Chapter 21: The Vocabulary of Skills

Home • Index  • ← Previous • Next → • Browse by Topic

The executive dashboard showed a striking number.

Sixty-three percent of the workforce had "Python" listed as a skill.

Leadership found this encouraging, given the organization's stated ambition to build stronger internal data and AI capability. Perhaps the talent existed already, waiting to be better deployed, rather than requiring extensive external hiring or expensive reskilling investment.

Then someone asked a more specific question.

What did "Python" actually mean in this dataset?

The answer, once investigated, proved uncomfortable. For roughly fifteen percent of employees claiming this skill, it reflected genuine, current professional capability, people who wrote production Python code regularly as part of their actual job responsibilities. For another twenty percent, it reflected a single online course completed roughly eighteen months earlier, one someone had enrolled in during a period of general professional development enthusiasm, never subsequently applied in any practical context. For the remaining sixty-five percent, "Python" had been added to their profile because a well meaning manager, during a skills inventory exercise, had suggested employees list "any programming language you've heard of, even if you haven't used it much," resulting in widespread, essentially meaningless self reported entries bearing little relationship to genuine current capability.

The dashboard's impressive sixty-three percent figure was, in any meaningful practical sense, almost entirely fiction.

This is not merely a data quality problem, though it certainly involves data quality. It reflects something more fundamental: the absence of shared, precise vocabulary for what a "skill" actually means, what level of proficiency different claims represent, what evidence, if any, substantiates these claims, and how quickly claimed skills should be considered to decay in relevance absent ongoing application.

This chapter explores why building genuine skills intelligence, the foundation increasingly essential for effective internal mobility, workforce planning, and responsible AI powered talent matching, requires far more disciplined vocabulary and evidence architecture than most organizations currently maintain.

By the end of this chapter, you should be able to ask: what does "skill" actually mean in your organization's current data architecture? What evidence, if any, substantiates skill claims in your systems? How do you handle skill decay, the reality that claimed capability without ongoing application often becomes less reliable over time? And how can AI powered skills matching avoid amplifying, rather than correcting, the kind of essentially meaningless data illustrated in this chapter's opening example?

The Precision Problem

Human language naturally tolerates significant imprecision around skill claims in everyday conversation.

When someone says, "I know some Python," this statement's actual meaning depends heavily on conversational context, the relationship between speakers, and shared understanding regarding what threshold of capability the phrase implies in this particular exchange. A casual conversation with a friend tolerates far more imprecision than a formal job interview, which itself tolerates more imprecision than a legally binding professional certification.

Enterprise skills data architecture, however, often inherits this natural human linguistic imprecision without adequately accounting for how this imprecision undermines the data's usefulness for the increasingly consequential purposes organizations now want skills data to serve.

If skills data merely supported casual conversation, imprecision might not matter significantly. But contemporary organizations increasingly want skills data to support genuinely consequential decisions: identifying candidates for internal mobility opportunities, informing workforce planning regarding build versus buy talent decisions, powering AI matching algorithms connecting employees to projects or learning opportunities, and increasingly, informing decisions about organizational capability and strategic workforce investment.

These consequential uses require precision that casual, unstructured self reported skill claims simply cannot reliably provide.

This suggests organizations need genuine architectural discipline around skills vocabulary, moving beyond the natural imprecision of everyday language toward something more rigorous, capable of supporting the increasingly consequential decisions organizations want this data to inform.

Building a Skills Taxonomy

A skills taxonomy provides structured, consistent vocabulary for describing capability, moving beyond the essentially unlimited, inconsistent variation that emerges when individuals freely describe their own capabilities using whatever terminology feels natural to them personally.

Effective skills taxonomy development involves several important design decisions.

Granularity level determines how specifically skills are defined. Overly broad categories, "technology skills" or "communication skills," provide insufficient precision for meaningful matching or analysis. Overly granular categories, distinguishing dozens of hyper specific variations within what practitioners would generally consider a single coherent skill area, create unwieldy complexity and inconsistent application, since individuals may reasonably disagree regarding which of many overly specific micro categories best describes their actual capability.

