Where the Cracks are Starting to Show: Mentoring Under Pressure

‍ ‍A companion piece to "Will mentors replace leaders in a hybrid world?" (15 April)

‍Last week's post argued that mentoring behaviors may be better suited than traditional leadership behaviors to generating commitment and retention in hybrid, project-based organizations.  That argument stands.  But it is worth looking at the same landscape from a different angle, because mentoring is not only a solution to the problems facing modern organizations.  It is also a practice that modern organizations are quietly reshaping, and not always in ways that strengthen it.

‍Three forces are at work in mentoring right now.  Each is well-documented in the research literature.  What is less well-documented, and where the interesting questions sit, is how they interact.  Because when you put them together, the picture that emerges is not straightforwardly encouraging.  Early cracks are appearing in the mentoring model that most of us learned, and anyone serious about mentoring in the next five years needs to understand where those cracks are and what they mean.

Three forces, briefly

The first force is the shift from permanent employment toward project-based work.  The numbers vary by source, but the direction is consistent.  Somewhere between a third and a half of the US workforce now does some form of freelance or project-based work, and the trajectory is upward in most developed economies.  The academic concepts that describe this shift – the protean career (first named by Hall in Careers in Organizations, 1976, and developed across a quarter-century of work summarized in Hall, 2004) and the boundaryless career (Arthur and Rousseau, 1996) – are no longer theoretical curiosities.  They describe how a large and growing fraction of the working population actually lives.

‍The second force is hybrid and remote working, which has now settled into a stable pattern rather than reverting to pre-pandemic norms.  Around 27% of the US workforce continues to work remotely, with strong preferences for not returning to the office full-time.  A steadily building body of research, including a recent five-wave longitudinal study of 512 employees by Efimov and colleagues, documents that prolonged remote work erodes organizational belonging, particularly for employees who joined after the shift and never experienced the in-person equivalent.

The third force is the integration of generative AI into daily professional work.  Workplace adoption of generative AI rose from 22% in 2023 to 75% in 2024, one of the fastest enterprise technology adoption curves ever recorded.  AI now routinely handles tasks that five years ago would have generated a conversation with a senior colleague: feedback on a draft, structured reflection on a difficult meeting, a sounding board for a decision.

‍Each of these, on its own, has been analyzed to death.  The harder question is what happens when you put them together.

Where the first cracks appear

Consider what traditional mentoring assumes.  It assumes that the mentor and mentee share an organization long enough for a relationship to develop and pay off.  It assumes sufficient physical, or at least synchronous, contact for the mentor to observe the mentee in action and offer feedback grounded in that observation.  And it assumes that the mentee's developmental conversations will happen primarily with humans, so that the mentor remains a meaningful source of guidance rather than competing with a more convenient alternative.

‍All three assumptions are weakening simultaneously.

‍The tenure assumption is the most straightforward casualty.  If a mentee is likely to leave within two or three years and may spend significant stretches of that time on projects outside the mentor's sightline, the developmental arc that mentoring traditionally relies on is compressed.  The mentor who plans to "see them through the next five years" is planning for conditions that may no longer apply.  This does not make mentoring useless, but it does mean its shape has to change.  Recent research on protean and boundaryless careers suggests that mentees increasingly need help building portable reputations and perceived employability across organizations, rather than progression within a single organization.  That is a different kind of conversation from the one most mentors were trained to have.

‍The observation assumption is weakening differently.  Hybrid working has not destroyed mentoring relationships, but it has thinned them.  A mentor who sees their mentee twice a month on video is working with a much narrower data set than a mentor who used to share an office floor.  The incidental observations – how someone handles pressure, how they show up in meetings they did not know the mentor was watching, what they are like when they think nobody important is looking – are largely gone.  What replaces them is the mentee's own self-report, which is useful but not the same thing.  Mentors are being asked to offer guidance based on evidence they cannot independently verify.  Good mentors have always had to do some of this.  Now they have to do it all.

The interaction nobody is talking about

It is the third force, AI, where the most interesting interaction with the other two sits.  The conventional conversation about AI and mentoring is an either/or: either AI coaches will replace human mentors, or they will not.  The evidence on that question is genuinely conflicted, but the question itself is probably the wrong one.

‍The more useful question is what happens to human mentoring when AI occupies the developmental space adjacent to it.  Consider the interaction carefully.

‍A mentee in 2026 has access, twenty-four hours a day, to a patient and competent interlocutor who will help them think through almost any professional problem.  That interlocutor remembers nothing between sessions, cannot observe the mentee in context, has no stake in the mentee's long-term development, and cannot hold the narrative of a career across years.  But it is available at 11 pm on a Tuesday, and a human mentor is not.

