Steam Engines in Wooden Carts
Why bolting AI onto legacy structures fails the J-curve.
Around 1900, factory managers replaced their steam engines with single large electric motors but kept the old shafts and belts. Productivity stagnated for decades. That is the Dynamo Paradox — a new-world power source forced into old-world structures.
Nordic management is in exactly the same trap with AI today.
The productivity paradox and the law of the J-curve
Economist Paul David showed the full impact of electrification arrived in the 1920s — about 40 years after the first power plant. The new technology required additional investment and organisational changes that could not be made overnight.
Erik Brynjolfsson calls this the J-curve of productivity. Initial AI adoption can briefly *reduce* productivity as resources flow into integration, training and workflow redesign. Manufacturing has seen drops of up to 1.33 percentage points in the early phase.
This is the moment many Nordic leaders get scared and retreat to the safe Copilot mindset. The Bold Ones see it through and ride the steep climb where market share and efficiency grow exponentially.
| Benchmark | Steam era / traditional IT | Electrification / autonomous AI |
|---|---|---|
| Power source | Centralised steam engine / IT | Small electric motors / Ecosystem of Agents |
| Architecture | Rigid shafts and belts | Flexible workflow |
| Control logic | Manual control | Proactive orchestration |
| Productivity | Incremental (10%) | Exponential (10x) |
| Key risk | The Copilot Trap | Structural rigidity |
Tiny motors in every machine
The 1920s productivity leap arrived when someone realised the true power of electricity was decentralisation. The belts came down. Each machine got its own small motor. Factory floors were rearranged by workflow, not by power source.
That is the key insight of the Self-Driving Enterprise. AI is not a big machine in the basement — it is the Ecosystem of Agents, a distributed network of specialised intelligent engines placed directly at the heart of the work.
The architectural nervous system runs in layers:
- Agent Runtime — a secure environment for agents.
- Identity & management — every agent has a digital identity and access only to authorised data.
- Communication Fabric — standard protocols (Agent2Agent, MCP) so agents share memory and coordinate tasks.
- Model abstraction — flexible use of different AI models per task.
- Observability & Safety — continuous monitoring of accuracy and ethical compliance.
The Great Rethink: cutting the straps
Many current processes are designed around human limits: limited memory, slow coordination. The Great Rethink is the courage to dismantle them and rebuild on the agents' terms.
If your decision chain is slow and hierarchical, an AI assistant in the middle is just a slightly faster way to wait. True reengineering replaces email coordination with direct agent integration. The sales agent already knows inventory; the logistics agent already knows the weather. They synchronise via the Mesh — no email required.
This requires the vulnerability Brené Brown calls for. 'We've always done it this way' is the most dangerous phrase in the SDE world.
Strategic Questions of Fate
Six questions that tear apart old market structures:
- Is email a leader's biggest obstacle? Inbox Dread starts the day with other people's priorities. Agents can handle 90% of routine communication, freeing cognitive capacity for Covey's Quadrant II work.
- Radical control and ecosystem — can a leader trust autonomous systems enough to focus on morning coffee and direction?
- Customer service without waiting — agents act like a five-star concierge, addressing issues before the customer notices them.
- Psychological safety — when survival energy is freed, courage becomes available.
- Mastery vs. completion — the operator is dead, the orchestrator is born.
- The death of product-centricity — Christensen's Jobs to be Done: the customer doesn't buy an elevator, they buy smooth vertical movement.
The Strategic War Room: Wind Tunnel and Prophet
The traditional strategy process is a once-a-year exercise based on history and best guesses. In SDE, strategy is dynamic and lives in a Wind Tunnel.
The process starts by identifying Questions of Destiny. Agents called Prophet and Navigator run 10,000 simulations of different scenarios. Management no longer 'decides' the strategy — they choose the most resilient path the simulations surface. The result is predictive immunity against market arrhythmias.
New market gravity
AI does not create value through automation alone — it changes the unit economics of the business model. When the cost of routine tasks drops to near zero, you can scale operations that were previously impossible. Marketing campaigns 100× cheaper and 100× more personalised will change the gravity of the market.
| Metric | Traditional model | SDE model |
|---|---|---|
| Scaling | Labour-intensive, linear | Technology-intensive, exponential |
| Cost structure | High variable costs | High fixed, near-zero variable |
| Decision | Days or weeks | Milliseconds |
| Customer value | Standardised mass | Ultra-personal concierge |
Gluing technology onto old structures is the most expensive mistake in history. A productivity leap requires complete reorganisation of the workflow on the agents' terms — not on the terms of human limitations.
Identify one cross-functional process where you would dare to redesign the org chart, not just the tooling.