Rewiring for AI 2.0: The End of the ‘Two-Pizza Team’
For the past two years, the standard corporate approach to AI has been to identify individual use cases and drop AI tools into existing processes. That era is officially over. This week, McKinsey released the updated second edition of its landmark ‘Rewired’ playbook, and the data is unequivocal: companies that fully apply the framework are achieving an average EBITDA uplift of 20% and a 3:1 return on every dollar invested.
The critical shift in this new edition is the move from individual use cases to end-to-end workflow redesign. McKinsey is now explicit that the highest-value AI practice is reimagining entire workflows from the ground up.
The Upskilling Trap: Why Your AI Training is Already Obsolete
The C-Suite Radar this week looks at what it really means to rewire an organisation around end-to-end AI workflows. And that raises the question I keep coming back to: who is actually going to lead that work?
The same McKinsey research that produced the Rewired playbook also published a separate survey this week, and the findings should give every leader pause. 75% of US workers expect their roles to change due to AI in the next five years. Only 45% have gone through a recent upskilling programme.
That gap is not really about training budgets. It is structural.
Traditional corporate training programmes are fundamentally incompatible with the speed at which AI is developing.
The Speed of Obsolescence
McKinsey makes a brilliant point in their research: imagine an AI training programme designed in January 2026 around one specific model. By the time that programme is rolled out to the workforce, it is already partially obsolete. Why? Because five major frontier AI models were released in February alone, alongside almost weekly feature updates.
When you try to train your workforce on specific tools, you are fighting a losing battle against the product release cycle.
The Lighthouse Approach
The solution is not to build faster training modules but to change how learning happens.
McKinsey advocates for a ‘lighthouse’ approach. Instead of rolling out company-wide, classroom-style training, organisations should select small, high-performing teams to act as visible role models and peer coaches. You equip these teams with the best tools, give them a real business problem, and let them figure out how to solve it using AI.
The results are striking. In one example cited by McKinsey, a lighthouse team working on procurement reimagined its approach to supplier management. They reduced manual sourcing and analysis effort by roughly 90% and identified approximately €80 million in value for the client.
More importantly, that team then becomes the training engine for the rest of the organisation. They share their workflows, their prompts, and their results with peer teams. It is peer-to-peer learning in the flow of work, which is the only kind of training that can keep pace with the technology.
What This Means for You
If you are waiting for your employer to roll out a comprehensive AI training programme, you are going to be waiting a long time. And when it finally arrives, it will likely be teaching you how to use the tools from six months ago!
Your AI fluency is your responsibility. Find the ‘lighthouse’ individuals in your network – the people who are actually building and experimenting – and learn from them.
This piece first appeared in Talent Pools Nexus, our weekly read for leaders in research, insights and consulting. Subscribe on LinkedIn.
