AI Eroding Leadership Skills: Why 78% of HR Leaders Are Worried, and How to Redesign Your Pipeline

A new Talogy survey of 207 senior HR, talent acquisition and L&D leaders found that 78% worry AI is causing a long-term loss of critical leadership skills, and three independent outlets covered it in the same week. That convergence matters more than the number itself: it means the people running talent pipelines are seeing the same gap from different seats.

This is not another recap of the survey stats. Bauer Media has already built a working response, an apprenticeship redesign called Amplify that recruits on potential and explicitly teaches the skills AI is displacing. This piece turns that model into an audit you can run on your own programme this hiring cycle, covering AI eroding leadership skills, apprenticeship redesign for 2026, and what to actually change in a curriculum, not just what to worry about.

If you run leadership development, graduate schemes or apprenticeships, the practical question is not “is AI a threat.” It is which specific entry-level tasks your own tools have already taken over, and what those tasks used to teach.

What did the Talogy survey actually find about AI and leadership skills?

Talogy surveyed 207 senior HR leaders, talent acquisition managers and L&D experts across the UK and US in September 2026. 78% said they are concerned about a long-term loss of critical leadership skills as AI takes over entry-level work, and 95% said AI is automating work in their organisation at least moderately.

The same survey found 78% of respondents struggle to assess AI-related skills in candidates and staff, and only 38% feel prepared to adapt job descriptions and career paths for an AI-saturated entry-level workforce. Talogy’s Ali Shalfrooshan, Director of Science Strategy, put the core problem plainly: teaching employees to work alongside AI tools isn’t enough, because leadership development has to build human capabilities like critical thinking, adaptability and collaboration on top of that.

Treat the 207-person sample as directional, not definitive. It is a real signal from senior practitioners, not a population-level estimate, so use it to prioritise where to look rather than as a precise industry-wide figure.

What does “loss of critical leadership skills” actually mean in the survey’s own terms?

In Talogy’s own framing, the skills at risk are the ones people used to build by doing entry-level work badly first: critical thinking, adaptability, resilience, clear communication, collaborative thinking, emotional intelligence and values-based leadership. These are not AI-literacy skills. They are the by-products of unglamorous, repetitive early-career tasks that AI now completes faster and better.

That distinction is the whole story. A junior analyst who used to build ten flawed forecast models before getting one right was learning to reason under uncertainty, not just building spreadsheets. If AI now produces the ninth-draft-quality model on the first try, the analyst never struggles, and the reasoning never gets built. The task got done. The learning that used to ride along with it did not.

Why did three outlets cover the same Talogy survey in one week?

HR Dive, Learning News and TrainingJournal all published on this survey within days of each other in late September 2026, which is a stronger trend signal than a single-sourced write-up. Convergent coverage like that usually means the underlying finding matches something practitioners are already seeing in their own data, not just a well-timed press release.

That is the difference between a trend worth acting on and a one-off survey headline. When three separate editorial desks independently decide the same 207-person study is the story of the week, the finding is doing work their audience already recognises.

What is Bauer Media’s Amplify apprenticeship, and why does it matter here?

Amplify is a new paid apprenticeship programme from Bauer Academy, offering entry-level roles across Bauer Media’s UK and European operations in content, communications, finance, data and technology. It recruits candidates on potential rather than prior media experience, then deliberately teaches the skills AI has started to erode: proofreading and attention to detail, professional communication, time management and problem-solving.

Courtnay McLeod, Managing Director of Bauer Academy, has been explicit that skills like proofreading “can be compromised with increased use of AI,” which is why Amplify treats them as things to teach on purpose rather than assume new hires already have. That is the part most coverage of the Talogy survey skips: a named organisation has already built the fix and is running it now, not proposing it as a future idea.

What Amplify does differently from a standard graduate scheme

A conventional graduate scheme screens for existing competence and assumes the job itself will finish the training. Amplify inverts that. Assessment centres evaluate raw capability and attitude rather than a media CV, and the curriculum assumes AI tools are already doing the first draft of most tasks, so the human apprentice is taught the judgment layer explicitly: reviewing AI output critically, communicating decisions, and managing their own time and priorities without a manager filling every gap.

Which entry-level skills has your own AI tooling started substituting for?

Run this as a working session with your training needs analysts and the managers who supervise year-one hires, not as a survey. Pull the job description and the actual weekly task list for every entry-level or apprentice role, then flag every task an AI tool in your organisation now does most of, or all of, unsupervised.

For each flagged task, ask what a person used to learn by doing it badly a few times before getting it right. That is the skill at risk, and it is almost never the task itself. A support agent who no longer drafts first-response emails from scratch is not losing “email writing.” They are losing the practice of reading a customer’s tone and deciding how much empathy versus how much information a reply needs, a judgment call that used to get corrected by a supervisor a dozen times before it became instinct.

Once you have this map, feed the “skill it used to build” column straight into your learning objectives for the redesigned role. If a task disappeared, its objective does not disappear with it. It moves to a deliberate exercise instead.

