Every other stance in this guide is about my own relationship with the mirror. This one is heavier for me, because it is about the relationship I build for everyone else. I run a company, so I know the weight of it: when you lead a team, a department, or a firm, you are not merely using AI — you are deciding how dozens or thousands of people will meet it. And most of us, under pressure, are getting the decision exactly backwards. I have felt the pull toward the wrong answer myself.

The trap the spreadsheet sets for you

The default logic of cost and efficiency whispers one word: replace. Cut the headcount, automate the task, treat the model as a cheaper worker. It looks like leadership and it books like a win — and I feel that whisper every time I look at a budget. But it is the collective version of the pharmakon from the first series: you take a short-term productivity gain and quietly hollow out your people's judgment, which is the exact capacity you will need to oversee the very AI you bought. Push it far enough and you get an organisation of operators instead of experts: brittle, unable to catch its own machine's confident errors, with no one left who can tell where the competence ends. You will have optimised your company into a building full of puppets, one efficient quarter at a time.

Augment, not replace — as a culture, not a slogan

The 2026 leadership research is unusually blunt on this: AI should augment, not replace, human capability — and the leaders who say so and then back it with actions (training and redeployment, not job-cut threats) are the ones whose adoption and innovation actually accelerate. Building the culture that lets AI be used responsibly and creatively at scale, one report argues, “may be the most important investment companies can make in the AI era.” My job is to build an organisation where the machine raises the floor and frees my people up to the judgment and human-premium work — not one that empties them into interchangeable hands.

Five disciplines that make the difference

• Redeploy the freed capacity; don't just bank the cut. AI lifts the novice toward the expert. Use that to move people up to higher-value work, not simply out. The pure-cut is a value trap: it wins the quarter and loses the decade.
• Protect your talent pipeline. Here is the mistake that eats companies alive: if AI does all the entry-level work, your juniors never do the hard reps — and you never grow the seniors you will need in five years. Cutting entry-level roles “weakens leadership pipelines and long-term capability.” I try to deliberately keep humans learning through the productive struggle, even when the machine could do it faster.
• Keep the judgment — and the ability to oversee. You now manage three kinds of contributor: humans, AI agents, and automation. Design, on purpose, how humans and AI share judgment and accountability. Reward the person who catches the machine's error. Do not let the whole org deskill into automation bias.
• Own your knowledge. Your company's crown jewels — and the collective intelligence of your people — should not be poured, unexamined, into a rented mirror you do not control. Where does your organisation's mind live, and on whose terms? I treat sovereignty as a board-level question now, not an IT footnote.
• Make the mirror argue. An AI that launders bad decisions with confident, fluent output is a governance risk, not a productivity gain. Build a culture that makes the machine disagree, and that keeps a human answerable for every consequential call.

Treat people as people

The human element is not soft. Transparency, AI governance, and reducing bias in AI hiring tools are rising leadership priorities precisely because organisations that build trust — transparent AI, meaningful human oversight, honesty about what is changing — earn both employee confidence and customer loyalty. Surveillance-by-default, AI “companions” deployed to paper over understaffing, and quiet replacement dressed as augmentation all corrode that trust, and it does not come back cheaply. Manager development is the top people-priority of 2026 for a reason: your frontline leaders are the bridge, and coaching a human through change is the one thing the machine cannot do for you.

The deepest reason this is your job specifically

The first series borrowed a warning from the philosopher Bernard Stiegler: the pharmakon cannot be managed by individual willpower alone — it needs a collective, institutional frame. As a leader, you are that frame for the people you lead. This is the responsibility I feel most heavily. An individual can decide to keep their own faculties sharp; but only I can decide whether my organisation keeps its capacity to think, judge, and adapt — or lets it atrophy into rented systems until the company can no longer govern itself. That is not an HR nicety. It is the difference between an institution that endures and one that is quietly proletarianised from the inside.

The honest balance

None of this is an argument to refuse AI — an organisation that abstains loses to one that adopts, full stop. And I will name the real tension, because I live it: you face genuine competitive and cost pressure to cut, and “augment, not replace” can sound naive against a P&L. So do not take it as charity; take it as strategy. The replacement path is the value trap; the augmentation path is harder, but it builds the durable, capable, trusted organisation that still has a talent pipeline, still owns its judgment, and can still oversee its own machines when the others cannot.

Your people will relate to the mirror the way you build the organisation to relate to it. Build one that augments — that raises the floor, protects the learning, keeps the judgment, owns its knowledge, and treats people as people. Or build one of puppets, and watch it hollow out from the inside while the dashboard stays green. The leader's version of this whole guide is the one I repeat to myself before every big call: don't let your organisation fall in love with the mirror.