction can seem academic. Experienced managers learn which credentials matter, which assignments develop judgment, who can be trusted with an exception, and which ratings should be treated cautiously. Their local knowledge corrects for weaknesses in the formal system. The proxies work because the work changes slowly enough for their relationships to remain dependable.

Change exposes the arrangement.

Across industries, recurring patterns in role-based capability evidence show experienced people struggling when familiar work is redesigned, while apparently junior employees demonstrate exceptional judgment in unfamiliar situations. In manufacturing and transport, the person with the strongest technical history is not always the person best able to interpret new forms of operational evidence. In financial services and government, formal knowledge of policy does not necessarily predict judgment when a novel case falls between established categories. In healthcare and professional services, qualifications establish indispensable foundations but do not reveal who can translate expertise across boundaries when consequences are shared.

These are not arguments against experience, seniority, or professional standards. They reveal that capability belongs first to the work. Until the work is understood, there is no stable object against which any claim about supply can be tested.

This reverses the sequence used in most workforce decisions. Organizations usually begin with visible supply: the people employed, credentials held, skills declared, ratings awarded, courses completed, and roles occupied. They then infer what the enterprise should be able to do. But capability demand precedes capability supply. The work creates the requirement. Only then can leaders determine whether available knowledge, judgment, experience, relationships, technology, and authority are sufficient to meet it.

Without that sequence, even accurate information produces unreliable conclusions. A workforce may contain deep analytical expertise while lacking the ability to apply analysis at the point where operational decisions occur. A function may report strong AI readiness because specialists are highly qualified and most employees have completed foundational learning. Yet the consequential work may depend on frontline professionals recognizing when a plausible model output is incomplete, biased, or unsafe. Neither specialist credentials nor enterprise completion rates demonstrate that capability where it is required.

AI is particularly unforgiving of proxy logic because it changes the content of work without necessarily changing the role title. A professional who previously produced an analysis may now need to interrogate one. A manager who reviewed employee recommendations may need to govern recommendations assembled by people and machines. Expertise in the original task remains valuable, but the critical judgment may have moved. Yesterday’s evidence can retain its prestige after losing its fit.

Hierarchy creates the same distortion. Formal level is often treated as a shorthand for capability because authority and career progression usually reflect accumulated judgment. Yet recurring evidence across operational roles shows that consequential capability is widely distributed. A relatively junior employee may be the last person able to detect a safety deviation, protect a decision trail, challenge an automated output, or escalate a customer risk. The organizational chart records authority. It does not show where the work places consequence.

The proxy problem is therefore not confined to Human Resources. It shapes capital allocation, transformation governance, technology adoption, risk, operating model design, and strategy. Whenever a business case assumes that a qualified workforce can operate a new model, that experienced managers can absorb a change, or that completed learning has produced readiness, a workforce signal has become a strategic assumption.

Evidence fit for the decision

Direct capability evidence does not mean observing every employee perform every possible task. Nor does it require replacing several imperfect signals with one supposedly definitive score. Capability is contextual, and any system claiming to measure it without reference to work merely creates a more sophisticated proxy.

Evidence becomes useful when it compares what the work demands with what the organization can bring to that work on comparable terms. A relevant work sample may reveal application that a qualification cannot. Observation under representative conditions may show judgment that tenure cannot. Decision records, exception handling, escalation patterns, and the rationales people create may reveal how capability operates in practice. Performance can contribute when the conditions that produced it are understood. Qualifications, experience, learning, self-assessment, and managerial judgment remain valuable sources when the limits of each are preserved.

The aim is not certainty. It is decision-grade evidence: evidence strong enough to distinguish among materially different choices and honest enough to show where an important claim remains untested.

This standard should rise with consequence. A manager assigning a low-risk piece of work can reasonably rely on observation, experience, and judgment. A board approving a major transformation, a leader promoting someone into a role with enterprise-wide consequences, or an organization introducing AI into high-stakes decisions should demand more. The question is not whether proxies may be used. It is whether they can bear the weight of the decision placed upon them.

That distinction changes the meaning of familiar information. A qualification becomes one source, not a conclusion. Experience becomes relevant history, not proof of future fit. A performance rating becomes a prompt for inquiry, not a portable capability score. Learning completion becomes evidence that an opportunity was provided, not that organizational capability improved.

Perhaps organizations have spent decades measuring almost everything surrounding capability while rarely measuring capability itself. They can report how many qualified people they employ, how long those people have served, how well they performed, what learning they completed, and where they sit in the hierarchy. Still unanswered is the question on which the decision depends: can the required work be performed, at the necessary standard, under the conditions that now apply?

The problem is not that organizations use proxies. The problem is that they mistake proxies for evidence.

The more important the decision, the less comfortable leaders should become relying solely on proxies.

Executive reflection questions

1. Which consequential decisions in our organization currently rely on qualifications, experience, ratings, or completion as if they directly demonstrated capability?

2. Where has the work changed enough that our most trusted capability proxies may no longer predict what we need to know?

3. What evidence would allow us to judge capability against the actual demands and conditions of the work before the decision becomes difficult to reverse?