Executive summary: it is not what you think
The prevailing narrative says AI will automate the mundane and free humans for creative, strategic work. A closer look shows the opposite. AI is surprisingly good at creative tasks and surprisingly weak at nuanced judgement.
The middle of the cognitive pyramid — evaluation and judgement — is the new human frontier. That is where your competitive advantage must be built.
1. The great misconception: the cognitive complexity pyramid
The comforting story is that AI handles the boring stuff. The reality is more disruptive. AI is effective at both ends of the pyramid: it summarises and reports (comprehend), and it writes code and creates designs (create). The area where human intelligence remains defensible is in the middle.
| Cognitive task | Expected AI role | Actual AI role |
|---|---|---|
| Create (new knowledge) | Complement | Substitute — AI is surprisingly good |
| Evaluate (judgement) | Complement | Complement — the new human frontier |
| Apply (methodology) | Substitute | Substitute |
| Comprehend (describe) | Substitute | Substitute |
Insight for CEOs. Your competitive advantage may not lie in pure creativity, but in the nuanced, context-aware judgement of your team. The ability to critique, evaluate and decide in complex situations is the human-centric frontier.
2. The data: productivity, wages and job growth
Contrary to fears of a "white-collar bloodbath", the data so far tells a different story. A PwC analysis of nearly one billion job ads shows clear divergence between AI-affected and non-AI industries.
| Metric | AI-affected | Non-AI | Takeaway |
|---|---|---|---|
| Productivity growth | 3.0× | 1.0× | AI sectors are three times more productive |
| Wage growth | 1.5× | 1.0× | AI is creating higher-paying roles, not eliminating them |
| Headcount growth | 0.8× | 1.0× | Slight decline, more than offset by wage and productivity gains |
Insight for CEOs. The real risk is a talent gap. The question is not "will AI take jobs?" It is "will your company be on the right side of the productivity divide?"
3. The jagged frontier: you cannot predict AI's next move
AI performance across tasks that look similar to humans is wildly uneven. A model may excel at a complex coding task and fail at a simple reasoning problem.
| Task | Human-perceived difficulty | AI performance |
|---|---|---|
| Math problems | High | High (9/10) |
| Legal analysis | High | Medium-high (7/10) |
| Creative writing | High | High (9/10) |
| Coding | High | High (9/10) |
| Medical diagnosis | High | Low–medium (4–6/10) |
| Humour | Low | Low (3/10) |
| Ethical judgement | High | Low (4/10) |
| Financial trading | High | Low (3/10) |
| Strategic planning | High | Medium (5/10) |
4. The task-analysis matrix: where to focus
| Substitute | Complement | |
|---|---|---|
| Core | High-risk zone. Your differentiator becomes a commodity. Redesign the value proposition urgently. | Sweet spot. AI amplifies your unique value. Invest here to build your AI-powered moat. |
| Non-core | Automate. Free up time for higher-value work. | Efficiency gains. AI helps with routine support tasks. |
5. The expert–novice gap: AI does not flatten talent uniformly
Conventional wisdom says AI will lift novices and narrow the gap with experts. The reality is domain-dependent.
| Domain | Impact on the gap | What it means |
|---|---|---|
| Material discovery | Widens — experts pull further ahead | Experts use AI to explore far more complex hypotheses |
| African entrepreneurs | Widens | Experienced founders leverage AI for more sophisticated strategy |
| Legal analysis | Narrows — novices catch up | AI gives juniors expert-level research capabilities |
| Writing quality | Narrows | AI dramatically improves average performers |
| Customer chat | Narrows | AI gives new agents the knowledge of experienced ones |
| Coding efficiency | Narrows | Copilots help juniors write production-quality code |
Insight for CEOs. AI will likely make average employees significantly better — and your best people even more exceptional. The value of a truly outstanding expert may increase. The value of a "good enough" generalist may not.
6. Three traps leaders fall into
| Trap | Description | Antidote |
|---|---|---|
| The automation trap | Cutting cost by automating existing tasks, missing the chance to create new value | Ask: "What can we now do that was previously impossible?" |
| The intuition trap | Predicting where AI will succeed from human intuition about difficulty | Test AI empirically on your specific tasks. Ignore general benchmarks. |
| The talent trap | Assuming AI makes everyone equally productive — under-investing in top talent | Identify your domain experts and turn them into AI-powered super-experts. |
7. A 90-day action plan
| Window | Action | Owner | Success metric |
|---|---|---|---|
| Days 1–30 | Task audit across major functions using the matrix | COO + HR | Complete map of tasks by quadrant |
| Days 31–60 | Empirical pilots on 3–5 tasks in the Core Complement quadrant | CTO + business units | Measured productivity uplift per task |
| Days 61–90 | Re-skilling roadmap focused on judgement, evaluation and AI collaboration | CHRO | Plan covering the top 20% of the workforce |
8. Conclusion: the new human frontier
AI is not simply automating the bottom of the cognitive pyramid and freeing humans for the top. It is disrupting both ends at once, while leaving a critical gap in the middle — evaluation, judgement, contextual decision-making.
That gap is not a weakness to be exploited by AI. It is the new human frontier. The organisations that will thrive are those that build their competitive advantage around uniquely human capabilities: evaluating complex trade-offs, exercising contextual judgement, deciding in ambiguous, high-stakes situations.
Do not ask "How do we use AI to do what we already do, but cheaper?" Ask instead: "What can we now do that was previously impossible — and how do we build an organisation that excels at the judgement-intensive work AI cannot do?"
Memo prepared by Álvaro de Nicolás · June 2026. For board and executive use.