ldnucu.com | L&D News | 14 October 2026
Shadow upskillingThe hidden AI shift
Employees are building AI skills that sit outside job descriptions, pay structures and the view of the learning function. Formal systems are stretched too thin to lead the change, so capability is being built off the books.
95 out of every 100 HR professionals report that employees use AI tools every single day in roles where AI is not part of the official job description.
Each block stands for one in a hundred. A proportional picture of each reported percentage, not a headcount.
The stretched function
A strategic mandate, a runway to sweep
The isolved 2026 HR Leaders Report surveyed over a thousand US human resources decision-makers and found a contradiction at the center of the function. 99% say HR is viewed as strategic within their organization. Yet 47% report spending four or more hours every day fielding repetitive employee queries, close to half the working day.
An HR or learning leader mandated to map the enterprise's future capability is answering questions about PTO approval, password resets and policy clarifications. It is an air traffic controller too busy sweeping the runway to guide the planes.
The cost shows up in readiness. The workforce skills gap widened by 11 points over the past year and now sits at 59%, and over half of respondents call the shortage an outright talent crisis. Facing budget pressure, 34% of organizations are investing in AI and automation instead of hiring new headcount, while lacking the strategic bandwidth to reskill the people who would use it.
Select a bar to see what it measures
Share of surveyed HR decision-makers who say the HR function is viewed as strategic within their organization. Almost universal.
Buying software does not build the capability to operate that software.
Off the books
When the job description stops describing the job
With no formal function guiding the transition, employees are not waiting for curricula to be published. They are quietly rewriting their own roles. The Oyster report calls it shadow upskilling.
Two-thirds of companies have added AI requirements to non-technical roles: marketing coordinators, operations staff, project managers. Almost half of those companies, 46%, never formally communicated the changed expectations to the people doing the jobs.
53% of employers say they prefer internal upskilling to external hiring. Yet 62% of organizations changed neither the job title nor the compensation when they added the requirements. It is asking a barista to start repairing the espresso machines while paying only for the coffee.
Formalize the role, one piece at a time
A conceptual model of the gaps the reporting describes. It does not predict results.
0 of 4 formalized
Select a piece to add it to the structure. The AI skill currently hangs off the edge of the job architecture.
The incentives compound it. When workers automate their own tasks, only 38% are rewarded for it, and almost a quarter are simply assigned additional work. An employee who uses a large language model to turn a four-hour task into twenty minutes, and is rewarded with a doubled workload for showing management, has every reason to keep the skill hidden.
Informal experimentation shows agility. Left undocumented, it creates organizational exposure. If a marketing coordinator is quietly using an unapproved AI tool to generate external copy and L&D has no visibility, there is zero governance: potential data security risks, compliance exposure, and inconsistent outputs because nobody is verifying the capability. L&D risks irrelevance if skills development stays a decentralized, unverified activity that happens in the shadows.
A guided pathway
From shadow skill to verified capability
To turn undocumented experimentation into organizational capability, enterprises have to move beyond static course libraries to guided, credentialed pathways. Pearson has recently announced a blueprint: Career Studio for Work, currently being piloted in the healthcare and information technology sectors.
The platform pairs discipline-specific AI tutors with dedicated human success coaches. The machine handles the technical side and the human handles the behavioral side. The AI provides immediate diagnostic feedback when a learner meets a conceptual obstacle, which scales cognitive support. The human coach tracks milestone progression and provides behavioral accountability.
The pathways culminate in industry-recognized external credentials such as Certiport certifications: foundational, globally recognized exams that validate digital literacy and technical skills. It is not a certificate of attendance. Anchoring the pathway to an external standard means the enterprise is verifying standardized capability, not issuing an internal completion badge.
A learning management system is often a room full of gym equipment. Nobody is checking your form.
In this model the AI tutor tells you where your form is failing conceptually, and the human coach acts as the personal trainer who holds you to the regimen.
