Dr. Christophe Kolb is founder and CEO of Taller and co-author of Cognitive Kin: How to Work, Win, and Make Meaning with Agentic AI.
AI is entering knowledge work through the smallest unit of production: the task. It drafts the memo, summarizes the transcript, generates the code stub, organizes the first deck and finds the precedent. A worker’s job title may remain familiar, but the economic content of the job is changing beneath it.
If routine cognitive output becomes cheaper, leaders can move faster with smaller teams, at least in the short run. Many of the tasks now being compressed were once the training ground for professional judgment. The first memo, first model, first document review and first reconciliation did two jobs at once: they shipped, and they taught. Remove the routine work too casually, and the firm saves hours today while draining tomorrow’s pool of people able to verify, decide and take responsibility.
The new scarcity is verification capital.
Treat AI As Task Recomposition
Every job is a portfolio of tasks, and AI is repricing them unevenly.
AI lowers the cost of search, drafting, summarization, classification, coding scaffolds and first-pass analysis. These activities have digital inputs, recognizable output formats and reviewable quality. Human value then migrates toward question selection, exception handling, judgment, persuasion, coordination and accountability. When answers become abundant, the scarce executive skill becomes judging which answer to trust.
In research I co-authored on how AI recomposes entry-level roles, we modeled five professions task by task. For every 100 hours of pre-AI work, between 57 and 73 survived. The pattern matters more than any single number. The work that remains clusters around client context, source grounding, integration, sign-off risk, stakeholder management and judgment under pressure.
The employment data is starting to agree. Using ADP payroll records, Stanford researchers find employment for 22- to 25-year-olds in the most AI-exposed occupations now sits 19% below where it would have been had it kept pace with their less-exposed peers, while employment for experienced workers held steady.
AI compresses routine cognition; it raises the premium on people who can frame the problem, test the answer and own the consequence.
Redesign Junior Work Around Verification
The traditional apprenticeship model relied on incidental learning. Juniors pulled comparables, cleaned data, built slides, checked citations, read cases, updated tickets and reconciled accounts. Much of that work was tedious, but it also taught pattern recognition. A junior banker learned why a comparable felt wrong. A paralegal learned why a clean-looking citation failed. A developer learned how a small dependency broke a living system. An auditor learned how a variance revealed a process story.
AI can remove enough of this exposure to create a capability gap. The answer is purposeful apprenticeship design.
Every AI-enabled workflow should define four things: question rights, review standards, escalation triggers and named ownership. Those protocols turn AI into a place where juniors are taught rather than bypassed. Juniors should spend less time producing first drafts from scratch and more time comparing AI output against source material, identifying failure modes, explaining their verification logic and presenting tradeoffs to senior reviewers.
Build A Judgment Ladder
Firms need a new ladder for early-career work to replace the lost opportunities for incidental learning. The first rung should be assisted production: using AI to create drafts, code, models or summaries. The second rung should be verification: checking grounding, provenance, assumptions, edge cases and compliance. The third rung should be exception handling: diagnosing why a plausible answer fails. The fourth rung should be accountable recommendation: explaining a decision under uncertainty to a client, partner, manager or product owner.
That ladder should be engineered into daily work. Firms can build libraries of verified work and run supervised audits of AI output in which juniors hunt for omissions and hallucinated confidence. They can rotate early-career staff onto exception teams, hold reviews on errors, near misses and judgment calls, and even use simulations for rare, high-stakes cases. Juniors can be tasked with writing short “confidence notes” explaining why an output should be trusted, revised or rejected.
This gives leaders a practical rule: every hour saved from routine production should be partially reinvested in structured verification and judgment practice. Pure efficiency extraction may improve this quarter’s utilization, but capability reinvestment builds the bench of future partners, principals, managers and architects.
Measure The Capability Pipeline
AI adoption goes wrong when firms measure output alone. Leaders need metrics for the apprenticeship system itself.
How many juniors encounter ambiguous, consequential work each month? Which error types do they catch, and which do they miss? How many AI-assisted drafts can a senior responsibly supervise before attention becomes the bottleneck? How quickly do juniors move from assisted production to credible recommendations? Which decisions require named ownership, and where does ownership become blurred?
These are management metrics, talent metrics and risk metrics at once. A firm that generates more documents, tickets or analyses with fewer learning loops may look productive while its expertise base erodes. A firm that pairs AI with disciplined review can increase output and accelerate judgment formation at the same time.
The Executive Choice
AI gives leaders cheaper answers. It also forces a choice about the architecture of expertise.
A weak organization will let AI hollow out junior work and call the result productivity. A strong organization will separate production from learning, then rebuild it with intention. It will give juniors better tools, richer feedback, more explicit standards and earlier exposure to protected ambiguity. It will treat verification as a core professional skill. It will make accountable judgment the center of workforce design.
The winners in AI-enabled knowledge work will be the firms that understand the hidden apprenticeship tax. They will save time, then invest some of that time in the people who must carry responsibility when the answer is consequential. Cheap cognition is easy to buy, but sound judgment takes time to build.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