Finding appropriate granularity requires genuine judgment, typically informed by how the organization actually intends to use this data. Workforce planning purposes might tolerate broader categorization than specific project staffing decisions requiring more precise capability matching.

Skill relationships and hierarchies help organize taxonomy into coherent structure, showing how specific skills relate to broader skill families, and how related skills connect to each other in ways that support intelligent matching, someone with strong Python capability likely has meaningfully related, if not identical, capability relevant to other data science programming contexts, even without explicit claim regarding every specific related skill.

Standard naming conventions prevent the kind of inconsistent terminology that undermines data usability, ensuring "Python," "Python programming," and "Python development" are recognized as referring to the same underlying skill rather than being treated as three distinct, unrelated data points due to inconsistent naming.

Regular taxonomy maintenance ensures the skills vocabulary remains current as genuine capability requirements evolve, retiring skills that have become obsolete, adding newly relevant skills as they emerge, and periodically reviewing whether the taxonomy's granularity and organization continue serving organizational need effectively as circumstances change.

This taxonomy development represents genuine investment, requiring thoughtful design rather than simply adopting whatever generic skills list a technology vendor happens to provide as default. Organizations serious about genuine skills intelligence typically need to invest real effort customizing and maintaining taxonomy that authentically reflects their specific organizational context and strategic priority, rather than assuming generic vendor provided taxonomy adequately serves their particular needs without meaningful customization.

Proficiency Levels and What They Actually Mean

Beyond simply identifying which skills someone claims, meaningful skills architecture requires some mechanism for capturing proficiency level, since claiming a skill exists at all provides limited useful information absent some indication of how developed that capability genuinely is.

Common proficiency frameworks typically define several distinct levels, though specific terminology and level count varies across different taxonomic approaches.

A basic awareness level might indicate someone has fundamental conceptual understanding without significant practical application experience, perhaps having completed introductory coursework or possessing theoretical knowledge without substantial hands on professional application.

An applied practitioner level typically indicates genuine, regular practical application within actual professional context, someone who uses this skill as meaningful part of their actual current job responsibilities, with demonstrated capability to apply this skill effectively to solve genuine problems.

An advanced or expert level typically indicates sophisticated mastery, often including capability to handle particularly complex or novel applications of this skill, and frequently including capability to teach, mentor, or provide expert guidance to others regarding this skill area.

Simply establishing these proficiency level categories does not automatically ensure consistent, meaningful application across the organization. Without genuine calibration effort, different individuals, and different managers validating skill claims, may apply dramatically different standards regarding what qualifies for each proficiency level, undermining the taxonomy's intended precision through inconsistent application.

This suggests organizations need genuine calibration investment, providing clear behavioral anchors describing what specific evidence or capability characterizes each proficiency level for particular skills, along with training and consistent application guidance helping ensure reasonably consistent interpretation across different individuals and different organizational contexts applying these proficiency definitions.

Evidence Types and Their Reliability

Perhaps the most crucial architectural question involves what evidence, if any, substantiates skill claims, since unsubstantiated self reported claims, as this chapter's opening example illustrates, often prove essentially meaningless for genuinely consequential organizational decision making.

Different evidence types carry meaningfully different reliability.

Self reported claims without additional substantiation represent the weakest evidence category, subject to significant inconsistency based on individual differences in confidence, self awareness, and interpretation regarding what threshold justifies claiming particular skill or proficiency level.

Manager validation adds meaningful additional reliability, since managers typically possess direct observational experience regarding employee's actual demonstrated capability within real work context, though this evidence type still carries some limitation, since managers may lack sufficient technical expertise to accurately assess certain specialized skill claims, or may apply inconsistent standards across different employees they supervise.

Credential and certification evidence, formal completion of recognized training programs, professional certifications, academic credentials, provides more standardized, external validation, though this evidence type primarily demonstrates that someone completed particular training or assessment, which does not always perfectly correlate with genuine practical application capability in real world professional context.

