Companies are spending millions bringing artificial intelligence into the workplace.

Microsoft Copilot.
ChatGPT.
Claude.
Gemini.

Governance frameworks are being created. Data protection requirements are being defined. Employees are encouraged to experiment.

And yet I keep observing something strange:

A lot of people still don’t really use AI.

Some don’t know how.

Some are afraid to ask.

Others use it quietly because they don’t want colleagues or managers to know.

And perhaps the most dangerous group uses it confidently — without checking whether the answer is actually correct.

This is where the real AI transformation problem begins.

It isn’t primarily a technology problem.

It’s a human problem.

The AI Nobody Wants to Admit They’re Using

I increasingly see a strange contradiction in the workplace.

Organizations want employees to use AI.

Employees know AI is becoming important.

But many still seem uncomfortable admitting that they need it.

Why?

Maybe using AI feels like admitting a weakness.

Maybe asking how Copilot works feels embarrassing when everyone around you seems to understand it already.

Maybe someone worries:

“If my manager realizes AI helps me do this, will they start wondering why they need me?”

That fear isn’t imaginary.

Microsoft and LinkedIn’s 2024 Work Trend Index found that 52% of people using AI at work were reluctant to admit using it for their most important tasks. Another 53% worried that using AI for important work could make them appear replaceable.

That tells us something important.

Giving employees access to AI does not automatically create AI adoption.

Trust matters.

Culture matters.

Training matters.

And psychological safety matters.

Then There Is the Opposite Problem

Some employees have already discovered AI.

But they make another mistake.

They trust it too much.

They ask a question.

Copy the answer.

Paste it into an email, presentation, report or analysis.

Done.

Except sometimes it isn’t done.

The answer contains assumptions presented as facts.

A source doesn’t exist.

A link leads nowhere.

The conclusion sounds convincing but is based on speculation.

Or the writing suddenly sounds nothing like the person supposedly writing it.

And that’s when AI becomes visible for all the wrong reasons.

This is something I find particularly important from a Governance, Risk and Compliance perspective:

AI output is not accountability.

The human using the system remains responsible for understanding, challenging and validating the result.

AI can accelerate your thinking.

It should not replace it.

AI Should Make Your Capabilities Stronger

This is where I believe we’re missing one of the greatest opportunities of the current AI transformation.

AI should not make humans less capable.

It should make us more capable.

Think about a normal working day.

Emails.

Meetings.

PowerPoint presentations.

Excel files.

Research.

Documentation.

Summaries.

Planning.

Searching through documents.

Preparing a first draft.

These are exactly the kinds of activities where AI can remove friction.

And we already have evidence that it can.

In Microsoft’s early research with Microsoft 365 Copilot, users completing a combination of searching, writing and summarizing tasks were 29% faster than people performing the same tasks without Copilot.

Seventy percent of users reported being more productive, and 85% said Copilot helped them reach a good first draft faster.

This is not about replacing someone’s expertise.

It is about reducing the time between having expertise and applying it effectively.

Thirty Minutes Changes More Than You Think

Microsoft and LinkedIn later identified what they called AI “power users.”

These weren’t necessarily programmers.

They were people who had developed habits around using AI.

They experimented.

They tried different prompts.

They asked themselves before starting a task:

Could AI help me with this?

And importantly, they didn’t stop when the first response wasn’t perfect.

Power users reported saving more than 30 minutes per day.

Thirty minutes doesn’t sound revolutionary.

But multiply it.

Five days a week.

Around 220 working days per year.

That can represent more than 100 hours.

Now multiply that across hundreds or thousands of employees.

Suddenly AI literacy isn’t an IT topic anymore.

It becomes a productivity and competitiveness issue.

The Skills Gap Is Already Becoming Visible

There is another part of this transformation that worries me more.

I meet people who genuinely want to learn.

But they are overwhelmed before they even begin.

They hear:

LLMs.

Agents.

Prompt engineering.

RAG.

Claude Code.

APIs.

Automation.

And they think:

“This isn’t for me.”

But that’s where I think the AI industry sometimes gets it wrong.

A normal employee doesn’t need to understand how to build an AI agent on day one.

They need to understand:

How can this help me tomorrow morning?

How can Copilot help me summarize a Teams meeting?

How can it help me structure a Word document?

How can I analyze information in Excel?

How can I prepare a PowerPoint presentation?

