Leadership in the AI Era

Learning Outcomes

This chapter puts the reader in a position to:

Explain how people historically became senior, and why accumulated judgment was built on bottom-rung repetition.

Describe why AI automates exactly the entry-level rungs on which juniors used to build judgment.

Explain why the resulting break in the judgment pipeline is a hidden crisis that surfaces only years later.

Apply the Ladder-Rebuild Model to name the judgment an automated rung built and design a deliberate substitute for it.

Distinguish incidental learning from deliberate practice, and explain what tacit knowledge the old repetition transmitted.

Produce a judgment-map for an entry-level role plus one deliberate substitute for the learning its automated work provided.

Opening Vignette

In November 2025 three Stanford researchers published an analysis of payroll records from the largest payroll processor in the United States, covering millions of workers month by month. The finding that matters here is narrow and hard. Workers aged 22 to 25 in occupations most exposed to AI showed a 16 percent relative decline in employment, and the decline held after controlling for what was happening to their employers generally. In the same occupations, more experienced workers were stable or still growing.¹

The shape of that finding is the whole problem. The damage landed on an age band inside occupations that were otherwise healthy, and it showed up in headcount while wages held steady, which is the signature of work that has stopped existing. The concentration was in roles where the technology substitutes for the person, and the roles where it substitutes most cleanly are, almost by definition, the ones we hand to people who are learning.

Chapter 11 followed that force into a worker’s livelihood and handed this chapter its brief; Chapter 9 raised the team-level version, when it asked how a leader keeps routing learnable work to junior humans even where an agent is faster. This is where both land. Consider a professional-services firm, a composite drawn to make the mechanism visible, in which the document review, the first-draft memo, and the basic model build are now done by machine. Nothing is obviously wrong. The partners are more productive, the clients are billed less for routine work, and the firm will not discover what it has done for about six years, when it looks for people ready to be senior and finds that it stopped making any.

A partner, watching the transformation, asks a quiet question that the productivity numbers do not answer: if no one does the grunt work anymore, where do the next partners come from? The senior judgment the partner relies on, the feel for which clause is dangerous, which number is wrong, which deal is off, was built, in the partner’s own case, by grinding through the thousands of documents that the machine now handles. The rung the associate is being spared is the rung the partner climbed, and removing it may quietly remove the path to the top of the profession.

This is the chapter’s problem, and it is largely invisible because its cost is deferred. Automating the entry-level work looks like pure gain in the moment and shows its price only in a decade, when the organization discovers it has no one ready to be senior. The apprenticeship ladder by which juniors became experts has been cut at the bottom, and unless a leader rebuilds the path deliberately, the pipeline to senior judgment fails silently while every current metric says the change is working.

This chapter argues that AI automates the entry-level work on which juniors built judgment, cutting the apprenticeship ladder, so the leader must deliberately rebuild the path to senior capability, which osmosis no longer supplies.

How People Became Senior

Expertise has always been built from the bottom up, through repetition that looked like drudgery and functioned as training. The junior lawyer reading contracts, the junior analyst building models, the junior accountant reconciling accounts, the junior doctor taking histories: each was doing work that was valuable to the firm and, at the same time, was accumulating the pattern recognition, the error-spotting instinct, and the domain feel that would eventually make them senior. The repetition was the curriculum, even though no one called it that.

Senior judgment accumulated as a byproduct of doing the junior work. No one designed the document review as a training program; it was real work that needed doing, and the exposure to volume and variety over years did the teaching. A junior who reviewed 10,000 contracts developed, without being taught explicitly, a sense of what a dangerous clause looked like, and that sense was the foundation of the senior judgment the profession valued.

The apprenticeship was self-financing because the junior work had to be done anyway. The firm needed the contracts reviewed, so it paid juniors to review them, and the juniors became experts as a side effect the firm did not have to fund separately. Nothing in the firm’s accounts recorded it, which is why its removal is so easy to miss: when the work stops being necessary the training stops with it, and no line item registers that a training program has just been eliminated.

AI Automates Exactly Those Rungs

The work AI automates first, in the professions, is the entry-level work that built judgment. This is a direct consequence of what the technology does well: the document review, the first-pass research, the boilerplate drafting, and the routine analysis are structured, high-volume, and pattern-based, which makes them ideal for automation and which also made them ideal training. The same features that made the work a good curriculum make it a good target for the machine.²

The bottom rungs are removed first, and the education they delivered goes with them. Where a profession once had a wide base of junior work through which every entrant passed on the way up, it increasingly has a machine doing that work and a smaller number of senior people directing it, with the rungs between them missing. The associate is spared the grind, and is thereby spared the education the grind delivered, and the profession is left with a structural gap between entry and expertise where the apprenticeship used to be.

