Key takeaways
Over a third of Claude users surveyed by Anthropic expect AI to be able to do most or nearly all of their work tasks within the next year (Anthropic Economic Index, June 2026). Reskilling a workforce takes far longer than that. Most HR leaders we work with feel this gap every quarter, because the technology roadmap changes faster than a workforce plan can be approved, let alone delivered.
The question more boards are now putting to their CHRO is a simple one: how many people, with which skills, will we need in three years? Nobody can answer it with confidence, and pretending otherwise produces plans that age badly. Accepting that uncertainty is uncomfortable, and it is also the starting point for workforce planning that works.
The Three Clocks is the pattern that makes workforce transformation so hard in the AI era. Technology, regulation and people change at three different speeds, and a company can't wait for the fastest clock to settle before it acts.
The scale of change on clock three is significant. The World Economic Forum expects 170 million roles to be created and 92 million displaced by 2030, a disruption equal to 22% of today's jobs (WEF Future of Jobs Report 2025). In the same report, if the global workforce were 100 people, 59 would need reskilling or upskilling by 2030, and 11 of them would be unlikely to receive it. Employers already feel the pressure, with 63% naming the skills gap as the key barrier to their transformation (WEF, 2025).
As a result, the clock that matters most for a CHRO is the slowest one. People and skills take longest to move, so decisions about them have to be made well before the technology picture is clear. Workforce planning capability is the organisational ability to make those decisions well, and to revise them continuously as all three clocks move.
Classic strategic workforce planning rests on two assumptions: there is one forecast to plan against, and the plan can be written in job titles and headcount. AI undermines both, in four ways.
Some of these data points come from Anthropic, an AI developer. Independent institutions such as the WEF reach the same conclusion, and everything that follows works whichever AI tools a company runs, because it maps tasks rather than products. Taken together, the four shifts mean the classic plan (one forecast, written in titles, refreshed once a year) fails on both of its core assumptions.
So how do you plan a workforce when the future won't hold still?
Our answer is to build the capability to re-plan and to treat any single plan as a snapshot. In our work with partners across the German-speaking region, that capability rests on four practices that build on each other and keep running as a loop: Understand, Map, Prove and Scale.
Workforce decisions are only as good as the leadership team's shared view of what AI can already do. When the CHRO, the business-unit heads and the CFO each carry a different picture of the future in their heads, every workforce conversation turns into a debate about assumptions.
The Understand practice closes that gap. In a Workforce Safari, senior leaders, HR teams and works council representatives visit companies that are further ahead, work hands-on with AI, hear from experts and play through a scenario game built on their own workforce. The game asks one question repeatedly: does this decision still make sense if AI moves slowly, moderately or very fast? The moves that pass in every scenario are the no-regret moves. Typical examples include AI literacy for every employee (which the EU AI Act has required under Article 4 since February 2025), building a network of internal AI champions, and redesigning one high-volume workflow end to end.
The Safari is a starting point, and it becomes a habit through regular signal reviews that keep senior leaders close to what is changing.
Try this: At your next leadership meeting, put the three Anthropic scenarios on one page and ask each leader which one your current workforce plan assumes. Where the answers differ, you have found your first real conversation. Then list the moves that make sense in all three, and start one of them this quarter.
The role and task map is the instrument most CHROs need and few have. It turns an abstract debate about "AI exposure" into a concrete picture of which work changes, how, and what people need to learn.
Take a role most HR leaders know well, the HR business partner. Drafting job descriptions and answering routine policy questions are already largely automatable. Workforce analytics is augmented, because AI prepares the analysis and the partner interprets it. Coaching a manager through a difficult team conflict stays human. A new task also appears: owning the HR assistant workflow, checking its answers and improving it over time. Seen task by task, the role shifts toward judgement and coaching, and the skills gap becomes specific enough to close.
In a From-To Role and Task Atlas, we break 5–15 critical roles into tasks, tag each one as human, augmented, automated or new, and write the "to" version of the role along with the skills delta and a bridge path. The result is a set of living role cards with scenario tags, refreshed every quarter, plus a list of emerging roles such as AI workflow owner or agent operations.
