The 12-Month AI Gap: Why Leaders Are Accelerating While Workers Are St – Good Machine

TL;DR

45% of leaders expect AI agents in workflows within 12 months. Only 30% of workers agree. The gap is not about technology — it's about trust, communication, and who gets a seat at the redesign table. (Adecco Group, May 21)
73% of organisations have deployed or are piloting AI. Only 18% have reskilled a majority of their workforce. The deployment-to-training ratio is roughly 4:1. (Aon, May 5)
85% of business leaders admit they feel pressure to appear further along than they actually are. Nearly as many say most companies in their industry would be caught "AI-washing" if investigated. (Milken Institute-Harris Poll, May 4)
41% of workers received zero employer AI support in the past year. Yet 69% believe AI can create more opportunities than it eliminates — if done right. The workforce is not the bottleneck. (Milken-Harris)
The industries least prepared: Hospitality (4.02 gap score), Healthcare (3.74), Financial Services (3.69). Frontline-heavy sectors face the widest chasm between AI exposure and workforce readiness. (Lightcast/Resume Now, April 24)

What Happened

In the span of four weeks — late April to late May 2026 — four independent research organisations published major studies on the AI workforce. They used different methodologies, surveyed different populations, and asked different questions. They landed in the same place.

On May 21, the Adecco Group released The Human Premium: Leadership Beyond the Algorithm, a survey of 2,000 C-suite executives across 13 countries whose organisations collectively employ 8.6 million workers. The headline finding — 45% of leaders expect AI agents integrated into workflows within 12 months, while only 30% of workers share that expectation — is the narrowest expression of a wider pattern. Only 22% of leaders are highly confident their organisations are developing future-ready workforce capabilities. Only 36% say their talent strategy clearly demonstrates that AI will create opportunities for employees. Only 39% are involving employees directly in job redesign.

Two weeks earlier, on May 5, Aon published its Human Capital Trends Report. The numbers were even starker: 73% of organisations have deployed or are piloting AI, but just 18% report that a majority of their workforce has participated in AI reskilling or upskilling in the past 12 months. Only 28% have hired anyone with AI expertise. Fewer than a quarter have created a Head of AI role. And only 28% have fully operational AI governance guidelines with oversight mechanisms in place.

The day before that, on May 4, the Milken Institute and Harris Poll released Bridging the Technology Advancement Gap: A New Consensus for the AI Era. This study captured something the others didn't: the performative dimension. 85% of business leaders admitted to feeling pressure to appear further along than they actually are. Nearly as many agreed that "most companies in their industry would be caught 'AI-washing' if investigated." Meanwhile, 41% of workers said they'd received zero employer AI support in the past year.

And on April 24, Lightcast data analysed by Resume Now produced the 2026 AI Workforce Preparedness Rankings, quantifying the AI Skills Gap Score across ten industries. Hospitality led the list at 4.02, followed by Healthcare (3.74), Financial Services (3.69), and Logistics & Warehousing (3.69). The pattern was consistent: industries with large frontline or operational workforces face the widest gaps, and those gaps are amplified where hiring is already difficult.

These are not four different stories. They are four camera angles on the same event.

What It Actually Means

The gap has a shape, and it's not what most commentary assumes

The standard narrative is: AI is advancing fast, workers can't keep up, we need more training. That narrative is wrong in two important ways.

First, it frames the gap as a skills deficit when the data shows it's primarily a leadership deficit. Workers are not resisting AI. The Milken-Harris poll found that 69% of workers believe AI can create more opportunities than it eliminates with the right approach. The Study.com State of AI Jobs and Skills Report (May 12) found that 54% of employees want to improve their AI performance, and 67% say they need two hours or fewer per week to do so. When employers make training available, 70% of workers complete it (WEF, January 2026). The workforce is willing. The structure isn't there.

Second, it frames the gap as a training problem when the data shows it's also a trust and communication problem. The Adecco finding that only 36% of leaders have a talent strategy that clearly demonstrates AI will create opportunities — and only 39% are involving employees in job redesign — is not about course catalogues. It's about whether workers believe their employer has a plan that includes them. When 45% of leaders expect AI agents in workflows within a year but only 30% of workers share that expectation, the 15-point gap is not a knowledge gap. It's a credibility gap.

The "AI-washing" admission is the most important number in the dataset

The Milken-Harris finding that 85% of business leaders feel pressure to appear further along — and that most believe their industry peers would be caught AI-washing if investigated — is the kind of statistic that should make boards uncomfortable. It means the public-facing AI narrative at many organisations is systematically ahead of the internal reality.

