AI will expose weak leadership before it replaces workers

This week, I want to move the question upwards.

We spend a lot of time asking what AI will do to workers. But we may not be asking enough what AI will do to leadership. Because AI will not only change the work. It will expose the people responsible for directing, developing and improving that work.

My view is simple. AI will expose weak leadership before it replaces workers.

AI can accelerate production. Leaders must still define progress.

AI is already changing the baseline

This is not a theoretical shift anymore. Microsoft and LinkedIn's 2024 Work Trend Index reported that 75% of global knowledge workers were already using AI at work, and that 78% of AI users were bringing their own AI tools into the workplace.[1]

That matters. It means AI adoption is not waiting for the perfect board-approved strategy. It is already happening inside teams. People are using AI to draft. To summarise. To prepare. To rewrite. To explain. To compare. To get started. To improve.

That changes what employees can reasonably produce. It also changes what they should reasonably expect from leadership.

If AI helps people move faster through the basics, then leaders cannot keep adding value only at the basic level. The leadership bar has to move up.

Low-level oversight becomes less valuable

Many leaders and managers have historically added value through oversight. Review this deck. Fix this email. Clean this proposal. Send me a status update. Summarise the meeting. Format the document. Make the argument clearer. Turn this into something presentable.

Some of that still matters. But much of it is becoming less differentiated. AI can help with first drafts, summaries, tone, structure, risks, meeting notes and turning rough points into cleaner language.

So the leader who only adds value by correcting the basics becomes more exposed. Not useless. Exposed.

Because the team may increasingly need less help with basic production and more help with context, judgement and direction. That is where leadership becomes more important, not less.

The evidence says AI can raise the floor

The idea that AI can lift baseline productivity is supported by serious research.

A Harvard Business School and Boston Consulting Group study found that consultants using GPT-4 completed 12.2% more tasks, completed them 25.1% faster, and produced work rated more than 40% higher in quality on tasks within AI's capability range.[2]

A separate study published in the Quarterly Journal of Economics found that generative AI assistance increased productivity by 15% on average among more than 5,000 customer-support agents, with the largest gains among less-experienced and lower-skilled workers.[3]

The pattern is important. AI can help more people reach a stronger starting point. It can raise the floor.

But once the floor rises, leadership has to move. A leader cannot simply sit one level above basic output if basic output is now easier to produce. The leader has to help people move from output to value.

The jagged frontier makes leadership more important

The productivity story is not one-directional. The Harvard and BCG study also introduced the idea of the "jagged technological frontier". AI helps on some tasks, but can hurt performance when used outside its capability range and trusted too much.[2]

That is a leadership issue. Because employees will not always know where the frontier sits. They may not know when AI is helping. They may not know when AI is guessing. They may not know when the answer is fluent but shallow, or polished but wrong. They may not know when a human conversation is still required.

This is where weak leadership gets exposed.

A weak leader says: "Use AI and move faster."

A stronger leader asks: "Where does AI help here, where could it mislead us, and what needs human judgement before we act?"

That is a completely different standard. The first drives speed. The second drives better work.

Output is no longer the leadership goal

AI makes it easier to produce more. More documents. More slides. More summaries. More meeting notes. More proposals. More polished emails.

But more output is not the same as more value. This is where leaders can easily celebrate the wrong thing. A team may look busier. A department may produce more. A manager may receive faster drafts. A project may generate more documents. But none of that proves the work is better.

The better leadership questions are different:

• Did this improve the decision?
• Did it clarify the customer problem?
• Did it reduce confusion?
• Did it surface the risk earlier?
• Did it help the team prioritise?
• Did it develop the person doing the work?
• Did it move the business forward?

AI can accelerate production. Leaders must still define progress.

Weak leadership hides behind activity

Weak leadership has always been able to hide behind activity. Meetings. Updates. Documents. Reviews. Process. Status checks. Follow-ups. Decks. Reports. Activity can feel like control. It can also become a substitute for direction.

AI makes this more dangerous, because AI can increase the amount of activity very quickly. A weak leader may ask for more analysis because it is now easier to generate. More options. More scenarios. More summaries. More slides. More versions.

But more analysis does not automatically create a better decision. Sometimes it creates more fog.

Strong leadership is not about asking for more material. It is about helping people decide what matters.

The leadership premium is moving

If AI reduces the value of basic correction and low-level oversight, the leadership premium moves elsewhere.

• Context. People need to understand why the work matters, not only what must be produced.
• Prioritisation. AI can generate many options, but leaders must help teams choose what deserves attention.
• Judgement. Teams need to know what to trust, what to challenge and what to verify.
• Coaching. If AI helps people produce cleaner first versions, leaders need to look beneath the output and develop the thinking.
• Standards. The minimum standard rises when everyone has access to better tools.
• Decision-making. AI can assist the thinking, but leadership still has to own the trade-off.

