A founder runs a candidate team through the usual evaluation: problem-solving strength, resilience under pressure, drive, cross-team behavior, depth of experience. The same five areas every diligence process probes. The scorecard weights have not changed in five years.
The market underneath the scorecard has. Three of the five areas now have machine equivalents available by subscription. The weights did not move. The weights are the error.
Which of the five areas can AI actually replace?
Parts of three, and the parts it replaces are the parts the old scorecard priced most confidently.
Intellect rents. Problem decomposition, research synthesis, first-pass design, code. Not perfectly, and not without judgment wrapped around it, but a capable model performs a useful fraction of what raw problem-solving muscle used to gate. When a capacity becomes rentable, its scarcity price collapses. The team that pays a hiring premium for generic problem-solving strength is buying yesterday's scarce asset at yesterday's price.
Experience externalizes. The useful definition of experience is the conversion of wins into repeatable systems, and systems now include the agent kind: specs, harnesses, escalation rules, review gates. A team's experience compounds into infrastructure that executes without any individual present. The asset stops living in tenure and starts living in the repo and the specs, which is why What the Spec Doesn't Decide, the Model Will reads as an experience argument in disguise: a spec is extracted experience, deposited where the organization can rent it forever.
Behavior extends. Cross-team work becomes cross-system work. Handoffs to agents need contracts, escalation needs design, and the discipline that used to be called collaboration now includes agent handoffs with ready, done, authority, failure, and evidence. The behavior station did not disappear. It widened.
What remains unrented: personality in the operational sense, resilience and adaptability under real load, and motivation, the drive that pushes an idea past the indifferent middle months. No model supplies a team that survives its first failed launch. No subscription ships the founder who keeps going when the deck stops working. These two areas were always the hardest to measure, always the flabbiest columns on the scorecard, and they are now the ones doing the most load-bearing.
This is the redistribution: rentable capacity flows to machines, and evaluation weight flows to what cannot be rented. Human Judgment Is the Scarce Layer names the general principle. The five-area frame gives it an operational shape: the scorecard's center of gravity moves to the two columns nobody ever weighted well.
The mechanism: scarcity moves, weights do not
Prices track scarcity, and evaluation is a price. When intellect was scarce, problem-solving puzzles were the interview, and rightly so. When experience lived in skulls, years-of-experience was a defensible proxy, and rightly so. Both heuristics priced a real scarcity that has since shifted.
Weights lag scarcity because weights are installed by the era that created them, and nobody schedules a re-weighting. The founder in the opening scene is not careless. He is running last decade's honest instrument on this decade's market. Every major tooling shift produces this lag, and the teams that notice first buy undervalued capacity: the resilient operator, the driven builder, hired at the price the market still assigns because the market's scorecard has not moved either. Small Teams, Real Owners, Fast Executives describes the end state: agents prepare, owners decide. The deciding layer is hired on exactly the two areas the old scorecard underweighted.
There is a flip risk, and it is the honest counter-argument. Dismissing the rented areas entirely is the same error in reverse. Judgment about when to rent, what to rent for, and how to wrap control around a probabilistic worker is itself high-order behavior plus experience: AI Needs a Harness Before It Needs Autonomy is the design document for that judgment. Renting well is a skill. The claim is not that intellect and experience stop mattering. It is that they matter as judgment about rented capacity, and that the market still prices them as the capacity itself.
What this is not
Not an AI-changes-everything piece. The redistribution thesis is narrow: three areas have rentable equivalents, two do not, and evaluation weight follows the unrentable. Everything else about team-building holds, including the chain: initiatives still pass through all five stations, and a team still ships at its weakest area. The stations have changed owners, not the chain.
Not a call to hire only for grit and charisma. Resilience and drive without coverage produce inspiring teams that ship badly. The re-weighting is about marginal dollars and marginal attention, not about a new single criterion.
And not a prediction that the weights stop moving. The unrentable two are scarce today. If adaptability itself becomes rentable, the center of gravity moves again, and the discipline of re-weighting becomes the durable skill. The scorecard is a living document or it is a museum.
The move
Re-weight deliberately, once, this quarter:
- List the five areas. Mark which ones your organization currently buys at a premium that a subscription also sells. That premium is the leak.
- Move the freed weight to the two unrentable areas, and make them concrete in evaluation: resilience observed under a real setback, drive observed on work nobody required.
- Audit the experience station for externalization: how much of the team's extracted experience lives in systems and specs rather than tenure? That ratio is the modern experience metric.
- Schedule the next re-weighting. Scarcity moves faster than the instrument that prices it.
Take-home
- Reread your hiring rubric and mark which scored criteria a capable model now performs for the price of a subscription. The premium on those rows is the budget leak.
- Interview harder for the two areas that never had good instruments: what the candidate did after a real failure, and what they built that nobody asked them to build.
- Treat specs, harnesses, and escalation rules as your experience balance sheet. Tenure is what experience used to be called.
Related reading:
- Teams Ship at Their Weakest Area, Not Their Average
- Human Judgment Is the Scarce Layer
- AI Needs a Harness Before It Needs Autonomy
- What the Spec Doesn't Decide, the Model Will
- Agent Conversations Without Contracts Are Faster Confusion
- Small Teams, Real Owners, Fast Executives
If your evaluation rubric still pays last decade's premiums for capacities that now rent by the hour, the scorecard is the competitive disadvantage. If you want help re-weighting the five areas for the agent era, book a discovery call.
FAQ
Which of the five team areas can AI replace?
Parts of three. Intellect rents: models now perform a useful fraction of problem decomposition, synthesis, and first-pass execution, collapsing the scarcity price of generic problem-solving. Experience externalizes into systems, specs, and harnesses that run without their author. Behavior extends to human-agent handoffs governed by contracts and escalation rules. Resilience and drive remain unrented.
Why does evaluation weight shift to resilience and drive?
Because prices track scarcity. When a capacity becomes rentable by subscription, its scarcity value collapses, and the marginal hiring dollar buys more where scarcity remains. Resilience under real load and drive through indifferent months were always the hardest areas to measure, which is why the old scorecard underweighted exactly the areas now doing the most load-bearing.
Does this mean intellect and experience stop mattering?
No. They matter as judgment about rented capacity rather than as the capacity itself. Knowing when to rent, what to delegate to an agent, and how to wrap control around probabilistic work is high-order behavior plus experience. Teams that dismiss the rented areas entirely make the same pricing error in reverse.
How should hiring change in the agent era?
Audit the rubric for criteria a subscription also sells, and move that weight to the unrentable two, tested concretely: what the candidate did after a real failure, and what they built that nobody required. Treat the ratio of extracted experience living in systems and specs, rather than tenure, as the modern experience metric.
When should a team re-weight its evaluation rubric?
Whenever scarcity moves, which is faster than most teams schedule for. A deliberate re-weighting once a year is a reasonable floor, with an immediate trigger whenever a new rental class appears for something the rubric prices at premium. The scorecard is either a living document or a museum.