This creates a vicious cycle. Leaders overstate readiness externally → internal teams feel pressure to maintain the appearance → actual workforce preparation gets deprioritised because it's slower and less visible than press releases → the gap widens → leaders feel more pressure to appear further along.

The Aon data confirms the downstream effects: 81% of employers cite operational efficiency as a key AI objective, 80% cite automating routine tasks, and only 35% cite workforce upskilling and reskilling. The stated priorities and the actual investment are misaligned by roughly 2.3:1.

The governance vacuum is a time bomb

Only 28% of organisations have fully operational AI guidelines with oversight mechanisms (Aon). Less than half have a team responsible for AI governance. Fewer than a quarter have a Head of AI.

This matters because the EU AI Act's AI literacy requirements are now in effect, obligating employers to ensure staff have sufficient AI literacy. Organisations deploying AI agents into workflows without governance frameworks, without workforce training, and without a Head of AI are not just taking business risk — they're accumulating regulatory exposure that will surface at the worst possible moment: when something goes wrong.

Why the Gap Exists: Ambition vs. Readiness

The gap is not mysterious. It has three structural causes, and they compound.

1. Deployment cycles are shorter than human cycles

AI tools can be procured and deployed in weeks. Building workforce capability — genuine, role-specific, applied AI literacy — takes months to quarters. The organisation's technical velocity and its human velocity are operating on different clocks. As Adecco CEO Denis Machuel put it: "AI may move at software speed, but organisational trust moves at human speed."

This is not a problem that faster training solves. It's a problem that sequencing solves. Organisations that deploy AI and then ask "how do we train people on this?" have already inverted the order. The organisations that close the gap — what BCG calls "AI Leaders" — integrate workforce preparation into the deployment timeline, not after it.

2. Efficiency is easier to measure than capability

81% of employers can point to a cost-saving or time-saving metric from AI. Far fewer can point to a workforce capability metric. Efficiency has a CFO-friendly number attached. Workforce readiness doesn't — or at least, not one that shows up in the same quarterly cycle.

This creates an incentive asymmetry. The executive who champions an AI deployment that saves 20% on a process gets recognised. The executive who champions a workforce upskilling program whose returns materialise over 18 months has a harder story to tell. The result is predictable: deployment gets funded, preparation gets deferred.

3. Workers are being informed, not involved

The Adecco finding that only 39% of leaders involve employees directly in job redesign is the mechanism behind the trust gap. When AI is done to workers rather than with them, two things happen. Workers disengage from the process — and they disengage from the training, because they don't see a path from the training to a role that still includes them. And leaders miss the operational intelligence that only frontline workers possess about which tasks AI should augment versus replace.

The 3-Step Bridge: Audit, Upskill, Redesign

The research points to a coherent three-step sequence. Organisations that execute all three — in order — close the gap. Organisations that skip steps or reorder them widen it.

Step 1: Audit — Know What You Actually Have

Most organisations cannot answer basic questions about their AI workforce readiness. The DataCamp 2026 survey of 500+ enterprise leaders found that only 23% can accurately measure AI ROI. The Study.com report found that only 32% of employees have a clear standard of what good AI use looks like.

An audit is not a survey. It is a structured assessment across four dimensions:

Tool adoption: Which AI tools are actually being used, by whom, for what, and at what frequency? Shadow AI — tools adopted without IT approval — is often the largest category.
Skill levels by role: Not "do you understand AI?" but "can you evaluate AI output for accuracy in your specific job context?" The Study.com data identifies four universal competencies: output evaluation (only 44% confident), prompt construction, task decomposition (only 34% confident), and safe/compliant use (only 30% confident — and the highest-risk gap).
Governance coverage: Do you have AI usage policies? Do employees know what they are? Are they followed? Only 28% of organisations have fully operational guidelines (Aon).
Trust and sentiment: Do workers believe AI will create opportunities for them, or eliminate them? The Adecco data shows only 30% of workers expect AI agents in their workflows within a year — meaning 70% either don't see it coming or don't believe it includes them.

The output of the audit is not a report. It's a baseline against which every subsequent action is measured.

Step 2: Upskill — But Design It Properly

The DataCamp research is blunt: "Most organisations are not failing to offer AI training. Rather, they are failing to design it effectively." The Study.com data confirms: only 18% of employees who received training said it prepared them to work independently. 25% said it helped with simple tasks only.

