Listen first
LDNUCU | 26 October 2026
Training to spot AI errors
74% of employers buy workplace tech faster than they train staff to audit it.
74 percent of North American decision-makers admit their organizations buy workplace technology faster than they help employees learn to use it. Organizations buy the licenses, turn them on, and assume access automatically generates competence in the workforce.
74 of every 100 decision-makers. Blocks show a percentage, not individual respondents.
The gap
Everyone is using it. Few feel ready.
The contrast between raw tool usage and actual employee confidence is very sharp right now. Organizations buy the licenses, turn them on, and assume access automatically generates competence in the workforce.
The consequences show in the latest Gulf Cooperation Council study from Al-Futtaim. Across six GCC markets, 82 percent of professionals use AI at least weekly, and 73 percent report productivity improvements. Only 39 percent feel highly confident using the technology.
The numbers show a large majority folding these tools into daily routines while guessing their way through the workflows. Organizations are treating AI deployment like buying employees high-end gym memberships without providing any personal trainers. Leadership teams buy the licenses, send out a single welcome email, and then act surprised when people either injure themselves, which in this context means unauthorized shadow IT or data breaches, or stop going altogether.
Pick two rails to measure the distance
Select a rail to read what it measures. Select a second rail to see how far apart two findings sit.
The training deficit behind those outcomes is stark. Only about 20 percent of the GCC workforce receives extensive training on these systems, and another 25 percent receive none at all. The exclusion is particularly pronounced among older workers. Among workers aged 45 and older, 47 percent receive no formal AI training from their employers.
Organizations are rolling out systems that alter the fundamentals of work while leaving nearly half their experienced workforce to figure out the mechanics on their own. That forces hesitation. Thirty-three percent of respondents in the GCC study cite data security as their primary barrier to using AI. People freeze up because they do not understand the technical boundaries. Nobody is teaching them the underlying architecture, so they do not know what proprietary data can safely be shared with an external model.
Structured learning generates a measurable operational return that addresses that hesitation.
Readiness and results
ASCM and the Digital Supply Chain Institute, joint report.
The cost of untrained AI
Where preparation is missing, workarounds appear.
The operational cost of ignoring structured training shows up clearly in supply chains. A joint report from the Association for Supply Chain Management, ASCM, and the Digital Supply Chain Institute reveals that 83 percent of supply chain leaders consider their workforces unprepared for AI. Organizations with AI-ready workforces report a 71 percent AI project success rate, while unprepared organizations sit at 28 percent.
Going back to the gym analogy, when employees lack official training and fail to receive functional approved tools, they build their own unauthorized workarounds to get through the workday. A recent US and Canada workplace technology survey reported that 43 percent of Canadian employees use unauthorized tools on a monthly basis, feeding proprietary company data into public AI models to hit their targets.
Make a prediction
Leaders and employees were asked whether employees have clear guidance on responsible AI use.
Decision-makers reported 89 percent. Estimate the employee figure, then reveal it.
The governance disconnect that runs alongside that shadow usage is striking. The same survey shows 89 percent of decision-makers believe their employees have clear guidance on responsible AI use, yet only 62 percent of employees agree. Leaders think checking a box on a compliance module translates to daily governance.
When employees rely on unauthorized tools to hit targets, what happens when those tools begin automating foundational analytical work? That is the exact work where junior employees historically learned the nuances, the edge cases, and the specific rules of their trade.
The central question
The work that taught judgment is the work being automated.
ASCM CEO Abe Eshkenazi outlines the danger of this transition. Employees historically built their critical judgment by running the analyses themselves. They worked the spreadsheets manually. They learned to spot workflow exceptions by processing data line by line, developing a sense for when a specific number looked anomalous.
Handing those foundational tasks to an AI while pushing junior workers into decision-making roles removes their ability to evaluate the output. A junior supply chain analyst who has never built a demand forecast manually lacks the historical context to challenge a system.
If an AI tool does the foundational analysis, the what of a task, instructional design has to redesign entry-level training to teach the underlying logic, the why. Otherwise, when the AI hallucinates a spike in inventory requirements, the analyst rubber-stamps the error because they lack the judgment to recognize it as a hallucination.
Build the analyst's foundation, then send the forecast for review
AI output: inventory requirements Conceptual illustration, no real data
Build the foundations the analyst has, or leave them out, then send the forecast for review.
A conceptual model of the reasoning described by ASCM's CEO. It shows no measured data and no real forecast.
The shortcut that does not work
Microlearning cannot stand in for hands-on judgment.
L&D cannot rely on traditional shortcuts to fix this judgment gap. You cannot throw microlearning or rapid video content at someone to substitute for the hands-on experience of spotting workflow exceptions.
A scoping review of 17 microlearning studies in JMIR Medical Education confirms the limits of that approach. The review shows that while spaced learning produces knowledge and recall gains, it results in minimal measured behavioral change or workplace outcomes. Memorizing a compliance rule on a digital flashcard does not equip an employee to identify a subtle system error during an operational crisis.
Climb the evidence, one step at a time
Scoping review of 17 microlearning studies, JMIR Medical Education. This edition reports the pattern the review found, not the size of any effect.
Shorten the move from education into work
Pearson's Lost in Transition research calculates that delays moving from education into work cost the Australian economy 12.5 billion Australian dollars annually in lost earnings.
Two kinds of mastery
Trades now ask for dual capability.
