The companies racing hardest into AI are quietly dismantling their ability to govern it

The more aggressively an organisation automates with AI, the more it will depend on experienced human judgement to oversee the result. Yet those are precisely the people being made redundant right now - often with AI itself as the cover story. The data says this is not efficiency. It is fragility.

There is a paradox running through almost every AI transformation programme of 2026, and very few boards are naming it.

The harder a company pushes AI into its operations, the more it will need experienced people to govern what it has built - to notice when an autonomous system is plausibly wrong, to challenge an output that looks right but isn’t, to remember what happened the last time a model optimised for the wrong thing. And at the exact moment that capability becomes mission-critical, organisations are systematically removing the people who hold it.

The numbers are not subtle. Across the first quarter of 2026, the tech sector alone cut close to 80,000 roles, and by Nikkei’s analysis roughly half were attributed to AI and workflow automation. Mercer reports that the share of firms planning to reduce junior roles jumped from 17% to 43% in a single year. Deutsche Bank analysts have a name for what is happening underneath the headlines — “AI redundancy washing” — and even OpenAI’s own chief executive has acknowledged that some companies are blaming AI for cuts they would have made regardless.

Strip away the narrative and a pattern emerges: juniors cut at the bottom, experienced staff pushed out near the top, and AI cited as the justification for both. The result is an organisation that is hollowing out the human layer it will most need to run the machines.

The experience crisis is the AI crisis

Deloitte’s State of AI in the Enterprise 2026 - a survey of more than 3,200 leaders - found that only one in five companies has a mature governance model for the autonomous AI agents they are already deploying. Agentic AI is scaling fast; the oversight is not.

That gap can only be closed by people. And the competencies AI oversight actually demands are not the kind you acquire on a two-day course. They are contextual understanding — knowing when an output is plausible but wrong. Organisational memory — recalling what happened the last time a system produced a similar recommendation, and what it cost. Experience-based judgement, for the moments when the model is uncertain or the outputs conflict. Ethical oversight, to catch a system reproducing bias. And composure when something fails critically and a decision has to be made quickly and calmly.

These are the capabilities that grow with years, not quarters. Which is why the most damaging move a company can make is the one many are making by default: assigning the cheapest, least experienced people to monitor the most expensive, highest-stakes systems. Automate the routine work, thin out the experienced workforce, and you do not become more efficient. You become more fragile - in ways that stay invisible until a crisis makes them obvious.

A documented failure, not a hypothetical

This is not a cautionary thought experiment. In 2023, IBM announced that AI would replace up to 7,800 roles, with a strategy weighted toward removing older, more expensive employees in favour of automation and younger talent. Within roughly eighteen months the company had quietly acknowledged that the productivity gains were more selective than promised. Institutional knowledge had walked out the door, and the systems that needed watching lacked monitors who could recognise when they failed subtly.

The board lesson is uncomfortable and precise: when you automate, do the people left monitoring the AI actually have the experiential and contextual competence to recognise a subtle failure — or have you put the cheapest people in charge of the most expensive systems?

The pipeline you are erasing makes tomorrow’s leaders

The damage compounds in two directions at once.

Cutting junior roles removes the bottom of the pipeline that produces the experienced staff and mid-level managers of five years from now. Pushing out senior people removes the institutional memory at the top - and that knowledge is rarely documented, rarely transferable, and gone the moment they leave. Research consistently identifies the loss of institutional knowledge as a primary reason organisations lose momentum during leadership transitions.

The cost shows up in the numbers boards do track. Planned, well-managed leadership handovers retain an estimated 85–95% of revenue through the transition; unplanned, scrambled ones retain only 60–70%. Internal successors reach full productivity in six to nine months; external hires take twelve to eighteen. And the macro figure is sobering: S&P 1500 companies are estimated to lose close to $1 trillion in combined market value every year through inadequate succession planning. Meanwhile 39% of companies have no succession plan at all, and 80% lack confidence in their own leadership pipeline.

These are not HR statistics. They are shareholder-value statistics — and the board is accountable for them.

Why diversity needs a 2.0

For two decades, the diversity conversation in Northern Europe has been overwhelmingly about gender - and that work was right and necessary. But while attention was fixed on gender, a different bias crept in through the back door: age.

The evidence is now hard to ignore. In recent surveys, 90% of employees over 50 report having experienced age discrimination at work; 64% have witnessed or experienced it in the past year; and 22% say they feel actively pressured out of their jobs because of their age. The World Economic Forum puts the figure at 76% of employees globally - with hiring managers favouring younger candidates precisely for AI-related roles, on the lazy assumption that digital fluency tracks with youth.

It doesn’t. The technology gap between older and younger professionals narrowed from 31.1% to 10.7% between 2022 and 2025, and only around 5% of older workers report any significant difficulty with new technology. The “older people can’t handle the tech” story is a cultural reflex, not a finding.

This is what we call Inclusion 2.0 - or Diversity 2.0: extending to age the same strategic seriousness that gender diversity has earned. Not as an ethical nicety, but because the data is unambiguous that mixed-age teams make better decisions, are more innovative, and are more resilient under pressure. A 25-year-old and a 55-year-old approach a problem differently; the younger brings digital fluency and fresh framing, the older brings context, relational insight and organisational memory. The complementary team is not a compromise. It is the strongest combination - and it is exactly the configuration a serious AI strategy requires underneath it.

The governance trap most boards haven’t priced in

There is also direct legal exposure. AI-driven screening tools are now used in around 14% of hiring processes - roughly triple the 2023 figure - and they do not challenge existing bias; they learn it from historical data and repeat it at scale. The EU AI Act classifies hiring and HR tools as high-risk AI systems, which places governance responsibility squarely on the board. Recent cases in both the US and EU have already turned AI hiring tools into the subject of regulatory action and litigation, in some instances holding the vendor liable as an agent of the employer.

So the board cannot treat this as an operational HR detail. The same body deploying AI agents is liable for the AI deciding who gets hired - and for ensuring automation does not quietly destroy the experience base needed to govern any of it.

What the board owns

We put the central question of all this to one of the most experienced chairs in the Nordic region - a former four-star general and chief of defence who now chairs companies across media, banking, cyber security and leadership development. Asked what a chair’s job really is, he reframed an entire military career in one line: his primary role, in every command, was to see far in the distance what would hit the organisation next, and to make sure that at board level they were prepared to navigate it.

That is the capability at stake. The goal is not to hire a new specialist for every risk - one for AI, one for cyber, one for succession. It is to retain and cultivate experienced transformation leaders who can carry all of it into a single decision. Automating the routine while dismissing the experienced is the precise opposite of building that capability.

Succession, and the experience it depends on, is not an HR task. It is a value decision - for shareholders and stakeholders alike. And the board owns it.

Understand the full data picture - and what to do about it.

The LCG Succession & Talent 2026 workbook brings together research from more than thirty global sources - DDI, AARP, McKinsey, Deloitte, Russell Reynolds, SHRM, LinkedIn and the Global Board Survey 2026 - into a board-ready framework: a readiness assessment, an age-bias audit, an AI-oversight competency map, and the questions every board should be asking before the next vacancy, crisis or automation decision makes the gap visible. It is the practical guide to building a Diversity 2.0 strategy that strengthens your AI rollout while retaining and redeploying the skills of your most experienced people.

Download the LCG Succession & Talent 2026 workbook →