How do you assess re-taught skills without a generic “AI skills” test?

Skip any assessment that asks candidates to describe AI tools or rate their comfort with them. Those measure familiarity, not the judgment you are trying to protect. Instead, build scenario-based assessments where the candidate reviews a flawed AI output, a wrong forecast, a tone-deaf customer reply, a buggy function, and has to identify what is wrong and fix it under a time limit.

That format tests the exact muscle AI removed the natural practice for: catching your own and others’ mistakes before they ship. If you need a structured way to build these, scenario-based assessment design gives a repeatable method for writing branching, judgment-based scenarios rather than knowledge-recall quizzes.

What good looks like in practice

A strong scenario assessment for this purpose has three parts: an AI-generated artifact with a deliberately planted flaw, a time-boxed review task, and a follow-up question asking the candidate to justify the fix out loud or in writing. The justification step is what separates candidates who caught the error by luck from those who reasoned their way to it, and it is the part most organisations skip.

Why do only 38% of HR leaders feel prepared to adapt job descriptions?

Most job descriptions for entry-level and apprentice roles still list tasks, not judgment. When the tasks get automated, the job description technically still matches what the person does day to day, so nobody flags it for review, even though the actual skill being built has quietly changed or disappeared. That is why only 38% of HR leaders in the Talogy survey feel ready to rewrite these roles: the trigger for a rewrite never fires.

The fix is to write job descriptions and career pathways around the judgment skill, not the task, and to review them every time a task on that list becomes AI-assisted rather than human-led. That single change, task-based description to judgment-based description, is the difference between a role description that ages and one that keeps forcing the review it needs.

Why should succession planning start earlier, not just at senior level?

If judgment is being built later and more deliberately rather than accumulated automatically through years of task repetition, succession planning has to start tracking it earlier too. Waiting until someone is mid-career to assess their decision-making under ambiguity means you find the gap years after the moment you could have closed it cheaply.

Pipeline programmes that previously started identifying “high potential” staff five or more years into a career should pull that assessment forward to year one or two, using the same scenario-based methods described above. This is also where executive and leadership training design needs to connect back to entry-level programme design, rather than treating them as two unrelated tracks that only meet at the promotion committee.

What should you change in a graduate or apprenticeship programme this hiring cycle?

Start with the task-to-skill audit table above, then rebuild the first ninety days of the programme around the skills column, not the task column. Concretely, that means:

Step 1: Reweight recruitment toward potential
Move at least one assessment stage away from prior experience or portfolio review and toward a scenario-based judgment test, following Bauer Media’s model of recruiting on potential rather than a media CV.

Step 2: Rebuild onboarding around AI-output review
Replace “how to use our AI tools” onboarding modules with “how to critically review what our AI tools produce” modules, using real (anonymised) flawed outputs from your own systems.

Step 3: Add mandatory human-only practice reps
For each skill on your audit table, schedule a fixed number of AI-free practice repetitions in the first quarter, not as a punishment but as deliberate practice, the same logic used in any competency-based programme.

Step 4: Rewrite the job description around judgment
Replace task lists with the judgment calls the role requires, and set a review trigger for whenever a listed task becomes AI-assisted.

For the full mechanics of stitching these steps into a formal curriculum, how to build a corporate training programme covers the sequencing and stakeholder sign-off most teams skip when they redesign under time pressure.

How do you measure whether the redesign is actually working?

Do not measure completion rates or AI-tool adoption. Measure the judgment indicators directly: error-catch rate on reviewed AI output, time-to-independent-decision on a defined task category, and manager-rated readiness for ambiguity at defined career checkpoints. These need to sit alongside your existing L&D KPIs so the redesign is accountable to leadership the same way any other training investment is, rather than running as a side project nobody checks on after launch.

Review these numbers quarterly for the first year of any redesign, because the point of the exercise is to prove the judgment gap is closing, not just that a new module exists.

Is this specific to AI, or the same pattern automation has always caused?

Automation has always removed some form of practice, but AI is doing it faster, across more roles at once, and specifically in the reasoning and communication tasks that leadership development has traditionally relied on early-career work to build. Previous waves of automation mostly removed manual and repetitive tasks; this wave is removing first-draft cognitive tasks, which is exactly the terrain leadership judgment is built on.

That is why the response cannot be “wait for the market to adjust.” The organisations already moving, Bauer Media among them, are treating this as a curriculum design problem to solve now, not a labour-market trend to observe.

Conclusion

The Talogy numbers are a useful prompt, not a verdict: 78% of HR leaders worried, 95% seeing AI automate work, and only 38% feeling ready to adapt. What actually moves the needle is running the task-to-skill audit on your own entry-level roles this quarter, not waiting for a bigger study to confirm the concern.

Start with one apprentice or graduate cohort. Map their real weekly tasks against what AI already handles, rebuild the first ninety days around the judgment those tasks used to teach, and set one measurable indicator you will check again in six months. Bauer Media did not wait for consensus before building Amplify, and the programmes that adapt fastest from here will be the ones that treated this as a design problem in September 2026, not a talking point.