Effective upskilling has five characteristics, drawn from the converging research:

Role-specific, not generic. The WEF (January 2026) found that personalised, well-designed training tied to business goals produces strong engagement. The DataCamp survey found that 23% of leaders say third-party training paths aren't tailored to specific roles. A salesperson needs different AI skills than a compliance officer.
Hands-on, not video-based. 23% of leaders say video-based courses make it difficult to apply skills. Organisations with hands-on practice environments see 2.7x higher proficiency gains (Iternal, citing BCG data). Watching AI explained is not the same as using AI effectively.
Short and embedded. 67% of employees say they need two hours or fewer per week. Modules of 30–60 minutes tied to specific job tasks outperform half-day workshops. The most effective programs embed AI learning directly into daily workflows — teaching employees to use AI on their real tasks, not hypothetical scenarios.
Accuracy practice built in. The highest-reported skill gap across all employees is output evaluation — knowing when AI is wrong. Only 44% feel confident in it. Every training module should include evaluation and fact-checking components.
Progressive, not one-off. Begin with output evaluation and safe use. Build toward task decomposition and complex application. One-off sessions produce awareness without confidence, and adoption without judgment.

Step 3: Redesign — Involve Workers in the Process

This is the step most organisations skip, and it's the one that determines whether the first two steps produce lasting results or temporary compliance.

Job redesign means answering: Given what AI can now do, what should humans do instead? It is not a question that can be answered by leadership alone. The Adecco finding that only 39% of leaders involve employees in job redesign is not just a fairness problem — it's an information problem. Frontline workers know which tasks AI handles poorly, which edge cases matter, and which workflows will break if a step is automated without understanding its downstream dependencies.

The Milken-Harris data shows that 88% of business leaders agree individual companies cannot solve AI workforce readiness alone — a coordinated national response is required. But the company-level actions that matter most are: (a) making the talent strategy explicit about how AI creates opportunities, not just efficiencies; (b) involving employees in redesign before deployment; and (c) communicating the plan clearly enough that the 30% who currently expect AI agents in their workflow becomes 70%.

Stakeholder Landscape

Stakeholder	Position	What the Data Says
C-suite leaders	Accelerating AI deployment; 45% expect agents in 12 months	Only 22% confident in workforce readiness; 85% feel pressure to appear further along
Middle managers	Caught between deployment mandates and team readiness	Not directly surveyed in these studies — a significant blind spot
Frontline workers	Willing but unsupported; 41% received zero AI support	69% believe AI can create opportunities; 54% want to improve; 67% need ≤2 hours/week
HR/L&D functions	Under-resourced for the scale of the challenge	Only 35% of orgs cite upskilling as primary AI objective; <25% have Head of AI
Regulators	AI literacy requirements now in effect (EU AI Act)	Only 28% of orgs have fully operational AI governance
Technology vendors	Benefiting from deployment velocity	Incentivised to sell tools, not workforce transformation — a structural tension

Cross-Layer Implications

Security. When 73% of organisations deploy AI but only 28% have governance frameworks, shadow AI becomes the dominant risk surface. Employees using unapproved tools with sensitive data — because they haven't been trained on compliant alternatives — is not a hypothetical. It's the default state.

Talent market. The Aon finding that only 28% of organisations have hired AI expertise, combined with the Adecco finding that only 36% have a talent strategy showing AI creates opportunities, means the external AI talent market and the internal workforce strategy are operating in parallel universes. Organisations are competing for scarce AI hires while failing to build AI capability in the workforce they already have.

Regulatory. The EU AI Act's AI literacy provisions are not aspirational — they are obligations. Organisations deploying AI without workforce training are accumulating compliance debt that will crystallise at the first incident.

Wellbeing. The Aon report found that 84% of employers are confident their wellbeing strategy meets workforce needs, yet 72% of employees report high stress. Organisations with fully deployed AI were twice as likely to describe leadership's wellbeing commitment as "strong and visible." The implication is uncomfortable: AI deployment may correlate with better wellbeing outcomes not because AI reduces stress, but because organisations that are competent enough to deploy AI well are also competent enough to manage people well. The causality runs through organisational capability, not technology.

What This Means for Leaders

If you are a CEO or board member

Your AI narrative is probably ahead of your AI reality. The 85% pressure-to-appear-further-along statistic is a mirror. The most valuable thing you can do this quarter is close the say-do gap: commission an honest internal audit of AI workforce readiness, make the results visible to the board, and tie a portion of executive compensation to closing the gap — not to deployment velocity.

