Your Junior Employees May Already Be Your Best AI Strategists

AI adoption is growing from the edges inward

Picture the typical executive AI meeting. Senior leaders debate platforms, budgets, risk, and transformation roadmaps. Meanwhile, a first-year analyst has already built a faster way to research customers, prepare a briefing, compare contracts, and draft the follow-up email before the meeting ends.

The analyst may not call it transformation. But that is where transformation is beginning.

New research released this week offers an unusually detailed look at how people actually use generative AI inside organizations. After analyzing more than 17 million ChatGPT Enterprise messages across 1,764 organizations, researchers found that AI use spans functions and seniority levels—but its intensity declines as seniority rises. Early-career employees and trainees sent roughly eight to nine more messages per week than the average active user in the same organization. Executives sent fewer.

That finding should make leaders curious, not defensive.

The same study found that ChatGPT Enterprise output grew about sevenfold between June 2025 and March 2026. Nearly half of that growth came from organizations that had already adopted the platform—not just from new customers joining.

In other words, adoption is not a switch companies turn on. It deepens as employees discover more reasons to use the technology.

And those reasons are remarkably broad. More than half of active users performed documentation or technical-writing tasks. Nearly half performed technical digital work. Others used AI for research, planning, sales, marketing, legal work, data analysis, finance, and regulatory questions.

This is less like installing a new payroll system and more like introducing the spreadsheet. No central team can predict every valuable use. Employees discover applications while doing the work.

That creates a management problem. If leaders only track licenses, logins, or training completion, they can miss the real story: where AI is changing the flow of work, who is learning fastest, and which individual practices deserve to become organizational capabilities.

The power users may not have organizational power

Early-career employees have several reasons to become heavy AI users. They often perform the research, drafting, documentation, analysis, and coordination tasks where generative AI can provide immediate leverage. They may also be more willing to experiment because they have fewer established routines to protect.

But the people discovering useful applications are not always positioned to redesign the process around them.

A junior employee may reduce a four-hour task to 45 minutes, yet keep the method private because the organization has no approved way to share it. A team may celebrate the time saved without changing the downstream review process. Or a manager may simply assign more work, converting an innovation into invisible labor acceleration.

That is how companies end up with pockets of impressive AI use and very little enterprise transformation.

The danger runs in both directions. Unmanaged experimentation can expose confidential information, introduce unreliable output, or create processes that depend on one employee’s personal technique. Over-management can suppress the very experimentation the organization needs to learn what AI is good for.

The answer is not to stop employees from discovering. It is to give discovery a path into the operating model.

Productivity can come with a hidden tradeoff

A second study released this week examined Microsoft 365 activity across several large international companies. Among intensive AI users, the researchers observed a 21.2% increase in productivity-oriented actions and a 7.1% increase in communication actions over a 20-week period.

The result sounds positive—and it may be. But the imbalance matters. Work shifted relatively toward individual, document-focused activity. The researchers warned that organizations should watch whether efficiency gains weaken interpersonal communication and the exchange of diverse information that supports innovation.

This is an important reminder: AI does not merely help people perform the same work faster. It changes the mix of activity.

An employee who can draft, analyze, and summarize more independently may need fewer routine interactions. That can reduce meetings and email overload. It can also reduce mentoring, context-sharing, healthy disagreement, and the informal learning that helps junior employees develop judgment.

Companies should not assume that more output automatically means a stronger organization.

Turn hidden expertise into a visible system

Leaders do not need another enterprise-wide brainstorming campaign. They need a disciplined way to observe what is already happening.

Start by identifying teams with unusually strong, responsible AI usage. Ask them to demonstrate the work before and after AI—not just the prompt. Capture the original cycle time, handoffs, quality checks, data used, decisions made, and exceptions encountered.

Then choose a few practices worth testing beyond the original employee or team. A useful experiment should answer five questions:

1. Does it improve a business outcome that matters?

2. Can another employee reproduce the result?

3. Can the output be checked reliably?

4. Does it preserve data, customer, and regulatory safeguards?

5. What human interaction or learning might be lost if the practice scales?

If the answers are strong, formalize the workflow, assign ownership, train the next group, and measure the result. If they are weak, keep learning before scaling.

This approach respects both sides of the evidence. AI value often begins with individual initiative, but durable value requires organizational design.

The next leadership pipeline may already be forming

There is a larger talent implication here.

Early-career AI power users are not merely learning a tool. The best are learning how to decompose work, test alternatives, judge output, combine human and machine strengths, and redesign a process around a new capability. Those are management skills for an AI-enabled organization.

Executives should be asking which employees are demonstrating that judgment—and whether the company’s promotion, development, and recognition systems can see it.

The future AI leader may not be the person who speaks most confidently about transformation in the boardroom. It may be the person who quietly changed how the work gets done and can explain why the new way is better.

The organizations that learn to notice that difference will move faster than those waiting for innovation to arrive from the top.

Questions for executives

1. Where are employees already using AI more intensively than leadership realizes?
2. Which high-performing individual practices could become safe, repeatable workflows across a team?
3. Are our productivity gains strengthening—or quietly weakening—mentoring, collaboration, and organizational learning?