AI Fallibility Meets the Digital Skills Gap: The Hidden Barrier to Enterprise AI Adoption

Artificial intelligence is advancing faster than many organizations—and many workforces—can comfortably absorb it.

 

A recent BBC article examining why Japanese companies have been relatively slow to adopt AI highlights an uncomfortable reality: AI adoption is not simply a technology problem. It is a trust, skills and organizational-change problem.

 

Japan is particularly interesting because it combines several forces that many other economies are beginning to experience simultaneously:

 

1. an ageing workforce,
2. labor shortages,
3. highly experienced employees,
4. relatively cautious corporate cultures,
5. uneven digital skills—and
6. a low tolerance for technologies that occasionally produce incorrect results.

 

That last point matters.

 

Generative AI is probabilistic. It can be extremely useful without being infallible.

 

For industries in which errors can affect project cost, procurement, contracts, safety or profitability, a small AI fallacy rate can therefore create a disproportionately large trust problem.

 

But Japan is not alone.

 

The Japan Challenge Is Appearing Elsewhere

 

OECD research confirms the scale of Japan’s challenge.

 

Only 8.4% of Japanese employees report using AI at work, while older workers are less likely to use AI, participate in AI-related training or trust AI systems deployed by their employers. Only around 30% of Japanese AI users report receiving company-provided AI training, and Japan recorded the lowest level of employer-AI trust among the eight countries included in the OECD comparison.

 

What makes this important internationally is that similar combinations of demographic pressure, digital-skills gaps and organizational caution are emerging elsewhere.

 

South Korea: perhaps the closest Asian parallel

 

South Korea may be the strongest parallel to Japan.

 

It combines rapid population ageing with one of the world’s most technologically sophisticated economies. Yet sophisticated national infrastructure does not automatically translate into equal digital capability across generations.

 

Recent OECD analysis finds that participation in job-related training among workers aged 55–65 is well below 20% in Korea, while the gap in the use of specialized software between workers aged 55–65 and those aged 25–54 exceeds 35%—one of the largest age-related digital gaps identified by the OECD.

 

AI adoption in Korean businesses also remains uneven. OECD estimates indicate that AI adoption among Korean SMEs is higher than Japan’s but still significantly below countries such as Germany and Ireland. At the same time, more than 80% of Korean employment is concentrated in SMEs, making workforce adoption particularly important.

 

The lesson is important:

 

A technologically advanced economy can still have a workforce-adoption problem.

 

Southern and Southeastern Europe: ageing meets weak lifelong learning

 

A similar challenge exists across parts of Southern and Southeastern Europe, including Greece, Italy, Portugal, Spain and several Balkan economies.

 

The issue is not that these countries lack capable professionals.

 

Quite the opposite.

 

They contain enormous accumulated industrial, engineering and managerial expertise.

 

The challenge is transferring that expertise into increasingly digital operating environments.

 

Across the EU, around 35% of workers aged 55–64 participate in job-related training compared with 48% of workers aged 35–54. In Greece and Türkiye, participation among older workers falls below 10%.

 

The European Commission also warns that the employment benefits of AI are likely to be distributed unevenly, favoring highly skilled and prime-age workers unless targeted measures address weaker regions and more vulnerable groups.

 

Recent European evidence reinforces the same message: AI adoption is significantly influenced by workers’ skills, training availability and participation in organizational decisions—not simply whether an AI tool is technically available.

 

For enterprise digital transformation, this matters enormously.

 

An experienced 58-year-old estimator who understands construction economics better than most younger engineers is not a problem to be automated away.

 

That employee represents decades of corporate knowledge.

 

The objective should be to amplify that knowledge through technology.

 

Central and Eastern Europe: another emerging divide

 

Parts of Central and Eastern Europe face a related challenge.

 

The World Bank is actively developing digital-skills and AI-readiness programs across Europe and Central Asia because access to technology alone is proving insufficient. Recent programs in Romania, Türkiye, Bulgaria and elsewhere focus specifically on building the skills required for an AI-ready workforce.

 

OECD research similarly identifies several Southern, Central and Eastern European countries where older workers participate comparatively little in workplace training.

 

This means enterprise AI adoption risks becoming two-speed:

 

digitally confident employees accelerate while experienced but less digitally fluent employees become increasingly disconnected from the corporate operating system.

 

That would be a serious strategic mistake.

 

Central Asia: the problem is skills more than ageing

 

Central Asia requires a different interpretation.

 

Countries such as Kazakhstan and Uzbekistan do not currently replicate Japan’s demographic structure to the same degree.

 

Their challenge is primarily digital capability, institutional readiness and workforce reskilling.

 

The OECD has identified adult-learning participation and effective skills utilization as important issues for Kazakhstan, while the Asian Development Bank has emphasized that Uzbekistan’s ability to capture the benefits of automation and AI depends heavily on training and digital-skills development.

