← All articles

AI Leadership

How Junior Staff Learn the Work While Using AI

A practical apprenticeship exercise that combines a first attempt, AI assistance, an explanation, and mentor feedback on real work.

By , Co-Founder & CEOUpdated 4 min read
Wooden joints and component pieces are compared under a magnifier.

A junior employee submits a polished customer response but cannot explain why it promises a replacement. The wording looks professional. The employee has not learned how your replacement policy applies to that case.

You can keep AI assistance and teach the underlying judgment. Give the employee a task with a small amount of protected practice, then ask them to explain the decision before a mentor accepts the result. The practice should target a skill they need on the job, not a contest to see who can work without tools.

Pick one decision the employee needs to understand

For a hypothetical equipment supplier, the task is to answer a customer whose product arrived damaged. The employee must distinguish a shipping claim from a warranty problem, identify missing evidence, and avoid promising a remedy they lack authority to approve.

Use a training case with invented customer details. Give the employee your approved policy and ask for an initial recommendation before they consult AI. A mentor can then see the employee's reasoning rather than infer understanding from a finished email.

StageWhat the employee producesWhat the mentor checks
First attempt“I would request delivery photos and order details before approving a replacement.”Did they identify the missing facts and their authority limit?
AI assistanceA draft response using the supplied policy, with uncertain points flagged.Did the employee keep private details out of the training tool and avoid invented policy?
Explanation“I rejected the suggested refund promise because our policy requires a claims review.”Can they explain a change using the policy rather than the tool's confidence?
Mentor feedbackA revised response plus one rule to apply on the next case.Can they apply that rule to a different example?

The first attempt can be short. It does not need to recreate the whole task manually. Ask the employee to identify the decision, the evidence, and the unresolved question. That gives the mentor enough to diagnose the skill gap without adding hours of homework.

Teach the part AI can hide

A 2025 Microsoft Research study surveyed 319 knowledge workers about AI use and critical thinking. It examined self-reported effort and confidence; it did not measure permanent loss of ability. Treat it as a reason to examine how employees reason about AI-assisted work, not as proof that AI damages junior staff.

In the supplier example, the mentor should spend time on the replacement decision, not on whether the employee wrote the greeting alone. AI can help shorten or clarify an answer after the employee understands the policy. The learning goal concerns the promise your business makes to a customer.

Ask for one rejected suggestion. “The tool recommended a refund; I removed it because…” reveals more than asking whether the employee checked the output. If they cannot name a reason, revisit the policy together. If the policy itself leaves the decision unclear, fix the policy rather than blame the employee.

Make the next case less rehearsed

Give the employee a second case with one changed fact: the product failed six months after delivery instead of arriving damaged. Ask them which part of the earlier reasoning still applies. The mentor should look for transfer of understanding, not repetition of a memorized answer.

Keep a small record of the skill practiced, the feedback, and the next task. After several cases, the supervisor can reduce support for routine work while preserving escalation for unusual circumstances. Do not treat confidence or writing fluency as a substitute for that evidence.

Agree on what counts as understanding

In the training example, use a small acceptance rubric: the employee identifies the required evidence, stays within replacement authority, and explains the escalation. Mark each separately. Strong phrasing should not compensate for an incorrect decision, and an awkward first draft should not hide sound reasoning. Give the employee the rubric before the task so they know what they are practicing. When a mentor rejects the answer, connect the feedback to the specific skill rather than asking the employee to “be more careful.”

Our team adoption guide covers bringing AI into day-to-day work. For leadership that coordinates training, business ownership, and use-case decisions, see fractional CAIO support.

Luke Thompson

About the author

Luke Thompson

Co-Founder & CEO, The Ops Guide

Luke Thompson is Co-Founder and CEO of The Ops Guide. He writes about practical uses of AI, workflow automation, custom software, and SEO for growing businesses.

About The Ops Guide

The Ops Guide helps businesses improve how work gets done through custom software, workflow automation, AI systems, and SEO services.

The Ops Guide Brief

Get AI news you can use.

The developments worth your attention, what they change at work, and what to do next. Choose a daily briefing or a weekly selection.

Free to subscribe. Confirm by email. Unsubscribe any time.