The Decade Ahead: What You Need to Know Next

I study how organisations make sense of change. Over the last few years I have sat with global CEOs, Nobel Prize winners, heads of state, head chefs, Supreme Court judges and jugglers and asked the same blunt question: how is the world changing and what do people need to know next? The answer I keep coming back to is simple and unsettling. We are heading into a decade of disorientation. It will be the most disorienting time of many careers, and also one of the most consequential.

That contradiction—high disruption and high opportunity—is the heart of what leaders must reckon with today. Artificial intelligence is not a single event. It is a landscape of collisions: efficiency smashing into expertise, speed colliding with care, automation rubbing against craft. My work synthesising conversations with more than 150 Australian and global leaders, and the findings in the Skills Horizon report, point to one practical truth: this year you must stop only understanding AI and start making it work.

⚖️ The central clash: Efficiency versus expertise
🗣️ What leaders are actually saying
🔎 Three practical frameworks to make AI work
🧭 Reorienting people and processes
🛑 Common pitfalls leaders fall into
📈 Measuring what matters
🔁 A short checklist to start making AI work today
🌱 The leadership mindset: cheerful pessimist meets practical optimist
🔧 Policy, ethics and the slow work of fixing things
🧠 The human side: curiosity and craft
🚦 Final diagnosis: make AI work, but make people central
❓ Frequently asked questions
📚 A final note on resources
Suggested links to add

⚖️ The central clash: Efficiency versus expertise

Everybody loves a productivity stat. Many people report wins from generative AI and related tools. Surveys and research show at least two-thirds of knowledge workers see improvements in productivity. Executives expect entry-level tasks to be taken over by AI. Teams finish certain tasks about 25 percent faster when they use the right tools. These are real gains we should not dismiss.

But efficiency is not a free lunch. Work that gets done faster can also be done with less care, and less brain engagement. I have a name for the behaviour I see in some organisations: Vibers. These are well-meaning colleagues who offload tasks to an AI without pausing to check why, how or whether the AI’s output is fit for purpose. The result is what I call work slop—long, poorly crafted emails, shallow content and decisions made without the mental effort that builds expertise.

The research data is blunt. About 40 percent of organisations I’ve worked with have seen this work slop appear. And when people outsource cognitive work without reflection, measurable brain activation drops. Some studies report a persistent reduction of around 55 percent in cognitive engagement after using AI in certain ways. If that becomes standard practice, organisations risk hollowing out the pipeline of future experts.

So your first leadership problem is a balancing act: how to adopt AI to gain speed while preserving and cultivating deep expertise. Ignore that balance and you will speed your way into a shortage of true capability later.

What the imbalance looks like in practice

Senior talent amplified, junior talent flattened. Senior professionals use AI to extend their reach and do better work. Juniors who rely on tools for rudimentary tasks fail to develop the mental models they need to progress.
Task completion without learning. Teams finish more work, but they do not internalise the tacit knowledge or the decision-making heuristics that build craft.
False economies. Organisations hire fewer graduates or skip training because tasks are being automated. That reduces the long-term replenishment of skill.

🗣️ What leaders are actually saying

My conversations with leaders help make the trade-offs granular and actionable. Here are a few takeaways I keep returning to.

"Senior talent is getting amplified and getting better at their work. But junior talent seems to be getting worse."

—Raphael Sank, creative agency leader

"We must learn where that line is: what can be automated and what should still be imagined by humans."

—Rose Herzog, WPP Australia and New Zealand

"This is an existential challenge for creatives. We have to rethink who we are with AI by our side."

—Robert Thomson, media executive

"Think not only about what to do with AI—to make something faster or cheaper—but how to reorient your organisation to make use of it."

—Michael Springer, Sony AI

"We tried once and were wrong. You cannot see every corner with AI. You must constantly experiment, admit mistakes and pivot."

—Nicola L'Amoreau, Chief People Officer, IBM

"This is an opportunity to slow down and fix things along the way."

—Meredith Whittaker, Signal

Those quotes show the repeated pattern: leaders want the gains but also fear the losses. They want frameworks that let them experiment responsibly, preserve craft, and adapt when they are wrong.

🔎 Three practical frameworks to make AI work

When I advise boards and leadership teams, I use three simple frameworks. They are not technological blueprints. They are organisational practices you can implement right now.

1. The Triage Framework: Automate, Augment, or Preserve

Every process, role and task in your organisation should be evaluated against three bands.