 

This distinction matters.

 

In Japan, Korea and parts of Europe, companies increasingly need to digitally enable an ageing workforce.

 

Across much of Central Asia, companies need to rapidly build digital capability across a workforce operating in economies that are digitizing quickly.

 

The implementation response may therefore differ—but the enterprise requirement is similar:

 

technology must adapt to people, rather than expecting every employee to become a technologist.

 

AI Fallibility Makes the Adoption Problem Harder

 

Now add the second dimension identified by the debate around AI adoption:

 

AI makes mistakes.

 

This creates an unusual psychological problem.

 

• Organizations routinely tolerate human error.
• An estimator can miscalculate.
• A buyer can select the wrong supplier.
• A site engineer can enter an incorrect quantity.
• A project manager can make a forecasting error.

 

Yet employees often expect artificial intelligence to perform at a much higher standard.

 

Once users witness an AI hallucination or incorrect recommendation, trust can fall rapidly.

 

The answer cannot therefore be:

 

“Trust the AI.”

At DANAOS Projects, we believe a better principle is:

 

Don’t Trust AI Blindly. Control It.

 

Enterprise AI should be designed with the assumption that AI can occasionally be wrong.

 

That means moving from:

 

AI → Decision

 

to:

 

AI → Recommendation → Validation → Approval → Execution → Audit

 

This is particularly important in construction, EPC, shipbuilding, offshore, marine, mining and other project-based industries.

 

If an AI agent recommends purchasing €500,000 of steel, the organization should not depend exclusively on the language model’s judgment.

 

The system should be capable of verifying:

 

• Is this material included in the BoQ?
• Which WBS activity requires it?
• When is it required?
• Which Cost Code will absorb the expenditure?
• What quantity was budgeted?
• How much has already been consumed?
• Is there stock available?
• What prices were previously paid?
• Which suppliers are approved?
• Is sufficient budget available?
• Who has authority to approve the transaction?

 

This is where an ERP becomes essential infrastructure for AI.

 

Why ProjectVIEW ERP Changes the Trust Equation

 

ProjectVIEW ERP connects business transactions through three fundamental project dimensions:

 

BoQ ↔ WBS ↔ Cost Codes

 

This creates structured business context around operational decisions.

 

Procurement, budgeting, subcontractors, materials, machinery, labor, progress and cost control are not isolated AI prompts. They exist inside a common project environment with users, permissions, workflows and approvals.

 

ProjectVIEW already provides centralized role-based access, workflows, transactional logging and traceability. Every process can therefore be executed according to the responsibilities and authority of the employee involved. AI can consequently operate inside an enterprise control system rather than outside it.

 

That is a fundamental difference.

 

• An AI agent may recommend.
• ProjectVIEW provides the business rules.
• The responsible employee validates.
• The workflow controls authorization.
• The ERP records what happened.

 

Controlled Fallibility

 

We believe this leads to an important concept for enterprise AI:

 

Controlled Fallibility.

 

AI does not have to be perfect to deliver enormous value.

 

It needs to operate inside an environment capable of identifying, restricting and correcting mistakes.

 

The degree of autonomy should therefore correspond to the consequence of the decision.

 

For example:

 

• Low risk: AI summarizes project documents automatically.
• Moderate risk: AI prepares a material requisition for employee verification.
• Higher risk: AI recommends supplier selection but requires procurement approval.

 

Financial commitment: AI prepares the transaction but authorized management approves execution.

 

The objective is not maximum automation.

 

It is maximum useful automation within acceptable risk.

 

And What About Employees Who Are Not Tech-Savvy?

 

This is arguably where AI can become most valuable.

 

Traditional enterprise software asks employees to learn the software.

 

• Which menu?
• Which screen?
• Which fields?
• Which report?
• Which workflow?

 

That approach becomes increasingly problematic when companies have employees with vastly different levels of digital fluency.

 

AI allows us to reverse the relationship.

 

Instead of teaching employees how to speak ERP, we can increasingly teach the ERP how to understand employees.

 

Imagine an experienced site manager saying:

 

“We need another 20 tonnes of reinforcement steel for the retaining wall next week.”

 

The employee does not need to understand an AI model.

 

The AI assistant can interpret the request and ProjectVIEW can provide its enterprise context:

 

project → BoQ → WBS → material → warehouse → budget → supplier → workflow.

 

The system prepares the transaction.

 

The employee verifies it.

 

The authorized manager approves it.

 

This is a dramatically easier path toward adoption for employees who are experts in construction, not experts in software.

 

DANAOS Implementation Strategy: Adoption Before Automation

 

Technology architecture alone will not solve this problem.

 

Implementation strategy matters just as much.

