A skills visibility gap needs work evidence, not another confidence score
A Harris Poll survey of 755 U.S. HR and learning leaders found greater confidence in knowing which skills jobs need than in knowing whether employees possess them. That is a measurement problem, not proof of a workforce deficit.
What happened
University of Phoenix reported that 64% of surveyed HR leaders were very confident they knew the specific skills employees needed, while 43% were very confident employees possessed them. The online U.S. survey ran from 14 to 27 July 2026 among 755 HR and learning leaders.
Why it matters
The difference compares two perceptions held by the same professional group; it does not measure employee capability. Leaders should replace broad confidence with evidence tied to tasks, work products, conditions and recency before buying training or changing roles.
University of Phoenix reported a 21-point difference between two confidence questions in its Future of Skills Development survey. Among 755 U.S. HR and learning leaders, 64% said they were very confident they knew the specific skills employees needed to do their jobs, while 43% were very confident employees possessed those skills. Keeping pace with AI and other technology was the largest named source of uncertainty about future skill needs, cited by 27%.
The survey was conducted online by The Harris Poll from 14 to 27 July 2026. Respondents were employed adults aged 25 or older, held at least a manager or supervisor title, worked in HR or learning and development, and participated in talent acquisition, development or workforce planning. The stated 95% Bayesian credible interval is plus or minus 4.6 percentage points for the full sample and wider for subgroups. Respondents came from a panel of people who had agreed to take surveys.
Do not convert a perception gap into a skills deficit
The two percentages describe leader confidence, not observed employee capability. They do not establish that 21% of a workforce lacks skills, that AI caused a gap, or that training would close it. A manager may know a role profile yet lack current evidence about performance. Employees may also demonstrate capabilities that the organisation does not record. The gap is useful because it identifies an information problem, but it cannot diagnose its cause.
Start with decisions. Choose three workflows where a capability judgment will change staffing, access, training or role design. For each workflow, define the task, quality threshold, operating conditions and evidence window. A skill label such as problem solving is too broad until it is connected to a work product, a constraint and a review standard.
Build an evidence chain employees can inspect
Use several evidence types: reviewed work samples, structured observation, simulations for rare tasks, error and rework patterns, and employee self-description with examples. Record who assessed the evidence, which rubric they used and how recently the task was performed. A single manager rating or AI-generated inference should not become the permanent record of a person's capability.
Employees need access to the same evidence. Let them correct identity and task attribution, add context, challenge an assessment and see how a decision used it. This matters when systems infer skills from communications, activity logs or work products. Data collected for operations should not silently become an employment score; define purpose, retention, access and appeal before reuse.
The counterargument is cost. Direct observation is slower than asking leaders or mining profiles. A risk-based approach keeps it proportionate: require stronger evidence for promotion, redeployment or exclusion from work; use lighter signals to suggest voluntary learning. Sample workflows rather than attempting an enterprise-wide ontology first.
Measure whether visibility improves a decision. Track disagreement between assessors, employee corrections, evidence age, prediction of task quality and whether a development action changes performance. Do not treat course completion or a filled skills profile as proof of capability. The Skills Intelligence Role Dictionary can help bound role scope, but local evidence must still show what a person can do under relevant conditions.
Do not optimise the pilot for the number of skill labels populated. A smaller inventory with current, contestable evidence is more useful than broad coverage assembled from stale profiles. Publish coverage gaps explicitly and prevent missing evidence from being interpreted as missing ability. That distinction protects employees and gives leaders a clearer backlog for observation.
The practical decision is to fund a small capability-evidence service before a broad skills inventory. If the organisation cannot explain the task, evidence, reviewer, age and appeal path behind a skill claim, it has produced another confidence score rather than useful visibility.