A fair objection is that 15 roles won't cover 2,000 job profiles. The answer is to go deep where the headcount sits, use AI-assisted mapping for the long tail, and keep the result in your own HR systems so the organisation owns and maintains it.
Try this: Pick the three roles with the largest headcount in your organisation. With two role owners each, list ten core tasks per role and tag every task as human, augmented, automated or new. The share of tasks in each category is your first honest view of exposure.
Boards and works councils move on evidence. A redesigned unit with measured results answers questions that no strategy paper can, including how people actually respond when their work changes.
In a Guided Experiment, one unit spends 12–16 weeks redesigning its workflows, roles and team structures with AI. The best candidates are areas where AI is already in use, such as software development, customer service or finance operations, because the tools are there and the remaining constraint is how the work is organised. We measure four things throughout: productivity, quality, wellbeing and skills.
The human measures matter because they move first. Confidence in using AI, a sense of autonomy and the share of teams with an active AI champion are leading indicators of the financial result. When people feel capable and in control, they take ownership of the redesigned workflows, and that ownership is what keeps the productivity gain in place once the experiment ends.
The economics are easy to frame. As an illustration, a unit of 50 people at an assumed fully loaded cost of EUR 80,000 per head costs EUR 4 million a year, so a productivity gain of 5–10% is worth EUR 200,000–400,000 a year. Even at the lower end, that covers the cost of a well-run experiment within its first year. The unit also leaves with a playbook, a group of internal champions and a draft works agreement that the next unit can build on.
Try this: Choose one unit where AI tools are already deployed. Before anything changes, agree four baseline measures with its leader: one for productivity, one for quality, one for wellbeing and one for skills. Without a baseline, any result you report later will be treated as opinion.
In Germany and much of Europe, co-determination sets the pace of AI adoption. Under Article 26 of the EU AI Act, employers must inform worker representatives and affected workers before a high-risk AI system is put into use at the workplace. In Germany, works councils hold information, consultation and co-determination rights over the introduction of AI under the Works Constitution Act (BetrVG §87, §90 and §111). In Switzerland, employee representation and social partnership play a similar role in how change is introduced.
Organisations that treat the works council as a final approval step tend to add months to their timeline and lose trust along the way. Organisations that co-design the principles at the start usually find the opposite. When employee representatives help shape how AI is introduced, the legal requirement becomes a source of trust and speed. A Social Partner Lab gives management and the works council a neutral format to agree those principles together, before the first tool decision is made.
Said simply: the slowest clock speeds up when the people it governs help set it.
Try this: Invite your works council chair to the first scenario session, before any tool or vendor decision. Agree three shared principles for introducing AI, for example transparency about which tasks change, a reskilling commitment for affected roles, and a fixed review cadence.
The fourth practice, Scale, is what turns these steps into a lasting capability. A small Workforce Transformation Office runs quarterly signal reviews, refreshes the role and task map, tracks redeployment and coaches leaders through the decisions that follow. Each completed experiment raises new questions that feed back into Understand, so the loop keeps running as AI moves.
The headcount question deserves a direct answer. Some roles will shrink, and leaders should say so openly. For most European companies, though, the larger risk over the next decade is having too few people as experienced employees retire. AI is how a company keeps its output while its workforce ages, provided it protects the junior pipeline and builds bridges from exposed roles into growing ones.
The choice is simple: keep rewriting a static workforce plan every year, or build the capability to re-plan every quarter. Every year spent on the first option leaves the slowest clock, your people's skills, further behind.
In 90 minutes, we work with 4 to 10 of your HR and business-unit leaders (and, ideally, a works council representative) in a live-facilitated session. You leave with a clear view of which clock is limiting your organisation, a first task-level picture of three high-headcount roles, and a shortlist of no-regret moves you can start this quarter. For teams that want to go further, the natural next step is a four-week Workforce Scenario Stress Test for one business unit.