This demand for dual capability, combining physical trade mastery with critical judgment to operate AI-supported tools, is straining technical fields. Pearson and WorldSkills Australia recently partnered to address pipeline friction, noting that 49 percent of technician and trade occupations in Australia are in national shortage.
A mechanic today has to diagnose complex software faults alongside repairing physical mechanical failures. Pearson's Lost in Transition research calculates that delays in transitioning from education into work cost the Australian economy 12.5 billion Australian dollars annually in lost earnings, and that shortening the transition period from 24 weeks to 18 weeks could add 5 billion Australian dollars to the economy.
Governance moves to the middle
The manager is where the system asks for a signature.
This automation of foundational work extends beyond the factory floor and the supply chain. Enterprise HR and workforce management systems are absorbing foundational administrative tasks, which shifts a substantial governance burden onto middle managers. Recent vendor updates show the scale of that shift.
Productivity claims are accelerating across the vendor ecosystem, and each figure above is the vendor's own. Middle managers are being squeezed from both sides. They have employees building their own background automations with tools like EmployeeWorks, while facing strict new regulatory requirements for human oversight of automated decisions.
AceUp CEO Will Guillaume Foussier emphasizes that managers are expected to protect human judgment, empathy, and coaching as transactional work automates away. Yet the data suggests managers are unprepared for that responsibility. Gartner finds only 8 percent of HR leaders believe their managers currently have the skills to use AI effectively.
Regulatory obligations are moving faster than manager capability. The EU AI Act already mandated AI literacy starting in February 2025, and by December 2027 it requires trained human overseers for high-risk HR AI systems. California's No Robo Bosses Act, which takes effect in July 2027, mandates human judgment in consequential workforce decisions like hiring, firing, and promotions.
Organizations face project risks when they deploy agentic systems without prepared human overseers, leaving managers legally required to audit complex hiring or redeployment decisions generated by systems they do not fully understand. Gartner forecasts that over 40 percent of agentic AI projects will be cancelled by the end of 2027 due to cost and unclear value.
Move through time and watch the duties stack up
Build, because buying is blocked
When talent cannot be bought, capability has to be built.
With middle managers stretched thin, specialized AI talent scarce, and external entry-level pipelines tightening, organizations have to find new ways to build and verify internal capabilities. Buying talent on the open market is often not viable.
The cybersecurity sector shows how organizations are reacting to that scarcity. Fortinet reports that 68 percent of IT decision-makers struggle to find AI cybersecurity talent. Compounding that, 67 percent cannot get budget approval for new headcount to hire externally even if the talent existed.
The business risk is substantial. Fortinet notes 94 percent of organizations experienced a security breach in the past year, and 74 percent faced costs over 1 million dollars. Sixty-three percent cite a lack of cybersecurity skills as a leading cause of those breaches. In response to headcount constraints and external scarcity, 97 percent of organizations say they will pay for current employees to earn cybersecurity certifications.
Filter the numbers by what they say
All findings shown. The first two describe a market organizations cannot buy from. The last describes what they say they will do instead.
Employers are forced to build internally because traditional entry-level pipelines are narrowing. An audit by Extern of the education company Stride showed zero open internships out of 100 open requisitions. LDP Connect highlights a broader shift toward sporadic hiring over predictable structured cohorts. Learning and development functions are increasingly responsible for building foundational business skills for new hires who lack prior internship experience.
Public entities and employers are exploring new methods to align skills with actual labor demand. The Philippine Department of Labor and Employment is exploring a collaboration with LinkedIn using its Economic Graph to align workforce training with real-time hiring trends. SHRM India notes that future-ready organizations are shifting toward three-to-five-year skills forecasting and manager-led development conversations. AustCorp urges employers to separate mandatory licenses from historical degree preferences and focus on assessing learning agility through structured work samples.
Verification that outlasts day one
Six months of signals, read for friction rather than volume.
Rethinking how organizations verify capability leads directly to continuous onboarding. Modern HR technology is shifting away from standard first-day checklists toward a six-month intelligence loop. Acteon's compliance training commentary shows why simple checklists fail, citing research from the University of Melbourne and KPMG that found 39 percent of UK employees have used AI at work in ways that contravene organizational policies despite mandatory training norms. Checking a box in a learning management system does not alter behavior under operational pressure.
Building capability requires tracking behavioral signals across learning, workflow, knowledge-seeking, and collaboration over six months. That can easily sound like workplace surveillance. From an employee's perspective, how does an intelligence loop distinguish between measuring capability development and counting keyboard clicks?
Read the same six months two ways
Counting activity treats every signal alike. A manager who weaponizes that count turns it into surveillance.
The new hire stalls in a complex process. What does the manager do?
Illustrative new hire. This grid is a conceptual model and shows no measured data.
Effective workflow intelligence looks at process bottlenecks and friction rather than raw activity volume. Tracking how many times an employee searches a knowledge base is invasive if a manager weaponizes that data, whereas responsible capability measurement tracks when an employee moves from initial familiarization to independent operational proficiency. Identifying where new hires stall in a complex process allows a manager to provide targeted coaching and fix broken documentation rather than assigning another generic compliance module.
What L&D has to teach next
The critical skill is auditing.
As agentic systems absorb both entry-level analysis and transactional HR coordination, the critical skill L&D must teach is auditing. Workplace learning is no longer only about training people to operate systems. It requires training the workforce to build the domain expertise needed to spot when the system is confidently wrong.
LDNUCU | YouTube Podcast
ldnucu.com | Learning and Development News You Can Understand