Agentic AI refers to systems that can plan and execute multi-step workflows rather than individual tasks. Evidence from consulting, finance, law, and technology firms shows that these systems are automating the analytical and coordination work historically performed by junior professionals. As entry-level roles shrink and AI exposure increases, organizations may gain productivity now while weakening the experiential pipeline that develops future senior decision makers.

This article includes:

Companies mentioned: McKinsey & Company; Goldman Sachs; Morgan Stanley; IBM; Google; Salesforce.

In February, a prior analysis in this publication examined how AI adoption is compressing the apprenticeship pyramid in consulting, banking, and other professional services sectors. That earlier analysis documented slower entry-level hiring, tighter promotion pipelines, and increased reliance on experienced lateral hires rather than internal development. The structural risk identified at the time was that shrinking junior cohorts could weaken long-term leadership pipelines.

Recent evidence suggests a deeper mechanism: the tasks that historically trained junior professionals are themselves being automated.

Across knowledge industries, generative AI systems now perform foundational analytical and coordination work that has traditionally defined early-career roles. In consulting, for example, McKinsey & Company deployed an internal AI platform called Lilli that synthesizes firm knowledge, identifies relevant experts, and suggests research materials. Consultants report time savings of roughly 30 percent using the system, and scoping decks that previously required days of junior analyst work can now be assembled in a few hours.

Similar capabilities are emerging across the consulting sector. AI systems can synthesize multiple market research reports in minutes rather than the eight to twelve hours historically required for human research and synthesis.

Investment banking provides a parallel example. Goldman Sachs CEO David Solomon explained publicly that generative AI can draft an IPO prospectus close to completion in minutes. Work that previously required multiple analysts working for weeks can now be produced rapidly, with human professionals focusing primarily on final judgment and review. Solomon summarized the shift directly: generative AI can now produce documents that are “95 percent of the way there in a few minutes.”

Legal work shows similar compression. According to the Thomson Reuters Institute, AI tools used in large law firms now automate substantial portions of document review, legal research, and due diligence. Analysts have described the consequence bluntly: the traditional law firm pyramid, in which junior associates performed high-volume foundational work, is losing its foundation as AI compresses tasks that once required hours into minutes.

Technology companies illustrate the same dynamic in software development. At Google, more than 25 percent of newly written code is now generated by AI systems and then reviewed by engineers rather than written entirely from scratch. Salesforce CEO Marc Benioff recently linked a roughly 30 percent increase in engineering productivity to coding agents and stated that the company does not expect to hire additional engineers in the coming fiscal year because AI tools can perform a substantial share of underlying development work.

These examples illustrate a broader structural change in how organizations produce expertise. For decades, junior professionals developed judgment through repeated exposure to operational tasks such as research synthesis, financial modeling, document drafting, and preparation of analytical materials under senior supervision. The work itself generated the repetitions through which professionals learned how problems were framed, how decisions were justified, and how errors were corrected.

Generative AI systems increasingly perform many of those tasks directly. Organizations capture productivity gains while simultaneously removing the operational environment in which experiential learning historically occurred.

Employment data is beginning to reflect this shift. Research analyzing occupations with higher AI exposure finds that entry-level roles decline relative to experienced roles as automation intensity increases. One analysis suggests that a ten-percentage-point increase in AI exposure within an occupation corresponds with a measurable reduction in demand for entry-level workers while demand for experienced workers rises.

Industry observers have begun to articulate the long-term implications. One consulting expert summarized the dynamic in a recent discussion: clients increasingly want the senior expert rather than the full team of junior staff historically assigned to projects. The same expert warned that this compression of junior roles cannot persist indefinitely because organizations will eventually lack the experienced professionals that the system previously produced.

Across consulting, finance, law, and technology, the pattern is consistent. AI systems rarely replace senior judgment directly. Instead, they replace the analytical and coordination work that historically trained future decision makers.

The shift underway is not limited to automation of individual analytical tasks. Increasingly, AI systems are executing portions of entire workflows.

Traditional automation focused on discrete activities such as drafting text, summarizing research, or generating code. Agentic AI systems can plan, coordinate, and execute sequences of work that historically required junior professionals to move information across steps, tools, and stakeholders.

Those sequences were the environment in which apprenticeship occurred.

In consulting firms, a traditional junior workflow often involved gathering background research, synthesizing firm knowledge, building draft analyses, preparing presentation materials, revising work after senior feedback, and coordinating updates across teams. The Lilli platform deployed by McKinsey & Company now performs substantial portions of this process automatically by synthesizing internal knowledge bases, identifying relevant content, and assembling draft materials. Consultants report that workstreams that once required days of junior analytical effort can now be produced in a fraction of the time.

Financial institutions are deploying similar capabilities. Morgan Stanley introduced an AI Assistant used by a large share of the firm’s financial advisor teams. The system synthesizes research documents, prepares client briefings, and supports meeting preparation. Companion tools generate structured meeting notes immediately after conversations with clients. Tasks that historically required junior analysts to gather materials, draft summaries, and circulate updates are increasingly executed directly by AI systems embedded in the workflow.