How can I improve an email without losing my own voice?

How do I check whether an AI-generated statement is true?

And perhaps most importantly:

What information am I allowed to give the AI — and what information must remain protected?

That’s where AI education should begin.

Not with hype.

With everyday work.

We Don’t Need More Fear. We Need AI Literacy.

YouTube is full of incredible AI content.

New models.

Agents.

Coding tools.

Automations.

Benchmarks.

The problem is that much of this content is already several levels ahead of the average employee.

Meanwhile, millions of people are still trying to understand the first level.

And I believe we’re underestimating them.

They don’t need another keynote telling them that AI will transform everything.

They need someone sitting beside them saying:

Open the application. Let’s solve one of your actual tasks together.

That changes everything.

Companies Have Responsibility Too

Organizations cannot simply deploy an AI license and call that transformation.

Technology adoption requires more than access.

It requires:

• clear governance
• understandable rules
• data protection
• practical training
• role-specific examples
• permission to experiment
• room to make controlled mistakes
• continuous learning

The Microsoft/LinkedIn research provides another interesting signal: AI power users were more likely to work in organizations where leadership actively communicated the importance of AI, encouraged experimentation and provided role-specific training.

That matters.

Because transformation doesn’t happen when IT activates a license.

Transformation happens when behavior changes.

But Employees Have Responsibility Too

There is another side to this.

Companies can provide training.

Managers can encourage experimentation.

Governance teams can establish safe boundaries.

But eventually, every individual has to make a decision.

Do I learn — or do I wait?

Education doesn’t end when school ends.

It doesn’t end after university.

And it certainly doesn’t end after getting a good job.

AI makes continuous learning even more important.

I would go further:

Learning how to work with AI is increasingly becoming part of professional responsibility.

Not because everyone needs to become an AI expert.

But because the tools surrounding our professions are changing.

Will AI Take Your Job?

I think we’re asking the wrong question.

The question isn’t simply:

Will AI replace my job?

A better question is:

What happens when someone doing the same job learns to use AI significantly better than I do?

That’s the comparison that matters.

A person with professional experience, judgment and AI capability can potentially work faster, research more broadly, prepare better first drafts and spend more time on higher-value decisions.

That’s difficult to compete with by simply refusing to use the technology.

This doesn’t mean every employee who doesn’t use AI today will lose their job tomorrow.

But ignoring the development entirely is becoming a professional risk.

And risk is rarely about what happens today.

Risk is about recognizing what could happen tomorrow early enough to act.

AI Should Give Us Something Back

This is the part of the AI discussion I care about most.

I don’t want a future where humans become machines.

I want almost the opposite.

For decades, technology has promised to make our lives easier.

Yet somehow we ended up with more emails, more meetings, more notifications, more administration and more digital noise.

AI gives us another opportunity.

Used properly, it can take away some of that mechanical work.

Not so that we can squeeze another 30 tasks into the day.

But so that we can invest some of that time back into what technology cannot replace easily:

Thinking.

Listening.

Creativity.

Relationships.

Leadership.

Judgment.

And simply being human.

That, to me, is one of the most exciting possibilities of artificial intelligence.

Start Smaller

We don’t need to turn every employee into an AI engineer.

We need to lower the barrier to entry.

Start with one task.

One application.

One real problem.

Then repeat.

A 15-minute micro-training.

A practical Copilot exercise.

A safe environment to ask what might feel like a stupid question.

A clear explanation of what data can and cannot be used.

And then another exercise next week.

Small improvements compound.

Eventually, AI stops feeling like a foreign technology.

It becomes another professional tool.

Just like Excel.

Just like PowerPoint.

Just like email once did.

My Conclusion

The companies that succeed with AI won’t simply be the companies with the most advanced models.

And the employees who succeed won’t necessarily be the most technical people.

I believe the advantage will go to organizations and individuals who learn how to combine three things:

Human expertise.
AI capability.
Responsible judgment.

Technology provides the capability.

Governance provides the boundaries.

Education provides the confidence.

But the human still has to take the first step.

And perhaps that is the real AI challenge nobody wants to talk about.

We already have incredibly powerful tools.

Now we need to teach people not to be afraid of using them — and not to be afraid of thinking for themselves while they do.

---

Christian Hirschmann

Governance | Risk | Compliance | AI | Digital & Operational Resilience