Project based evidence, demonstrated application of claimed skill within actual completed work, arguably provides the strongest evidence category, since this reflects genuine practical application rather than merely theoretical knowledge or completed training, though this evidence type requires more sophisticated data architecture to capture systematically, connecting skill claims to specific project or work outcomes that substantiate genuine practical application.

Peer or client feedback evidence, structured input from colleagues or clients who have direct experience with someone's demonstrated capability, can provide valuable additional substantiation, particularly for skills where direct managerial observation may be limited, though this evidence type requires careful design to avoid the kind of inconsistency and potential bias that can affect any subjective evaluation process.

Sophisticated skills architecture typically incorporates multiple evidence types, weighted according to their relative reliability, rather than relying entirely on any single evidence category, recognizing that different evidence types offer complementary strength and limitation that, combined thoughtfully, provide more reliable overall picture than any single evidence type alone could achieve.

Skill Decay and Temporal Relevance

Skills, unlike some other data categories, genuinely decay in relevance over time absent ongoing application, a reality that skills architecture must explicitly address rather than treating skill claims as permanently, indefinitely valid once initially established.

This decay reflects several underlying dynamics.

Technical skills, particularly in rapidly evolving domains like software development or digital marketing, may become genuinely outdated as underlying technology, tools, or best practice evolves, meaning capability that was genuinely current and valuable several years ago may no longer accurately reflect current professional standard or expectation.

Even relatively stable skills tend to atrophy somewhat absent ongoing practical application, since human capability, particularly for complex technical or professional skills, generally requires ongoing practice to maintain peak proficiency, meaning claimed capability from several years ago, even if genuinely accurate at the time, may not accurately reflect current capability absent ongoing application during the intervening period.

This suggests skills architecture needs explicit temporal dimension, capturing not merely whether someone claims particular skill and proficiency level, but when this claim was established or last validated, and ideally, some mechanism for periodic revalidation or explicit decay modeling that appropriately discounts confidence in skill claims that have not been recently substantiated through ongoing application or explicit reassessment.

Different skills may reasonably warrant different decay timelines, reflecting how quickly genuine capability tends to atrophy or become outdated within particular skill domains. Rapidly evolving technical skills might warrant more aggressive decay modeling, treating claims as significantly less reliable after relatively brief periods absent recent application, compared to more stable, foundational skills less subject to rapid technological or best practice evolution.

This temporal architecture adds genuine complexity beyond simply capturing static skill claims, but this complexity reflects genuine underlying reality that skills architecture must accommodate if it hopes to provide genuinely reliable, current picture of organizational capability rather than accumulating increasingly outdated, unreliable historical claims that no longer accurately reflect genuine current capability.

AI and the Amplification Risk

AI powered skills matching and talent intelligence capabilities offer genuine potential value, helping organizations more effectively identify internal capability for project staffing, internal mobility opportunity matching, and strategic workforce planning purposes.

However, this chapter's opening example illustrates a crucial risk: AI systems trained on or matched against unreliable underlying skills data will amplify, not correct, this underlying data quality problem, potentially with even more consequential downstream impact given AI's capacity to process and act upon this data at significant scale.

If an AI powered talent matching system treats the essentially meaningless "sixty-three percent claim Python" data with the same confidence it would treat genuinely well substantiated skill data, it may confidently recommend individuals for opportunities requiring genuine Python capability who, in reality, possess minimal or no genuine practical capability in this area, having only nominally claimed this skill during a poorly designed skills inventory exercise without any genuine substantiating evidence.

This suggests AI powered skills matching capability must incorporate genuine awareness of underlying evidence quality and reliability, weighting or filtering skill claims based on evidence type and recency, rather than treating all claimed skills as equally reliable regardless of underlying substantiation, or lack thereof.

Responsible AI implementation in this domain requires the underlying evidence and vocabulary architecture this chapter describes, genuine taxonomy discipline, meaningful proficiency level definition, appropriate evidence weighting, and explicit decay modeling, functioning as necessary foundation before AI powered matching capability can provide genuinely reliable, trustworthy output, rather than assuming sophisticated AI algorithm alone can somehow compensate for fundamentally unreliable underlying data through some kind of algorithmic magic that, in reality, cannot exist.

