Why the core cannot become the twin
A large firm is a machine for producing the norm. Efficiency, repeatability, the quarterly close, the works council, the seven-to-nine-layer decision path. That is not a failure of character. It is the job the core was hired to do.
Adidas CEO Bjørn Gulden put the same idea in plain language when he took over a troubled company: get the naysayers out of the way. Too many layers, and one person can stop the move that actually matters. He did not form a working group to socialize the risk on the Samba. He took the people who had already said no out of the path, pulled production forward, and later pushed decision rights closer to the market. Sales and profit recovered. The board extended him.
The same mechanism applies to an automotive company. The core’s immune system will treat a twin that rewrites workflows as a threat. Process owners, legal, finance, and the works council are doing the jobs they were trained to do. Their no is rational on their local metric.
Collectively, it kills the only move that could change the trajectory. Inviting more of the core into the twin’s design doesn’t make it safer. It reconstitutes the veto machine as a steering committee and a Slack channel. The parallel run never accumulates enough instances to beat the incumbent on evidence.
The decision must come from the C-level
Only a C-level decision can cut through that clay. The twin needs a mandate that answers the obvious questions before they are asked, then locks the naysayers out of the existence question while inviting them into specific, bounded jobs. Finance instruments the parallel measurement. Legal and security design the evaluation band and the week-one rollback. The process owner can refuse a draft on the named workflows. They cannot refuse the twin itself.
One named executive must own that mandate and have hiring and firing authority, not a vote. The twin reports to the chief executive or one deputy, not to a monthly digital board. Money is ring-fenced past the next forecast. The twin reaches the live files in week two, not after the official platform program finishes. Without that short line to authority, the first act becomes a partners’ meeting, and the deviation never runs long enough to produce a number.
No transformation office, no multi-year roadmap, and no “sharing learnings with the core” can substitute for that cut. The core will not argue itself into becoming the edge.
It must be a separate project — or a separate organization
The practical form is an edge twin: a new unit, still inside the firewall, designed from the start on different principles. Work is routed through intelligence layers. Humans stay deliberately above the loop — setting constraints, validating outcomes, handling exceptions.
We rebuild two high-throughput workflows and run them in parallel with the core on the same live inputs. Measure cycle time, quality, cost, and risk against the incumbent. What loses is deprecated. The next workflow is taken. Over months, not years, the edge stops being the edge.
This cannot live as a project inside the existing operating model. It needs its own decision rights, evaluation criteria, and system of record for the workflows it owns. In many cases, the cleanest structure is a separate legal entity or a clearly ring-fenced subsidiary whose success metric is not the core’s quarterly close. The mothership’s processes stay where they are. The twin doesn’t depend on them for permission to exist or scale.
Backcasting comes first: define the future operating model, then work backward to the two chains that must run on day one. Name the input, output, gate, and system of record. Keep the room small enough that a Tuesday decision still happens on Tuesday. The CEO or the named deputy, two gate owners, finance, legal and security, and a build cell few enough to ship.
What the twin delivers
Development cycles shorten. New products can be produced at costs the core cannot match. Multiple innovation streams restart instead of one more cost-cutting program. The company regains a trajectory that is viable beyond the next round of savings.
The alternative is already visible: strategies discarded, products that no longer sell, and structures that keep saying no until the market has moved. The core will keep producing the norm. The twin is the deviation that is allowed to run long enough to produce evidence. Without that separation, and without the C-level decision that protects it, progress isn’t scheduled, and the company won’t evolute into an AI-native organization.