Automate: Low-risk, repetitive tasks where speed and consistency beat human judgement. Example: routine data extraction, standardised reporting.
Augment: Tasks where AI speeds work but humans make the final call. This is fertile ground for improving productivity while preserving learning opportunities. Example: draft content, research briefs, decision support systems.
Preserve: Work that builds expertise, requires deep care, ethics, craft or complex judgement. Keep humans central and use AI only as a lightly supportive tool. Example: editorial judgment, creative concepting, strategic decisions with high uncertainty.

Map every role and major workflow through this triage. The map becomes a governance blueprint. It also tells you where to direct learning resources and where to monitor for work slop.

2. The Prompt and Practice Playbook

Prompting is a new basic skill but it is not a shortcut to expertise. Treat prompting as a disciplined practice akin to interviewing. Teach people to:

Define the goal before prompting. What decision will this output inform?
Provide context. The better the context, the better the output and the more learning happens for the human using it.
Check for errors and bias. Make critique part of the normal workflow.
Annotate and save prompts and outcomes. Build an internal library of prompts tied to domain outcomes.

Good prompting increases efficiency and also creates a traceable learning record. That is how juniors can learn both from the tool and from feedback loops set up by seniors.

3. Experiment and Pivot: A Governance Loop

One of the most important organisational habits is to assume you will be wrong. Nicola’s experience at IBM is instructive. The executive team thought it had a one-year plan for AI adoption and then had to pivot.

Build a simple governance loop:

Run small, time-boxed experiments
Define success metrics before you start
Collect qualitative and quantitative feedback
Decide to scale, adapt or stop within a set review period

When leaders commit to short, transparent experiments, they reduce fear and increase learning velocity. It also creates a culture where admitting mistakes is normal and expected.

🧭 Reorienting people and processes

Technology changes faster than culture. The most successful organisations focus on people and processes first, then technology. Here are practical moves I recommend.

Make learning visible

Learning that is invisible gets ignored. Create mechanisms that make learning count as work. Examples:

Give credit and time for mentoring that explains how prompts were constructed and why outputs were edited.
Create a "prompt library" with outcomes and post-mortems.
Require a short annotation field on AI-generated work that outlines human verification steps taken.

Protect apprenticeships

Entry-level roles are not just labour; they are training grounds. If entry-level work disappears, you must design new apprenticeships that replicate learning opportunities. Options include:

Rotate juniors through augmentation-heavy roles with mentors
Create graded responsibility so mistakes are safe and instructive
Use AI as a teaching tool rather than a replacement tool

Build human-in-the-loop checkpoints

Workflows should incorporate checkpoints where humans must engage before the output becomes final. These are not onerous. They are short, high-quality interventions that maintain standards and transfer tacit knowledge.

🛑 Common pitfalls leaders fall into

Leaders get three things wrong more often than not. Naming them helps you avoid them.

Buying instead of building the habit. Organisations buy tools and expect behaviour change without investing in learning and governance.
Over-centralising decisions. Some leaders lock AI decisions up at the top. That slows learning at scale.
Ignoring ethics until it’s too late. Ethical lapses damage trust and are expensive in the long run. Treat ethics as a feature, not a blocker.

How to spot work slop early

Look for signals that efficiency is eating expertise:

Lots of AI-generated drafts that never get substantive human edits
Rising error rates in customer-facing outputs
Fewer internal promotions for technical roles
Decreasing attendance at learning sessions

If you see these, pause automation plans and ask hard questions about training, mentorship and quality checks.

📈 Measuring what matters

Productivity metrics are seductive but incomplete. If you only measure time saved, you risk rewarding behaviours that hollow out expertise.

I recommend a mix of metrics across three domains:

Performance: efficiency gains, error rates, business outcomes
Capability: internal promotion rates, task complexity handled by mid-level staff, observable skill growth
Trust and quality: customer satisfaction, ethical incidents, human verification rates

Pair quantitative data with qualitative narratives. Stories from mentors about how a junior learned a decision-making pattern are as important as minutes saved in a dashboard.

🔁 A short checklist to start making AI work today

Use this as your immediate, pragmatic to-do list.

Run a rapid inventory of tasks using the Automate/Augment/Preserve triage
Mandate an annotation on all AI-assisted deliverables describing human checks
Start a prompt library and require teams to contribute prompts and outcomes
Create short, time-limited experiments with clear metrics and a review cadence
Protect apprenticeship pathways and design new learning rotations
Establish human-in-the-loop signoffs for critical outputs
Track capability metrics as well as productivity metrics

🌱 The leadership mindset: cheerful pessimist meets practical optimist

I call myself a cheerful pessimist. That sounds odd but it’s relevant. The future is bright-ish. There are tremendous opportunities if we are deliberate. Being cheerful does not mean ignoring the risks. It means we plan for them, name them, and build systems that allow us to pivot when the path is wrong.