Garbage in, garbage out remains as true for sophisticated AI powered skills matching as for any other data driven system, perhaps even more consequentially true given AI's capacity to process this data at scale and produce confident seeming output that may mask underlying data unreliability from users who reasonably trust AI generated recommendations without necessarily understanding the potentially unreliable underlying data foundation these recommendations rest upon.

Governance and Ownership

Building and maintaining genuine skills architecture requires clear organizational ownership and governance, not merely initial taxonomy development followed by assumption that the system will somehow maintain itself indefinitely without ongoing deliberate stewardship.

Effective skills governance typically requires several elements.

Clear ownership for taxonomy maintenance, identifying specific accountability for periodically reviewing and updating skills taxonomy as organizational needs and industry landscape evolve, rather than allowing taxonomy to gradually become outdated absent explicit ongoing maintenance responsibility.

Evidence validation processes, ensuring genuine mechanism exists for substantiating skill claims through appropriate evidence types, rather than relying entirely on unsubstantiated self report that this chapter's opening example illustrates as often essentially meaningless for genuinely consequential purposes.

Decay modeling governance, establishing and periodically reviewing appropriate decay timelines for different skill categories, ensuring this temporal dimension receives genuine ongoing attention rather than being implemented once and then ignored as circumstances and technology continue evolving.

Quality monitoring, periodically assessing whether skills data actually demonstrates the kind of reliability and evidence substantiation the architecture intends, rather than assuming initial good design automatically ensures ongoing data quality without periodic verification.

Cross functional coordination, since effective skills architecture typically requires genuine collaboration between HR, learning and development, workforce planning, and often business unit leadership, ensuring taxonomy and evidence architecture genuinely reflects authentic organizational capability need rather than being developed in isolation without adequate input from those who will ultimately depend on this data for consequential decision making.

This governance investment represents genuine ongoing organizational commitment, not merely initial project effort, reflecting the reality that skills architecture, like other genuinely valuable enterprise data assets discussed in earlier chapters, requires ongoing stewardship to maintain genuine reliability and usefulness over time, rather than inevitably degrading toward the kind of essentially meaningless data this chapter's opening example illustrates absent this ongoing deliberate governance attention.

Counter-Perspective

"This Level of Rigor Is Impractical"

There is a legitimate counterargument suggesting this chapter's recommended discipline exceeds what most organizations can practically implement.

Building comprehensive taxonomy, calibrated proficiency levels, multiple evidence type architecture, and explicit decay modeling represents significant organizational investment that many organizations, particularly smaller organizations without dedicated capability specifically focused on skills architecture, may struggle to implement with the full sophistication this chapter describes.

This concern has genuine merit.

However, the underlying principle, that unsubstantiated, imprecise skills data provides limited genuine value for consequential decision making, remains valid even for organizations unable to implement the full sophistication this chapter describes.

Organizations with more limited capability can still meaningfully improve skills data reliability through more modest but genuine practice: focusing evidence and validation effort specifically on the smaller set of skills most critical to genuine strategic priority rather than attempting comprehensive coverage across every conceivable skill, implementing basic manager validation even without more sophisticated multiple evidence type architecture, and periodically prompting skill claim refresh even without fully automated decay modeling.

The goal is not necessarily achieving the full sophistication described in this chapter's more comprehensive discussion, but genuinely internalizing the underlying principle that skills data quality matters significantly for the consequential purposes organizations increasingly want this data to serve, and applying this principle proportionately given genuine organizational capability and resource constraint, rather than either ignoring this challenge entirely or becoming paralyzed by the apparent complexity of fully comprehensive implementation.

Case Note

A professional services firm sought to build an internal talent marketplace, allowing employees to discover and apply for internal project opportunities based on demonstrated skills and capability, reducing dependence on informal networks and manager relationships that had previously determined project staffing in ways that arguably disadvantaged employees without strong existing internal visibility or relationships.

