The Firm, Rewired: Part one
Strategy
What would we gain by redesigning work for AI?
Intelligence has just become abundant. The cost of analysis, drafts, and trying another approach has collapsed, so the scarce, valuable things have changed.
The Intelligence Supply Shock
Every process in your business was designed around one assumption: thinking time is expensive. Expert attention was limited and analysis was slow, so work got rationed through queues, sign-offs and specialists.
Agents break that assumption. Another analysis, another draft, another option explored now costs pennies and arrives in minutes. That makes AI more than a better tool. It is new productive capacity, reaching every desk the way electricity once reached every machine on the factory floor.
When Google's AI Co-Scientist compressed months of expert research into two days, the scale of the shift became concrete.
Scarcity Has Moved
When everyone can produce unlimited drafts, analysis and code, output stops being the thing you compete on. What stays scarce is everything around it: knowing what to ask for, judging what comes back, coordinating who does what, and owning the result.
By context we don't mean clever prompts. We mean what your business knows that the model doesn't: your priorities, your constraints and exceptions, your appetite for risk, how your pricing really works, what your regulator expects, and the unwritten knowledge of how things actually get done.
The evidence backs this up. The UK government's AI Safety Institute found AI roughly doubles performance on attention-heavy tasks, but delivers no measurable gain on tasks that need judgement, planning or prioritisation.
What becomes abundant
...so what becomes scarce
Generating text, code and analysis
Review capacity to vet what's produced
Exploring alternatives and branches
Judgement on which path to actually take
First drafts and routine synthesis
Reliable evaluation of whether it's any good
Executing individual cognitive sub-tasks
Coordination across people and modules
Delegating operational rights to agents
Named human accountability for outcomes
Producing polished output at scale
Trusted escalation paths when something fails
The New Economics
Here is the uncomfortable pattern behind the headlines: individuals using AI get faster, yet their companies often don't. Economists have a name for the cause: "so-so automation", AI that substitutes for human effort without changing what the business can actually do. A sole trader pockets the gain directly. In a larger firm, the same gain is swallowed by the approvals, reviews and rework that sit unchanged around it.
The firms that thrive won't simply be the ones with the most AI output. They will be the ones that redesign around what is now abundant, and manage by sampling and statistics rather than reading every word.
That gap is not a reason to wait. It is the reason the next two parts, management and competitive advantage, matter as much as the technology itself.
Intelligence is now abundant, context is the advantage. Let's engineer yours.