Walk the pathway. Select a moment on the stairs to see who or what responds.
The AI provides immediate diagnostic feedback when a learner meets a conceptual obstacle, so cognitive support scales without waiting for a person.
This is the structural move worth noticing: the model separates the cognitive support mechanism from the behavioral accountability mechanism. AI individualizes instruction and supplies real-time remediation, while human coaching is relied on for persistence and completion.
It also fits a rule LDNUCU applies to every enterprise learning development: content production is not learning, and course completion is not capability. Completion is a metric. Pearson has announced the model and is piloting it. No learner outcomes were reported.
Training that feels like the job
You cannot AI your way out of a jammed conveyor belt
AI coaching works when the job happens on a screen. A vital segment of the workforce relies on physical, non-automatable trades, and it faces a demographic threat: the retirement cliff.
In Milwaukee, workforce accelerator Per Scholas has partnered with The Campbell's Company on a 15-week, 400-hour advanced manufacturing technician program, backed by a $500,000 corporate investment. It targets a projected deficit of 1.9 million manufacturing workers by 2033, driven largely by retirements of veteran technicians.
The instructional design is what merits attention. The curriculum was not bought off the shelf as a vocational video module. It was built on empirical task analysis.
Trace how the plant floor became the lab. Select a stage.
Instructional designers spent a year shadowing over 200 technicians on Campbell's actual high-speed production lines.
Physical replication is not limited to factories. The US Army National Simulation Center at Fort Leavenworth recently acquired the Virtuix Omni One for infantry training, an omnidirectional treadmill that lets trainees walk, run, crouch and change direction while interacting with a synthetic environment.
The argument from this editorial analysis is that motor skill transfer requires authentic physical feedback. Muscle memory demands physical resistance and spatial awareness. Software operated by thumbsticks from a swivel chair cannot build resilient physical capability, and L&D teams building technical, medical or emergency response training should weigh the physical requirements of the job, not only the cognitive steps.
Behavioral capability needs a different kind of fidelity. UK rail operator Avanti West Coast has opened an Immerse simulation space at its Talent Academy in Crewe. It uses 270-degree video projections, synchronized multi-directional audio, dynamic lighting and scent diffusion to recreate a busy railway environment.
Staff practice high-stress scenarios such as severe platform crowding and volatile ticket barrier disputes. The space also lets them experience a station from the sensory perspective of neurodivergent passengers, including children with autism. The instructional claim is that realistic physiological stress, without real danger, changes how the brain encodes the memory. Learner engagement is not workplace performance, and conceptual agreement in a quiet classroom is not behavioral capability on a crowded platform.
The unexpressed
Put the invisible forces on the table
Bring each force up from beneath the table
In a standard planning retreat only explicit goals and resource allocations are on the table. The forces that govern how decisions are actually made remain off it.
The Training Magazine analysis asks why executive strategic planning retreats so consistently fail to change organizational behavior. Drawing on Peter Senge's theory of organizational learning, it argues that organizations learn not by absorbing new data but by shifting their collective mental models.
Typical sessions cover explicit goals and resource allocations and ignore what Senge's work calls the system of the unexpressed: the deeply held, unspoken assumptions that govern how decisions are made. An unvoiced fear of candor. Structural impatience for quarterly results. Silent assumptions about what is culturally acceptable. The things everyone knows and nobody says.
The methodology highlighted is tactile physical systems modeling, such as LEGO Serious Play. Executives build three-dimensional models of these forces and name them as objects. You are no longer criticizing the chief financial officer's impatience. You are pointing at a red block that represents the systemic pressure for rapid returns.
A physical proxy bypasses the brain's interpersonal defensive routines and depersonalizes the conflict, so leadership can surface hidden assumptions without the defensiveness that normally derails alignment. Unless learning surfaces and reshapes these mental models, stated strategic intentions will be overridden by legacy cultural habits.