 

This is particularly relevant in Japan, Korea, Southern Europe, the Balkans, Central Asia and other markets where change may need to be introduced progressively.

 

DANAOS implementations begin with As-Is and To-Be process analysis, GAP analysis and design blueprints, rather than forcing every organization into a generic implementation template. Training plans, change management, UAT, Go-Live strategy and post-implementation support are formal project deliverables.

 

The objective is to understand how employees currently work before changing how they work.

 

1. Train by role—not by software

 

An estimator does not need to learn everything ProjectVIEW can do.

Neither does a warehouse manager.

 

Neither does a CFO.

 

DANAOS uses role-based training tracks so users learn the processes relevant to their actual jobs. The current ProjectVIEW framework includes a formal Training Needs Assessment that considers ERP proficiency, organizational role, business processes, language requirements and user availability before defining each training program.

 

That is particularly important for less digitally experienced employees.

 

Training becomes:

 

“How do I perform my job?”

 

rather than:

 

“How do I operate this complicated ERP?”

 

2. Hands-on simulation before Go-Live

 

Users should not learn mission-critical technology for the first time in production.

 

ProjectVIEW training uses dedicated sandbox environments with realistic business scenarios, allowing employees to practice without affecting live transactions.

 

DANAOS implementations also incorporate a simulation period followed by hands-on support during deployment.

 

This is especially valuable for employees who are uncomfortable with new technology.

 

Confidence comes from repetition, not PowerPoint presentations.

 

3. Create internal Champions

 

One of the most effective ways to overcome resistance to change is not to make IT responsible for adoption.

 

It is to create trusted users inside each operational department.

 

The DANAOS ProjectVIEW ERP Champions program develops senior power users capable of supporting colleagues, understanding cross-module workflows, escalating issues and communicating with DANAOS after Go-Live.

 

An experienced procurement manager is usually more likely to trust another procurement professional than an external AI consultant.

 

This creates peer-led digital adoption.

 

4. Keep humans in the workflow

 

ProjectVIEW allows organizations to retain approval authority exactly where they need it.

 

Users and roles determine access rights, workflows determine authorization, and transactional history provides traceability.

 

AI automation can therefore be introduced gradually:

 

AI informs → AI recommends → AI prepares → AI executes within approved boundaries.

 

There is no requirement to jump immediately from manual work to autonomous operation.

 

5. Make learning continuous

 

ERP and AI adoption cannot be treated as a one-time training event.

 

ProjectVIEW supports self-paced digital learning, recorded training, online documentation, in-application guidance and role-specific refresher training. This is particularly important for ageing workforces because employees can learn at different speeds without being excluded from digital transformation.

 

6. Validate technology before trusting it

 

The same philosophy applies to AI itself.

 

ProjectVIEW’s current SaaS framework incorporates controlled testing environments, User Acceptance Testing and regression validation before production changes. Major changes require UAT and customer sign-off before production cutover.

 

This is precisely how organizations should approach AI.

 

Trust should be demonstrated—not demanded.

AI Could Actually Help Ageing Workforces

 

There is an important upside that is often missed.

 

AI does not necessarily threaten experienced workers.

 

It could become one of the technologies that allows them to remain productive longer.

 

The OECD notes that user-friendly AI applications may be easier for older workers to adopt than sophisticated specialist software because conversational systems can complement existing expertise.

 

This is potentially transformative.

 

The future of enterprise software may therefore move away from forcing experienced workers to memorize increasingly complex software interfaces.

 

Instead, employees communicate naturally while AI translates their intentions into structured enterprise processes.

 

That makes AI potentially more inclusive than the previous generation of enterprise technology, provided it is implemented correctly.

 

The Real Competitive Advantage Is Not AI

 

The companies that win the AI race will not necessarily be those that deploy the most AI agents.

 

They will be the organizations that successfully connect:

 

human expertise + reliable enterprise data + AI automation + business rules + governance.

 

This is particularly important across ageing economies such as Japan, South Korea and Southern Europe, and increasingly important across Central and Eastern Europe and rapidly digitizing Central Asian markets.

 

For DANAOS Projects, this defines the direction of ProjectVIEW ERP.

 

We envision ProjectVIEW as the central command center and orchestrator of project operations, weaving the enterprise data fabric while purpose-specific AI agents accelerate individual processes.

 

The ERP provides context and control.

 

AI provides speed and intelligence.

 

Employees provide experience and judgment.

 

DANAOS provides the implementation, training and change-management framework that brings the three together.

 

The objective is not to make experienced employees adapt to AI overnight.

 

And it is certainly not to pretend that AI will become infallible.

 

The objective is much more practical:

 

Make AI easy enough to use, controlled enough to trust and valuable enough that employees choose to adopt it.

 

Because the hidden barrier to enterprise AI is not the model.

 

It is whether the people who actually run the business are prepared to use it.