The automation improves this year’s output and degrades the next decade’s expertise, on different timescales. The efficiency case for this is real and should not be dismissed, which is what makes the problem hard. Having a machine review the contracts is better for the client and the firm in the short run, and no leader should pretend otherwise or preserve grunt work purely as a hazing ritual. The two effects appear on different timescales, which is what makes the trade so easy to lose: a leader optimizing for the visible present will systematically undermine the invisible future unless the trade is made explicit.

The Hidden Crisis

The crisis is hidden because the pipeline it breaks is long. An organization that stops developing juniors does not notice for years, because it still has the seniors it developed under the old system, and they mask the gap. The failure surfaces only when those seniors retire and there is no one behind them who built the judgment the old way, at which point the organization discovers a shortage it can no longer quickly fix, because senior judgment takes years to build and cannot be summoned on demand.

Nothing in current metrics signals the deterioration of the future judgment supply. That deferral is what makes the crisis dangerous. Productivity is up, clients are satisfied, and costs are down, and the one thing that is deteriorating, the future supply of senior judgment, does not appear on any dashboard until it is a shortage. A leader watching the numbers sees success, and the success is real, and it is quietly consuming the seed corn of the organization’s future capability without registering the consumption anywhere the leader is looking.

The task is to replace the ended apprenticeship deliberately while the automating continues. Stopping the automation would forfeit real value. The point is to recognize that automating it silently ends an apprenticeship that has to be replaced deliberately, and to build the replacement on purpose. The organization that automates the grunt work and does nothing else will have a judgment crisis in a decade; the one that automates the grunt work and rebuilds the ladder deliberately will not. The difference is whether a leader sees the invisible cost and acts on it, and the framework is the instrument for acting.

Framework: The Ladder-Rebuild Model

The Ladder-Rebuild Model is this chapter’s principal framework, and it rebuilds the apprenticeship as a deliberate design, since the osmosis that once did the work has ended. It works in four movements, applied to each entry-level role whose work is automating. Name the judgment the old work built; design a deliberate substitute that builds the same judgment without the automated repetition; sequence and assess the substitute so that it actually forms the capability; and keep the learner in proximity to the judgment being exercised.³

The first movement names the specific judgment the automated work used to build. Diagnosis: name precisely what judgment the now-automated work actually built. The document review built pattern recognition for dangerous clauses; the model-building built a feel for when a number is wrong; the reconciliation built an instinct for where errors hide. This diagnosis is harder than it looks, because the judgment was never made explicit while the work built it incidentally, and naming it requires a senior person to articulate what they know that the junior grind taught them, which they have often never had to put into words.

The second movement designs deliberate practice that builds the same judgment differently. Substitution: design a practice that builds the same judgment without the volume of automatable repetition. If the old learning came from reviewing 10,000 contracts, the substitute might be critiquing the machine’s review of curated hard cases, being handed the exceptions the machine flags as uncertain, or working through a designed sequence of instructive examples chosen for what they teach, with the week’s business needs set aside. The substitute deliberately supplies the judgment-building exposure that the automated workflow no longer provides as a byproduct.

The third movement sequences and assesses the substitutes so judgment actually forms. Sequencing and assessment is what turns a good intention into a working ladder. The substitutes have to be staged so that judgment builds in a sensible order, and assessed so that the organization can see whether the capability is actually forming, because a training substitute that no one checks degrades into a ritual as empty as the hazing the chapter warned against. The model describes a standing ladder, and treating it as a course to be delivered once forfeits it, because the judgment it builds is the organization’s future capability and requires the same standing attention as any other asset the organization depends on.

A fourth movement, easy to omit and fatal to omit. Proximity: keep the learner in the room where the judgment is exercised. The first three movements can all be satisfied on paper by a training program, and the old apprenticeship was never a training program. What transmitted was overheard reasoning, the partner thinking aloud about why this clause was the dangerous one, and none of the substitutes reproduce it if the junior is not present when it happens. Designing the practice and then delivering it at a distance rebuilds the syllabus and loses the thing the syllabus was standing in for.