When leaders accept that they will be wrong sometimes, they free their organisations to move faster and learn more. Nicola L'Amoreau’s candid admission that her team’s early plans were wrong should be the standard operating tone across leadership teams. Experiment, learn, adapt and repeat.

🔧 Policy, ethics and the slow work of fixing things

Technology has a speed advantage. Policy and ethics require a different cadence. Meredith Whittaker’s counsel to slow down and fix things is not anti-technology. It is an argument for deliberate stewardship. That stewardship looks like:

Transparent documentation of model use and data sources
Clear escalation paths for ethical concerns
Industry collaboration on standards where regulation is slow
Investment in explainability and audit trails for high-risk decisions

These interventions protect reputation and, crucially, keep your teams honest about the limits of their tools.

🧠 The human side: curiosity and craft

AI will exaggerate differences between those who cultivate the craft of their work and those who do not. Curiosity becomes an organisational superpower. Curiosity manifests as asking better questions, seeking the reasoning behind outputs and treating AI as a collaborator that tests your hypotheses rather than an autopilot that absolves you of thinking.

Practical moves to keep curiosity alive:

Host regular "why" sessions where teams explain reasoning behind key decisions
Celebrate failures that produce learning and document them
Reward people who mentor and explain AI-assisted decisions

🚦 Final diagnosis: make AI work, but make people central

The key message I have for leaders is crisp: you cannot outsource judgement and expect expertise to regenerate itself. If you want speed, you must design for learning. If you want automation, you must preserve apprenticeship. If you want ethical outcomes, you must build governance that is lightweight enough to move but robust enough to matter.

Do these things and you will turn the decade of disorientation into a decade of productive reinvention. Ignore them and you will find yourself efficient but shallow, fast but fragile. The choice is yours and now is the time to act.

❓ Frequently asked questions

How do I decide which tasks to automate with AI?
Use the Automate/Augment/Preserve triage. Automate repetitive, low-risk tasks. Augment tasks where AI speeds work but humans retain judgment. Preserve tasks essential for developing expertise and where ethical or creative judgment matters most.
Won’t using AI reduce cognitive skills and lead to a lazy workforce?
It can if you let it. Organisations that see reduced engagement do not pair AI with learning and human-in-the-loop practices. Require annotations on AI-assisted work, protect apprenticeships, and design checkpoints where humans must interpret and justify outputs.
How can we measure whether AI is improving capability and not just saving time?
Track a balanced scorecard: performance (time saved, error rates), capability (promotion rates, task complexity handled by mid-level staff) and trust/quality (customer satisfaction, incident reports). Combine metrics with qualitative narratives from mentors and reviewers.
What governance model works best for AI adoption?
Start small with rapid experiments. Define success metrics up front, review outcomes in short cycles and be prepared to pivot. Pair experiments with a lightweight governance board that focuses on ethics, safety and learning outcomes.
How do we prevent 'work slop' in our teams?
Detect it early by monitoring for lots of unedited AI drafts, rising error rates, and falling participation in learning. Intervene with mentorship, set standards for human edits, and require brief annotations explaining verification steps.
Should we centralise AI strategy or distribute it across teams?
Avoid extremes. Centralise standards, ethics and shared infrastructure. Distribute experimentation and domain-specific adoption so teams can learn quickly and adapt tools to their context.
Will AI take entry-level jobs?
AI will take many routine entry-level tasks, but that does not mean entry-level roles disappear. You must redesign entry roles to focus on learning, mentoring and ascending through graded responsibilities rather than merely performing repetitive work.
How do we build a culture that admits mistakes and pivots?
Institutionalise short, public experiments with clear review points. Reward transparency about what went wrong, not just success. Leaders must model admitting error and show how they pivoted based on evidence.

📚 A final note on resources

If you want a synthesis of research, interviews, and practical playbooks, look to the Skills Horizon work. The report bundles lessons from business, academia and policy into tangible actions for leaders and teams. Everything I have written here is grounded in that same lineage: collect evidence, design experiments, protect learning and build governance that is humane and iterative.

This decade will be disorienting, yes, but it will also be a period where good leadership matters more than ever. Be deliberate. Keep learning. Balance the tempting speed of AI with the stubborn slow work of building expertise. That is how you make AI work for your people, your teams and your organisation.