Initial skills data, migrated from an existing but poorly maintained skills inventory system, exhibited exactly the kind of unreliability this chapter describes, extensive self reported claims without substantiating evidence, inconsistent terminology across different individuals describing essentially similar capability using different language, and no meaningful mechanism for distinguishing genuine current capability from outdated or exaggerated claims.

Rather than attempting comprehensive immediate remediation across the firm's entire skills taxonomy, the implementation team adopted a phased approach, initially focusing rigorous evidence and validation effort specifically on the relatively small set of skills most directly relevant to the firm's highest priority strategic capability areas, artificial intelligence and data analytics capability the firm was actively trying to build and deploy more effectively across client engagements.

For this focused skill set, the team implemented genuine multiple evidence type architecture: project based evidence connecting skill claims to specific completed client engagements demonstrating genuine practical application, manager validation requiring explicit confirmation based on direct observed capability, and appropriate decay modeling given the genuinely rapid evolution characteristic of this particular technical domain.

For the broader remaining skill taxonomy, covering less immediately strategic priority skill areas, the team implemented more modest improvement, basic terminology standardization and simple manager validation, without the full sophisticated evidence architecture applied to the highest priority strategic skill set.

This phased approach allowed the organization to achieve genuine, reliable skills intelligence for the specific capability area most critical to current strategic priority, supporting genuinely trustworthy internal mobility matching for AI and data analytics related opportunities, while accepting more modest, though still meaningfully improved, data quality across the broader skill taxonomy where perfect precision, while theoretically desirable, was not immediately essential for current strategic priority.

The lesson extends beyond this specific example. Organizations need not achieve uniform, comprehensive sophistication across their entire skills taxonomy simultaneously. Thoughtful prioritization, focusing the most rigorous evidence and validation effort on genuinely strategic priority skill areas while accepting more modest improvement elsewhere, often represents more practical, achievable path toward genuine skills intelligence than attempting comprehensive uniform sophistication that may prove impractical given genuine organizational resource constraint.

Systems Lens: Shared Vocabulary as Coordination Infrastructure

In systems terms, shared, precise vocabulary functions as crucial coordination infrastructure, enabling meaningful communication and coordination across the organization regarding genuine capability, without which numerous downstream processes, internal mobility matching, workforce planning, AI powered talent intelligence, cannot function reliably regardless of how sophisticated the surrounding technical architecture might otherwise be.

This chapter's opening example, where "Python" meant three essentially different things depending on which employee subset one examined, illustrates how absence of this shared vocabulary infrastructure undermines coordination capability, since the organization's various systems and decision makers cannot meaningfully coordinate around genuine capability when the fundamental vocabulary describing this capability lacks sufficient precision and shared meaning.

Building this coordination infrastructure requires genuine investment, the taxonomy development, proficiency calibration, evidence architecture, and governance this chapter describes, functioning as necessary foundation enabling the kind of reliable, meaningful coordination around organizational capability that increasingly sophisticated talent management and AI powered capability increasingly depends upon.

Philosophical Digression

There is an old philosophical observation that naming things precisely represents a form of respect, taking the subject seriously enough to describe it with genuine accuracy rather than convenient, imprecise approximation.

When organizations allow "Python" to mean simultaneously genuine professional capability and essentially meaningless nominal claim without any distinction, this imprecision reflects, in some sense, insufficient respect for the genuine capability that some employees have worked hard to develop, capability that deserves more precise recognition than being lumped together, undifferentiated, with essentially unsubstantiated claims that happen to use identical terminology.

This precision matters not merely for organizational efficiency, though it certainly serves this practical purpose, but for a kind of fairness toward individuals whose genuine developed capability deserves accurate recognition, rather than being obscured within imprecise aggregate data that fails to distinguish authentic expertise from superficial familiarity.

Building genuine skills vocabulary precision, in this sense, reflects not merely technical data architecture discipline, important as this practical dimension certainly is, but a form of genuine respect for the real, differentiated human capability that precise, evidence based skills architecture allows organizations to actually see and appropriately value, rather than obscuring this genuine differentiation within convenient but ultimately misleading imprecise aggregate claims.

