AI rewiring organizations is the real shift of this moment — and most leaders are still missing it. In this special panel from the Kyndryl Institute and The Innovation Show, four of the sharpest minds in strategy and innovation explain why bolting AI onto old structures fails, and what to build instead.
Host Aidan McCullen is joined by Rita McGrath, Alexander Osterwalder, Ismail Amla and Usman Sheikh. Together they map how work, value and power are being redrawn — and why the winners may not be who you expect.
Why AI rewiring organizations starts with a rusty gate
Aidan opens with a lesson from his grandmother. As a boy, he painted over a rusty gate to earn his pocket money. The rust bled straight through, so he had to strip it back and start again.
That is exactly how most firms use AI today. They paint it over broken processes and outdated structures. However, AI rewiring organizations means stripping back to the metal first. Ismail Amla calls it the difference between bolting technology on and rewiring everything.
The death of the billable hour
Rita McGrath’s Wall Street Journal piece named the shift bluntly: AI and the death of the billable hour. For decades, law and consulting split an hour into six intervals and billed each slice. With AI, that metric collapses.
As a result, value moves to outcomes. One legal-AI firm now bills for the documents it delivers, not the hours it burns. Because the cost of production keeps compressing, the old pyramid of junior staff has far less to sell.
Power moves to the edges
McGrath’s research for her new book surprised her. The best leaders are not stepping back — they are deep in the weeds, yet without command and control. At NVIDIA, Jensen Huang reads staff “Top Five Things” memos on Sunday nights and comments directly, connecting himself to the front line.
Bayer, meanwhile, runs on 90-day cycles and separates a “professional home” from day-to-day tasks. Therefore politics pays less and good work pays more. This is the permissionless organisation in practice.
Agents, verification and the new bottleneck
Usman Sheikh warns of a quieter problem. AI speeds the work up, but the ability to check that work lags behind. Consequently, judgement becomes the bottleneck, and faster output that needs twice the review is no saving at all.
Ismail Amla adds that enterprises now manage agents like employees — onboarding them, reinforcing behaviour and exiting them. In regulated industries, however, the hard question is the audit trail: can you prove thousands of autonomous agents act in line with your mission?
AI-native versus legacy
Just twelve months ago, having hundreds of thousands of people meant labour arbitrage and advantage. Now, Amla notes, it can be a liability. Clients who once expected ten people now expect three people and a tool — and they expect to share in the value created.
South Korea’s Toss shows the other path. Founded by a dentist, it runs on directly responsible individuals who launch and retire work as the market responds. Around half of what the team ships does not stick, and that is by design.
What incumbents still own
AI-native startups carry no legacy and no psychological debt. Yet incumbents hold real assets. Everyone has the same models and compute, so advantage now comes from proprietary data, deep domain expertise and context.
Osterwalder’s warning is sharp: adopting AI without changing your business model is innovation suicide, the same mistake Kodak made with the digital camera. His advice is to fund many small teams like a venture capitalist and let the winner emerge, because no leader can pick it in advance.
Related episode on Rita McGrath and transient advantage
Usman Sheikh’s Frame Break essay on the verification bottleneck
Listen and watch
Hear the full conversation on Apple Podcasts, Spotify and YouTube.
About the host
Aidan McCullen is the 2025 Thinkers50 Innovation Award recipient, a keynote speaker on AI, disruption, innovation and change, host of The Innovation Show, and author of Undisruptable: A Mindset of Permanent Reinvention (Wiley).
More about Aidan · admin@theinnovationshow.io
Frequently asked questions
What is the episode “AI Is Rewiring Organizations” about?
It is a panel from The Innovation Show, produced with the Kyndryl Institute, on how AI is rewiring organisations — moving work away from billable hours toward permissionless, outcome-driven models. Host Aidan McCullen is joined by Rita McGrath, Alexander Osterwalder, Ismail Amla and Usman Sheikh.
Who are the panellists?
Rita McGrath, a Columbia Business School professor known for “transient advantage”; Alexander Osterwalder, creator of the Business Model Canvas and CEO of Strategyzer; Ismail Amla, SVP at Kyndryl and author of From Incremental to Exponential; and Usman Sheikh, author of Frame Break and MD of High Output Ventures.
What does “permissionless organisation” mean?
Coined by Rita McGrath, a permissionless organisation lets people act and make decisions without waiting for sign-off through layers of hierarchy. AI compresses the cost of doing work, so power shifts to the edges — as at NVIDIA, where Jensen Huang reads staff memos directly, and Bayer, which runs the company on 90-day cycles.
Why is the billable hour dying?
Rita McGrath argues that AI breaks the link between time spent and value delivered. When a tool can produce in minutes what once took a team hours, billing by the hour stops making sense. One legal-AI firm now charges for the documents it delivers rather than the time it burns, signalling a shift to outcome-based pricing.
What is the “verification bottleneck”?
Described by Usman Sheikh, the verification bottleneck is the gap between how fast AI produces work and how fast humans can check it. As output accelerates, human judgement — not generation — becomes the constraint, so faster work that needs heavy review delivers little real saving.
How can incumbent companies compete with AI-native startups?
The panel argues that since everyone has access to the same models and compute, advantage now comes from proprietary data, deep domain expertise and context. Alexander Osterwalder warns that adopting AI without changing the business model is “innovation suicide,” the mistake Kodak made with the digital camera.
Can one person really build a company with AI?
Yes. Rita McGrath cites a single developer who built a product alone and sold it to Wix for around $80 million within six months — an example of how AI enables very small teams to create outsized value.
How should companies manage AI agents?
Ismail Amla suggests treating agents much like employees — onboarding them, reinforcing the behaviour you want and retiring them when needed. In regulated industries, the harder challenge is the audit trail: proving that autonomous agents act in line with the organisation’s mission.