If you are a CHRO or Head of People

You have more leverage than you think. The data shows that when employers provide structured, role-specific, hands-on AI training, 70% of workers complete it. The workforce is not the bottleneck. Your challenge is designing programs that are applied (not video-based), role-specific (not generic), and progressive (not one-off). Start with output evaluation and safe use — the two competencies with the lowest confidence and the highest risk.

If you are a line manager

You are the most important person in this equation and the least studied. None of the four major reports surveyed middle managers directly. Your team's AI readiness depends on whether you protect time for learning, model AI use yourself, and involve your people in redesign decisions. The 67% of workers who say they need two hours or fewer per week to improve — give them those two hours. Protected. On the calendar.

If you are a worker

The data says your employer is probably not going to provide structured AI training in the near term. 41% of workers received zero support in the past year. The four competencies to develop independently: output evaluation (can you tell when AI is wrong?), prompt construction (can you get consistent results?), task decomposition (can you break your work into AI-manageable steps?), and safe use (do you know what data you can and cannot feed into AI tools?). Start with output evaluation. It's the highest-reported gap and the most career-durable skill.

Uncertainty Ledger

The 45% leader expectation may be aspirational, not predictive. The Milken-Harris finding that 85% of leaders feel pressure to appear further along suggests some of the 45% may be expressing ambition rather than concrete plans. If actual agent deployment in 12 months is closer to 25–30%, the gap narrows — but the credibility problem remains.
Middle managers are the missing data point. None of the four studies surveyed middle managers specifically. Their experience — caught between C-suite deployment mandates and team readiness — is likely the most predictive variable for whether AI adoption succeeds or fails at the operational level.
The 70% training completion rate (WEF) may reflect selection bias. Workers who opt into employer-provided AI training are likely already more motivated and technically comfortable. The harder population — those who haven't engaged — may require different approaches that the current data doesn't capture.
Industry rankings (Lightcast) measure relative preparedness, not absolute risk. Hospitality ranks worst on AI skills gap, but its market risk score (2.49) is moderate. Energy & Resources has a lower skills gap (3.52) but the highest market risk (3.47). The interaction between these two scores — not either alone — determines urgency.

Bottom Line

Four major studies in four weeks have produced a convergence that is rare in social science: the AI workforce gap is real, it's widening, and its primary cause is not worker resistance or technological velocity — it's a leadership credibility deficit. Leaders are deploying AI faster than they're preparing people, overstating their readiness externally, and under-involving workers in the redesign of their own jobs. The fix is not more training content. It's a sequenced commitment to audit honestly, upskill properly (role-specific, hands-on, progressive), and redesign with workers at the table. Organisations that close the say-do gap in 2026 will be the ones that still have a workforce worth deploying AI for in 2028.

Sources

Adecco Group, The Human Premium: Leadership Beyond the Algorithm, May 21, 2026. Survey of 2,000 C-suite executives across 13 countries. [Tier 1 — PR Newswire / primary research]
Aon, Human Capital Trends Report, May 5, 2026. Global survey of board directors and senior business and people leaders. [Tier 2 — industry analyst firm]
Milken Institute-Harris Poll, Bridging the Technology Advancement Gap: A New Consensus for the AI Era, May 4, 2026. Survey of 2,001 American adults including 1,280 workers and 502 business leaders (VP+, $2B+ revenue). [Tier 2 — established think tank and polling firm]
Lightcast / Resume Now, 2026 AI Workforce Preparedness Rankings, April 24, 2026. Analysis based on Lightcast's Workforce Risk Outlook data. [Tier 2 — labour market data provider]
DataCamp, The AI Skills Gap in 2026: Why Most AI Training Isn't Translating to Workforce Capability, March 2026. Survey of 500+ enterprise leaders. [Tier 3 — vendor research, useful with care]
Study.com, State of AI Jobs and Skills Report 2026, May 12, 2026. [Tier 3 — vendor research, useful with care]
World Economic Forum, The AI Perception Gap: How to Ensure Employers and Workers Are Aligned, January 2026. [Tier 2 — international organisation]
Iternal, AI Skills Gap 2026: Statistics, Causes & How to Close It, April 2026. Aggregates BCG, McKinsey, Gartner data. [Tier 3 — aggregator, used for framework synthesis only]