Further reading: George Lakoff and Mark Johnson, Metaphors We Live By; Anders Ericsson, Peak: Secrets from the New Science of Expertise; Cathy O'Neil, Weapons of Math Destruction.

Reflection Questions

1. What does "skill" actually mean in your organization's current data architecture, and how consistently is this meaning applied across different individuals and systems?
2. What evidence, if any, currently substantiates skill claims in your systems, and how reliable is this evidence given the framework this chapter describes?
3. Does your skills architecture account for temporal decay, or does it treat all skill claims as permanently valid regardless of how long ago they were established or last substantiated?
4. Where might AI powered skills matching in your organization be amplifying, rather than correcting, underlying data quality problems?

Given genuine resource constraint, which specific skill areas might warrant the most rigorous evidence and validation investment as strategic priority, following the phased approach this chapter's case example illustrates?

Key Takeaways

Skills data often suffers from essentially meaningless imprecision, unsubstantiated self reported claims using inconsistent terminology without genuine evidence substantiation, undermining reliability for the increasingly consequential purposes organizations want this data to serve.

Building genuine skills intelligence requires disciplined taxonomy development, calibrated proficiency level definition, multiple evidence type architecture weighted according to reliability, and explicit temporal decay modeling accounting for how claimed capability becomes less reliable absent ongoing application.

AI powered skills matching risks amplifying underlying data quality problems rather than correcting them, requiring genuine evidence quality awareness rather than treating all skill claims as equally reliable regardless of substantiation.

Effective skills governance requires clear ownership, evidence validation processes, decay modeling governance, and ongoing quality monitoring, representing genuine sustained organizational commitment rather than one time implementation project.

Organizations with limited resources can still meaningfully improve skills data reliability through thoughtful prioritization, focusing rigorous evidence architecture on genuinely strategic priority skill areas rather than attempting uniform comprehensive sophistication across their entire taxonomy simultaneously.

Optional Reading

George Lakoff and Mark Johnson, Metaphors We Live By This foundational work on how language shapes thought offers valuable background for understanding why precise vocabulary genuinely matters for organizational coordination, beyond merely representing pedantic linguistic preference.

Anders Ericsson, Peak: Secrets from the New Science of Expertise Ericsson's research on genuine expertise development offers valuable grounding for understanding what authentic skill mastery actually requires, useful context for calibrating meaningful proficiency level definitions within skills architecture.

Cathy O'Neil, Weapons of Math Destruction O'Neil's examination of how algorithmic systems can amplify underlying data problems offers directly relevant warning regarding the AI amplification risk this chapter discusses regarding unreliable skills data.

Josh Bersin's research and writing on skills based organizations offers practical, industry grounded perspective on contemporary skills architecture challenges and emerging practice, valuable supplementary context for this chapter's more foundational discussion.

Ronald Heifetz, Leadership Without Easy Answers While not directly focused on skills architecture, Heifetz's work on adaptive challenges offers valuable framework for understanding why building genuine skills intelligence often requires the kind of sustained organizational commitment and cultural change this chapter's governance discussion describes, rather than simple technical implementation alone.

Quiet Reflection

Somewhere in your organization's skills data right now, genuine expertise sits alongside essentially meaningless nominal claims, both using identical terminology, both appearing with superficially equal weight within whatever dashboard or system currently aggregates this information.

The employee who has spent years developing genuine mastery deserves better than having this authentic capability obscured within imprecise aggregate data that fails to meaningfully distinguish their real expertise from a colleague's essentially unsubstantiated claim based on a single forgotten online course.

Building genuine skills vocabulary precision reflects more than technical architecture discipline.

It reflects taking real human capability seriously enough to describe it accurately, allowing organizations to genuinely see and appropriately value the authentic expertise their people have actually developed, rather than obscuring this genuine differentiation within convenient but ultimately misleading imprecision that serves no one well, however tempting this imprecision's apparent simplicity might initially seem.

Cite this chapter: Roy, A. (2026). Chapter 21: The Vocabulary of Skills. In Designing the Architecture of Dignity. Retrieved from https://dignity.consciouscybernetics.org/chapter-21

Index  • ← Previous • Next → • Browse by Topic