Where the model misleads. Its diagnosis step assumes the automated work built one identifiable judgment, and much entry-level work built several at once, unevenly, in ways the people who did it cannot articulate on request. A leader who runs the diagnosis honestly should expect to end up with a partial list, and should design for the gap it leaves. The model also cannot tell a leader whether the judgment in question is still worth building. Some of what the old ladder transmitted was genuinely specific to work that no longer exists, and rebuilding it faithfully would be an expensive way to train people for a firm that is gone.

Deliberate Practice for Judgment

The substitution movement rests on a distinction the research on expertise makes precise: the difference between mere experience and deliberate practice. Anders Ericsson’s work established that expertise comes from deliberate practice, focused, effortful, feedback-rich activity designed to improve specific aspects of performance, and that years of simply doing an activity supply surprisingly little of it.⁴ The old apprenticeship worked because the volume of junior work happened to deliver a great deal of something close to deliberate practice, with real feedback from senior review and real stakes.

The rebuild shifts from incidental practice to deliberately designed practice. What the volume used to supply by accident now has to be supplied on purpose: focused exposure to instructive cases, effortful engagement with hard examples, and rich feedback on the judgment being formed. Designed well, the deliberate substitute can build judgment faster than the old grind, because it concentrates on the instructive cases and does not dilute them across the mass of routine work the machine now handles, which is the one real opportunity the broken ladder creates.

A deliberate design could build judgment faster than the accidental ladder it replaces. The opportunity is real and should be claimed. The old apprenticeship built judgment inefficiently, through years of repetition most of which taught little, so a design that isolates the instructive experiences has room to beat it. This is the constructive version of the chapter’s argument: the automation of the grunt work is a chance to replace an accidental, wasteful apprenticeship with a designed, efficient one, provided the leader takes up the design task and stops assuming the judgment will still somehow appear.

What Osmosis Transmitted

One caution keeps the rebuild honest: some of what the old apprenticeship transmitted was tacit, and that kind of knowledge is hard to design for. Ikujiro Nonaka’s work on knowledge creation drew the distinction between explicit knowledge, which can be written down and taught directly, and tacit knowledge, the feel, the judgment, and the know-how that resist articulation and are transmitted largely through shared experience and proximity.⁵ A great deal of what juniors absorbed on the old ladder was tacit: the explicit skill of reviewing a contract, and with it the unspoken norms, the war stories, the sense of how the work is really done, picked up by being present around seniors.

A technical rebuild can supply explicit judgment while starving the tacit knowledge osmosis carried. The tacit dimension is what a purely technical rebuild is most likely to miss. It is straightforward to design deliberate practice for the explicit, articulable judgment, the recognition of a dangerous clause, and much harder to design for the tacit knowledge that osmosis transmitted, because that knowledge was never explicit and passed through proximity, which instruction cannot reproduce. A rebuild that supplies the deliberate practice but eliminates the proximity, the junior no longer in the room where seniors work through hard problems, may build the explicit judgment and starve the tacit, producing technically trained juniors who lack the unspoken feel the old apprenticeship conveyed.

The rebuild must preserve proximity for tacit transmission, and it is never finished. Keeping juniors in contact with seniors working through the hard, non-automatable cases, so that the tacit transmission continues even as the routine volume disappears, is part of rebuilding the ladder. Rebuilding the apprenticeship, then, is one piece of a larger imperative, because the judgment that must be rebuilt is not built once and finished; expertise now has a shortening half-life, and keeping people current is a continuous task that the next chapter takes up.

Then and Now: From Osmosis to Design

Then. Professionals became senior through bottom-rung repetition: the articleship, the trainee desk, the junior review, in which years of necessary junior work built accumulated judgment as a byproduct that no one had to fund or design.

Now. Those rungs are now automated, so the junior work that built judgment by osmosis is largely done by machines, and the path from entry to expertise is cut where the apprenticeship used to be.

The rhyme. People still must become senior, and organizations still depend on a supply of accumulated judgment that takes years to build and cannot be summoned on demand. The need for the ladder is unchanged.

The break. The path to seniority is cut and must be rebuilt on purpose, through deliberate practice and preserved proximity, because the incidental apprenticeship that built judgment for free has ended.

The Practitioner’s Lens: The Practice Head Rebuilds the Path

A practice head or partner is worried, correctly, that there will be no next generation of seniors. The junior work that made the partner is being automated, the associates are faster and thinner in judgment, and the partner can see the shortage coming years before it arrives. The task is to redesign how juniors build judgment, deliberately, before the pipeline fails.

