Mentoring in the AI Era: What to Teach When AI Does the Basics

Mentoring in the AI era shifts from teaching execution to teaching judgment, communication and review. What the evidence says, how to run a session, pitfalls.

info Quick Definition
Mentoring in the AI era is guidance from a more experienced person to someone starting out, in a context where AI handles much of the basic execution. The focus moves from teaching people to produce toward teaching them to judge, explain and review what they produce.

What changes in what gets taught?

An AI tool can give you a first version of a screen, a summary or a journey map in seconds. What it does not do for you is decide which problem is worth attacking, justify a decision to a team, or check that a solution works. That is where a mentor adds the most.

A Toptal article (last updated May 2026) records two useful positions. Gytis Markevicius, a digital product designer, says AI can help a junior with the basics, such as persona examples, journey maps or definitions of concepts, but that for soft skills such as project management, leadership and communication, “I don’t think there’s anything better than to have someone mentoring you.” Darrell Estabrook, who describes himself as a design coach, worries that a junior will grab the most convenient AI model and assume its answer solved the problem, so he tries to encourage them to branch out from that.

There is also experimental evidence, though not from design. In January 2026, Anthropic published a randomized experiment with 52 software engineers, most of them junior, who learned a Python library they did not know. The group with an AI assistant averaged 50% on a later quiz and the group that coded by hand averaged 67%. The largest gap was on debugging questions, and the speed gain (about two minutes) was not statistically significant. The participants who kept more learning generated code and then asked follow-up questions to understand it, requested code together with explanations, or asked only conceptual questions. The authors flag their own limits: a small sample, a single programming task, and comprehension measured right afterwards. Read it as a hint that using AI without understanding what it returns costs learning, not as a rule for design.

Which skills weigh the most?

According to what the sources say, three groups:

  • Creative thinking and curiosity. Estabrook says that for someone starting out, mastering every technical process matters less than thinking creatively.
  • Explaining decisions. Besides asking whether a candidate can use AI, Estabrook says he wants them to show why they chose what they chose. It is the skill a mentoring session can practice directly.
  • Soft skills. Project management, leadership and communication, which Markevicius points to as the ground where a human mentor pays off most.

How to train each one is a proposal from this wiki, not a finding from the sources. For the context of someone starting out, see being a junior designer in the AI era and learning design with AI as a tutor.

How should a session work?

A useful session asks something specific of the junior:

  1. Bring the reasoning, not only the result. What they asked the AI, what they discarded, and why.
  2. Bring what they did not understand. A paragraph they cannot explain is the best starting point.
  3. Review one case with judgment. The mentor asks what was left unverified, not how good it looks.
  4. Close with a task. One thing the junior will do without AI to practice judgment.

Five fixed questions work for the case review, and the mentor repeats them every session: what problem was this solving? what other solution did you discard, and why? what did you verify before showing it? what part do you not fully understand? and what would you do if the main constraint changed? The first four reveal judgment. The fifth shows whether the junior can think beyond the screen. Repeating them has a practical advantage: the junior learns to anticipate them and arrives with answers already thought through.

How do you track progress?

A simple log is enough: date, case name, one decision taken, the reason, and one question left open. Each month, mentor and junior go through it and mark the decisions where they changed their mind. The list lets you see judgment evolve case by case.

The log also prevents a common error: grading by the latest screen. If a decision made three months ago still holds up, the progress is there even if today’s visual result is less striking.

What if the senior has no time?

Carly Ayres describes it in an April 2025 article: senior designers are burnt out, too busy or too bewildered by AI to mentor, and people entering the field do so without structure or support.

Two practical exits. For the senior: a short routine, such as half an hour a week reviewing one case with the five questions, works better than a long session that never gets scheduled. For the junior, Ayres advises asking for mentorship with curiosity about the person’s specific work, asking for advice rather than a complete career solution, making specific requests that are easy to accept (for example, offering a few time options), and remembering that mentorship runs both ways: juniors also bring a fresh perspective and curiosity.

When can human mentoring not be replaced?

When the topic is the relationship with a team, negotiating scope, or how to tell a client about a mistake. In those conversations AI can help prepare material or rehearse a message, but it cannot replace a person who knows the situation and can answer for it. It is the same line Toptal draws for soft skills.

Illustrative example

Illustrative example: a junior designer brings her mentor a sign-up screen generated with AI. The mentor does not comment on the colors. She asks: “What did you do with the phone field, and why?” The junior does not know. Together they review the case, and the mentor asks her to write, before the next session, three decisions in the screen and the reason for each. The session lasts twenty minutes. What mattered was not the screen but that the junior started justifying what she approves.

Common mistakes

  • Teaching tools instead of judgment.
  • Correcting the visual result instead of the reasoning.
  • Letting AI write the mentor’s feedback.
  • Mentoring only the technical side and forgetting communication.
  • Waiting for the ideal amount of time before starting.
  • Assuming a programming study applies unchanged to design.

Sources

Frequently asked questions

What should a mentor teach a junior designer today?
Judgment about the problem, communication and how to review what AI generates. According to Toptal, AI helps with the basics, but what weighs most in a junior is creativity and being able to explain why they chose something.
Does AI replace the mentor?
For basic concepts it can help. A designer interviewed by Toptal thinks that for soft skills such as communication and leadership, there is nothing better than having someone mentor you.
Does using AI affect how much a junior learns?
An Anthropic experiment with 52 engineers, most of them junior, found that people who learned a new library with an AI assistant scored 50% on a quiz, against 67% for those who coded by hand. It is a programming study, not a design one.
How should a junior use their mentor when working with AI?
By bringing their reasoning, not only the result: what they asked the AI, what they discarded and why, and what they could not understand.
What if senior designers do not have time to mentor?
Carly Ayres describes seniors who are burnt out, too busy or bewildered by AI. A practical answer is a short, recurring case review instead of a long session that never gets scheduled.