‍In a world where the mentee is isolated by hybrid working and increasingly defined by a portfolio of projects rather than a single career, the pull toward the AI solution is very strong.  It solves the access problem posed by hybrid working.  It solves the just-in-time problem inherent in project work.  What it does not solve, and structurally cannot solve, is the identity problem: the need for someone who can help you see the throughline across your projects, the pattern in your choices, the professional you are becoming rather than the task you are currently completing.

‍A 2025 global study commissioned by Workday captures the underlying tension.  82% of individual contributors agreed that employees will crave more human connection as AI usage grows.  Only 65% of managers agreed.  The 17-point gap between those two numbers is a warning sign.  It suggests that the people affected are already sensing something that the decision-makers have not yet processed.

What this means for mentors

If the diagnosis above is roughly right, it implies a specific set of moves for anyone taking mentoring seriously over the next few years.  Chapter 17 of From Manager to Mentor focuses specifically on virtual mentoring, and the argument there is that the core skills of good mentoring translate well across a screen.  That remains true.  But the landscape has shifted enough since that chapter was written to add some practical observations.

The first is that mentors increasingly need to know what AI is good at and what it is not, because their mentees are using it whether or not the mentor acknowledges it.  The mentor who pretends AI is not in the room is doing their mentee a disservice.  The mentor who understands the division of labor – AI for drafting, structured reflection, and task-level feedback; the human mentor for judgement under ambiguity, identity formation, and honesty that carries weight because it comes from someone who has something to lose by being honest – can be genuinely useful in a way that AI cannot replicate.

‍The second is that mentors in hybrid environments need to be more deliberate about what they used to do incidentally.  The corridor conversation has to be scheduled now.  The "how are you actually doing" question has to be asked explicitly rather than inferred from body language.  The mentor who waits for organic cues is waiting for a signal that no longer reliably arrives.

‍The third, and perhaps the most important, is that mentoring in a project-based world must broaden its scope.  The implicit question in traditional mentoring was "how do you succeed here?" The more useful question now is "how do you build a professional identity that travels?" That shift is not cosmetic.  It changes what a mentor listens for, what they ask about, and what counts as a good conversation.  It also has an uncomfortable implication for organizations: that the mentor who does this well is building loyalty to the person rather than to the firm, and that this is a feature rather than a bug.  People who feel genuinely mentored tend to stay; those who feel their mentor is merely a delivery mechanism for the organization's retention strategy do not.

The honest conclusion

Last week's post made the case that mentoring behaviors are well-suited to the world we now work in.  That case still holds.  But it would be complacent to treat mentoring as a finished solution that merely needs wider adoption.  The same forces that make mentoring important are also reshaping what mentoring has to do to remain useful.  The model most of us inherited – patient senior person, long-tenured mentee, shared organizational context, plenty of observation time – assumes conditions that are receding.

‍ The good news is that the underlying skills travel.  Listening, questioning, holding space for someone else's thinking, telling the truth with care: these survive the transition from office to hybrid, from single-employer to portfolio, from pre-AI to AI-saturated.  What changes is the context in which those skills must be deployed, and the kind of self-awareness the mentor needs about where their contribution fits within a mentee's broader developmental ecosystem.

‍The mentors who will matter most over the next five years are not the ones who do traditional mentoring more carefully.  They are the ones who have honestly looked at where the old model is cracking and adjusted their practice to match the world as it actually is.

Sources

‍Arthur, Michael B., and Denise M. Rousseau, eds. 1996.  The Boundaryless Career: A New Employment Principle for a New Organizational Era.  New York: Oxford University Press.

‍Efimov, Ilona, Annika Krick, Volker Harth, Jörg Felfe and Stefanie Mache.  2025.  "When do employees feel isolated when working from home?  Longitudinal trajectories, antecedents, and outcomes of workplace isolation during the COVID-19 pandemic." Frontiers in Psychology 16: 1601214.

‍Hall, Douglas T. 1976.  Careers in Organizations.  Pacific Palisades, CA: Goodyear.

‍Hall, Douglas T. 2004.  "The protean career: A quarter-century journey.  "Journal of Vocational Behavior 65, no. 1: 1-13.

‍Kundi, Yasir Mansoor, Sandrine Hollet-Haudebert, and Jonathan Peterson.  2021.  "Linking protean and boundaryless career attitudes to subjective career success: A serial mediation model." Journal of Career Assessment 29, no. 2: 263-282.

‍Nimmi, P. M., Clement Cabral, Georgia Thrasyvoulou, Marco Giovanni Mariani, and Gerardo Petruzziello.  2024.  "The impact of protean career on career sustainability: Mediating effect of perceived employability." Australian Journal of Career Development 33, no. 2.

‍Workday.  2025.  Elevating Human Potential: The AI Skills Revolution.  Global study commissioned by Workday, January 2025.

‍ ‍