The practice head forces seniors to articulate the judgment the old grind built. They sit with the senior people and force the articulation the old ladder never required: what, precisely, did the years of document review teach, and how would anyone know a junior had learned it? This diagnosis, naming the judgment the automated work built, is uncomfortable because seniors often cannot at first say what they know, and it is essential, because a substitute cannot be designed for a capability no one has named. The act of articulating it is itself valuable, turning tacit senior judgment into something the organization can deliberately develop.

The practice head designs the substitutes and defends them against short-term efficiency. They route the instructive hard cases and the machine’s flagged exceptions to juniors as deliberate practice, they build in the senior feedback that turns mere exposure into deliberate practice, and they keep juniors in proximity to seniors working the non-automatable problems so the tacit transmission survives. Each of these costs something now and pays only later, and the practice head has to defend them against a firm that will be tempted to capture the full efficiency of automation and let the ladder rot, which is the choice that produces the judgment crisis a decade out.

The practice head makes the invisible future cost visible and governed. The rebuild has to be argued on the strength of a cost the numbers do not yet show, which is a hard case to win against the visible savings of automation. The practice heads who succeed are the ones who make the pipeline a governed asset, tracked and invested in like any other, so that the firm sees the future capability it is building or consuming, well before the shortage becomes too late to fix.

Applied Exercise

This chapter’s applied work builds a judgment-map for an entry-level role and one deliberate substitute. Expect 3 to 4 hours for the judgment-map and one substitute. Choose a real entry-level role in the reader’s organization whose work is automating.

Step 1. Choose the role and its automating work. Name an entry-level role and the specific junior work, now being automated, through which its people used to build judgment.

Step 2. Map the judgment the work built. With input from senior people, put a name to the judgment the automated work developed, the pattern recognition, error-spotting, or domain feel, making explicit what the old ladder taught incidentally.

Step 3. Design one deliberate substitute. Design a deliberate-practice substitute that builds one named judgment without the automatable volume, such as critiquing machine output on curated hard cases, with the senior feedback that makes it deliberate.

Step 4. Preserve proximity and set the check. Specify how juniors will stay in proximity to seniors for tacit transmission, and how the organization will assess whether the judgment is actually forming. The deliverable is the judgment-map plus the one substitute with its proximity and assessment plan.

Executive Briefing

The chapter’s claims are below, in the order a leader has to act on them.

The apprenticeship worked because nobody had to pay for it or defend it. The junior work was economically necessary, so the training was invisible and self-financing, and the judgment accumulated through years of exposure that no one designed or budgeted. That is also why no one is defending it now. A system nobody built is a system nobody notices losing.

What made the bottom rungs good training is exactly what makes them automatable. Document review, first-pass research, and boilerplate drafting are structured, high-volume, and pattern-based. Those same properties made them ideal for building judgment and ideal for a machine. The education disappears with the grunt work, and it disappears first.

An organization can consume this asset for a decade before anything registers. A firm that stops developing juniors keeps its existing seniors for years, so productivity rises, clients stay satisfied, and costs fall. The shortage surfaces when those seniors retire and no one behind them built judgment the old way. Because no current metric signals the deterioration, the decision to consume the pipeline is never actually taken by anyone.

Naming the judgment the old work built is the step firms skip. The Ladder-Rebuild Model asks first what judgment the now-automated work actually built, which is harder than it sounds, because that knowledge was never made explicit while the work was building it. Then design a deliberate-practice substitute that builds the same judgment without the automatable volume, and sequence and assess it so the capability forms. Skip the first step and the substitute will faithfully train something else.

A technically correct rebuild still starves what osmosis used to carry. Deliberate practice, focused and feedback-rich, replaces the incidental practice the volume supplied, and designed well it can outpace the accidental ladder it replaces. Proximity has to be preserved alongside it, because watching seniors work transmitted tacit knowledge that no exercise captures. And the judgment is never finished, because expertise now has a shortening half-life.

Reflection Questions

The chapter argues that the old junior work built judgment as a byproduct no one designed. For a senior role the reader knows, what did the grunt work actually teach, and can anyone articulate it?

The judgment crisis is invisible because the pipeline is long. How would the reader’s organization know whether it is quietly consuming its future supply of senior judgment?

The Ladder-Rebuild Model begins by naming the judgment an automated rung built. Why is that naming so hard, and who in the organization could actually do it?