Professional services firms are beginning to articulate explicit strategies around these capabilities. IBM has described a future consulting architecture in which human professionals operate alongside large numbers of “digital workers” capable of executing analytical and operational steps across engagements. The firm has publicly discussed a model in which thousands of consultants work with far larger numbers of AI agents performing underlying execution tasks.

This represents a structural shift. Earlier generations of automation supported junior professionals while they moved through analytical processes. Agentic systems increasingly run portions of those processes themselves.

As a result, organizations retain the output while removing the experiential path that historically produced it.

Industry analysts have begun describing the resulting dynamic as a “skill experience paradox.” In this emerging model, junior employees are expected to develop analytical judgment while interacting with systems that bypass the intermediate steps through which that judgment historically formed.

The implications extend beyond productivity metrics. The traditional apprenticeship model relied on repeated exposure to complex, multi-step work processes where junior professionals learned how problems were framed, how data was interpreted, and how decisions evolved through interaction with experienced colleagues. These developmental repetitions occurred naturally because the workflow required them.

Agentic AI compresses that environment.

Organizations still produce research briefs, financial analyses, legal memoranda, and client presentations. What disappears is the sequence of intermediate steps through which junior professionals previously learned how to construct those outputs.

The productivity benefits of agentic AI appear immediately. The developmental consequences unfold gradually because the work that historically trained future experts is increasingly absorbed into the systems that now execute it.

The immediate impact of agentic AI is visible in productivity metrics: organizations produce analytical outputs faster while employing fewer people in foundational roles. The longer-term implication is less visible. When the workflow layer that trained junior professionals disappears, organizations may preserve output today while weakening the pipeline that produces experienced professionals tomorrow.

Historically, knowledge industries operated through a pyramid structure. Large cohorts of junior employees entered organizations, accumulated experience through repeated exposure to real work, and over time a smaller number advanced into managerial and senior advisory roles. The structure produced both operational capacity and a long-term leadership pipeline.

AI compresses the base of that pyramid.

Research across multiple sectors indicates that entry-level roles are declining more sharply than experienced roles in occupations with high AI exposure. Organizations increasingly substitute automation for junior labor while augmenting experienced professionals above it.

From a short-term operational perspective, this structure appears efficient. Firms reduce costs, accelerate execution, and concentrate work among smaller teams of experienced professionals supported by AI systems. The risk emerges later if the cohort that would normally progress into mid-career roles never existed.

Industry commentary increasingly describes this dynamic as a potential pipeline crisis. Corporate presentations and analyst discussions frequently raise the same question: if junior professionals form the pipeline for future experts and leaders, what happens when that pipeline narrows significantly?

Researchers studying AI-mediated work environments warn that organizations may be optimizing for near-term productivity while unintentionally weakening the development of professionals who would sustain long-term expertise. The financial benefits of replacing junior analysts with AI systems appear quickly, but the absence of those cohorts may only become visible years later when promotion pipelines thin.

Similar patterns have appeared historically in industries undergoing automation transitions. When entry-level roles disappear, organizations often attempt to compensate through lateral hiring, accelerated promotions, or acquisition of experienced talent. These approaches can temporarily address staffing shortages but rarely rebuild the underlying capability formation system.

For learning and talent leaders, the implication is structural. The challenge is no longer limited to delivering training programs efficiently. The deeper issue is designing the environments through which professionals develop judgment in an AI-mediated workplace.

Under the traditional model, organizations generated those learning environments automatically through the structure of work. Agentic AI reduces that structure.

Some organizations are beginning to experiment with alternatives, including programs that train employees to supervise AI systems, simulation-based development environments that recreate complex work scenarios, and accelerated rotational programs that expose junior professionals to higher-level decision-making earlier in their careers. These initiatives reflect an emerging consensus that capability formation may need to be designed intentionally rather than assumed to emerge naturally from daily work.

Whether these approaches can replicate the developmental density of the traditional apprenticeship system remains uncertain.

What is already visible is that organizations adopting agentic AI are reshaping the mechanisms through which their future leaders will be formed. Productivity gains from AI adoption are appearing rapidly. The consequences for leadership pipelines will likely emerge more slowly but could prove equally consequential.

Learning and Development Executive Intelligence is for CHROs, CLOs, and senior L&D buyers investing in internal talent development, training, and reskilling.

This is one of our six education and learning-related publications spanning K-12, Higher Education, and Workforce. Our education newsletters reach tens of thousands of senior decision-makers across the U.S. and key international markets.

Ping us at [email protected] if you’d like to learn more, explore Enterprise Subscriptions, or would like to partner in other ways.

The Intelligence Council is a next-gen B2B media and business intelligence platform built for people who make strategy, allocate capital, and carry operating risk.