That is the shift. Leadership becomes less about controlling production and more about improving contribution.

Leaders must make AI use visible

One of the biggest risks is invisible AI use. When employees use AI privately, leaders see only the final output. They do not see the prompt, the first answer, what was accepted, what was rejected, what was verified, or where the person added judgement.

That makes development harder. It also makes risk harder to manage. Microsoft and LinkedIn's Work Trend Index found that many people were bringing their own AI tools to work.[1] That should concern leaders, not because AI use is bad, but because unmanaged AI use creates blind spots.

The answer is not to shame people for using AI. The answer is to make the process visible. Ask:

• What did you use AI for?
• What did it give you?
• What did you change?
• What did you verify?
• What did you reject?
• What did you add that AI would not have known?

That is a better management conversation. It turns AI from a hidden shortcut into a visible learning tool.

Strong leaders coach judgement

This links directly to the previous article in the series. AI can make junior people faster. But leaders still need to make sure they are learning.

That means reviewing more than the final product. A polished AI-assisted output may hide weak thinking. So leaders need to ask people to explain their reasoning. Why did you choose this recommendation? What assumption are you making? What would a sceptic challenge? What evidence supports this? What did you leave out? What would change your mind? Where might AI have been wrong?

These questions develop judgement. They also show whether the person understands the work or simply produced it.

That is one of the biggest leadership responsibilities in an AI-enabled workplace. Not checking whether the paragraph is clean. Checking whether the thinking is strong.

The manager as bottleneck is under pressure

Some managers have historically been bottlenecks. They were the only person who could approve, structure, rewrite the work or make the call. They were the only person who knew the context.

AI puts pressure on that model. If a team member can get a strong first structure from AI, they do not need to wait three days for a manager to help them start. If they can summarise the meeting automatically, they do not need a manager to chase notes. If they can prepare five customer questions in minutes, they do not need a manager to spoon-feed the first version.

That is positive. It frees leaders to do higher-value work. But only if leaders are ready to move up.

The manager who enjoyed being the bottleneck may feel threatened. The leader who wants stronger people will feel relieved.

AI adoption needs leadership, not cheerleading

There is a difference between leadership and cheerleading.

Cheerleading says: "Everyone should use AI."

Leadership says: "Here is where AI helps us, here is where it does not, here are the standards, and here is how we will use it responsibly."

Cheerleading celebrates usage. Leadership defines value. Cheerleading pushes speed. Leadership protects judgement. Cheerleading asks for more output. Leadership asks for better decisions.

That distinction matters because AI adoption without leadership can create chaos. People use different tools. Different standards emerge. Sensitive information may be mishandled. Work becomes more polished but less reliable. Teams produce more, but learn less. Managers lose visibility of how work is being created.

This is not an AI problem. It is a leadership problem.

The organisation has to change too

This is not only about individual leaders. Organisations also need to change how they think about AI.

McKinsey's 2025 State of AI research found that high-performing organisations were more likely to be scaling AI across the business and investing more meaningfully in AI capabilities.[4]

That matters because AI value does not come only from giving people tools. It comes from changing how work is designed, governed and improved. Leaders need to create the conditions for AI to be useful: clear use cases, clear standards, clear governance, clear training, clear accountability and clear expectations for human review.

Without that, AI remains a personal productivity layer. Useful, but limited. The bigger opportunity is not simply that individuals work faster. It is that teams work better.

The role of leaders becomes more human

There is an irony in all of this. The more AI enters work, the more human leadership matters. Not softer leadership. Stronger leadership. Clearer direction. Better prioritisation. Sharper judgement. More coaching. More trust. More accountability. More context. More courage.

AI can generate options. Leaders must choose. AI can summarise information. Leaders must decide what matters. AI can prepare a first draft. Leaders must raise the standard. AI can help people move faster. Leaders must make sure they are moving in the right direction.

That is why AI does not make leadership less important. It makes weak leadership harder to hide.

The question

If AI can help with the basics, what should leaders spend more time doing? Coaching? Prioritising? Setting standards? Developing judgement? Clearing blockers? Creating context? Making better decisions?

Because the future of work will not only test workers. It will test leaders. And the leaders who only managed output may find themselves exposed. The leaders who improve thinking will become more valuable than ever.

---

Sources

[1] Microsoft and LinkedIn, "2024 Work Trend Index: AI at Work Is Here. Now Comes the Hard Part", 2024.

[2] Fabrizio Dell'Acqua et al., "Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality", Harvard Business School working paper with Boston Consulting Group, 2023.

[3] Erik Brynjolfsson, Danielle Li and Lindsey Raymond, "Generative AI at Work", Quarterly Journal of Economics, 2025.

[4] McKinsey, "The State of AI: Global Survey 2025", 2025.