Deliberate practice can build judgment faster than incidental repetition. What would a deliberate substitute for one automated entry-level task look like in the reader’s field?

A technical rebuild can supply explicit judgment while starving the tacit knowledge that proximity transmitted. How does the reader’s organization keep juniors in the room where seniors work through hard problems?

Key Terms

Apprenticeship ladder. The path by which juniors historically became experts, through years of bottom-rung repetition that built judgment as a byproduct of necessary work. Cut at the bottom when AI automates the entry-level work.

Ladder-Rebuild Model. The four-movement instrument introduced in this chapter (name the judgment the automated work built, design a deliberate substitute, sequence and assess it, and preserve the learner’s proximity to exercised judgment) for rebuilding the apprenticeship deliberately.

Deliberate practice. Focused, effortful, feedback-rich activity designed to improve specific aspects of performance, as distinct from mere experience. The mechanism the rebuild must supply deliberately once incidental repetition ends.

Tacit knowledge. The feel, judgment, and know-how that resist articulation and pass through shared experience and proximity, where explicit instruction cannot reach. Much of what the old apprenticeship transmitted, and what a technical rebuild risks starving.

Judgment pipeline. The supply of accumulated senior judgment an organization depends on, built over years and impossible to summon on demand. The thing the broken ladder silently cuts.

Incidental versus deliberate learning. The distinction between judgment acquired as a byproduct of doing necessary work and judgment built through practice designed for the purpose. The rebuild is a shift from the former to the latter.

Further Reading

On the nature of expertise, Anders Ericsson’s research is the essential foundation, most accessibly in Peak: Secrets from the New Science of Expertise (with Robert Pool, 2016) and originally in his 1993 work on deliberate practice. The central finding, that expertise is built by focused, feedback-rich deliberate practice, which experience alone does not supply, is exactly what a leader rebuilding the apprenticeship ladder needs, because it specifies what the substitute for the automated repetition must supply.

Turning to tacit knowledge and how it is transmitted, Ikujiro Nonaka and Hirotaka Takeuchi’s The Knowledge-Creating Company (1995) is the standard reference, and its distinction between explicit and tacit knowledge, and its account of how tacit knowledge passes through shared experience, is the caution this chapter draws on: a rebuild that supplies deliberate practice but eliminates proximity will build the explicit judgment and starve the tacit. Read together, Ericsson and Nonaka frame the two halves of the rebuild, the designed practice and the preserved proximity.

On the broader question of professional work under automation, the developing literature on the automation of legal, accounting, and analytical tasks is worth following as it matures, with attention to the difference between the automation of tasks and the development of the people who used to perform them. The chapter’s contribution is to insist that the second question, largely absent from the efficiency-focused discussion, is where the real long-term stakes for an organization’s capability lie.

References and notes

1. Erik Brynjolfsson, Bharat Chandar and Ruyu Chen, Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence, Stanford Digital Economy Lab, version of November 13, 2025, using high-frequency administrative payroll data from the largest payroll software provider in the United States. The findings drawn on here are the approximately 16 percent relative decline in employment for workers aged 22 to 25 in the most AI-exposed occupations after controlling for firm-level shocks; the stability or continued growth of employment for more experienced workers in the same occupations and for workers in less exposed fields; the observation that adjustment occurred through employment levels while compensation held; and the concentration of declines in roles where AI substitutes for human labor. The paper reports associations in recent data and does not settle the question of cause.

2. The claim that AI automates first the structured, high-volume, pattern-based entry-level work that also served as professional training is argued here as a reading of where the technology is heading in the professions, dated to the mid-2020s. No single measurement is offered in support.

3. The Ladder-Rebuild Model originates with this book. The chapter reasons it out as a design aid; no empirical study stands behind it, and each entry-level role whose work is automating will need its own version.

4. K. Anders Ericsson and colleagues on deliberate practice; see K. A. Ericsson, R. T. Krampe, and C. Tesch-Römer, “The Role of Deliberate Practice in the Acquisition of Expert Performance,” Psychological Review, vol. 100, no. 3 (1993), pp. 363–406, and Anders Ericsson and Robert Pool, Peak: Secrets from the New Science of Expertise (Houghton Mifflin Harcourt, 2016).

5. Ikujiro Nonaka and Hirotaka Takeuchi, The Knowledge-Creating Company: How Japanese Companies Create the Dynamics of Innovation (Oxford University Press, 1995), on the distinction between explicit and tacit knowledge and the transmission of tacit knowledge through shared experience.