“Why plan, when nobody knows anything?” — the objections that keep workforce planning on the shelf

For years I have been writing here about how AI changes everything we thought was settled in HR: how we think about diversity, and every element of the talent management framework — reward, career management, learning and development. The feedback was kind, and it usually came with a “but”: nice in theory.

So we spent two years turning the theory into tools. Not to win the argument — to make it testable. But before I show you anything, let me finally answer, in one place, the objections I hear every single time strategic workforce planning comes up. You will recognise them. You may have said some of them. I certainly heard all of them last month.

“Why plan at all, when there’s this much uncertainty?”

This is the big one, so let’s take it seriously. The premise is right: nobody knows how fast AI adoption will run, which capabilities land next quarter, or what your regulator will clarify in 2027. The conclusion is wrong, because it assumes planning means predicting.

It doesn’t. The whole point of dealing with uncertainty is to decompose it: break it into named assumptions, put a range on each, and see which decisions survive across the range. Run slow, base and fast adoption scenarios. Make the realization rate a slider, not a hope. When the assumptions are explicit, a CFO’s “why spend on reskilling if the automation might not come?” stops being a matter of conviction and becomes a comparison: here’s the cost if it comes and we weren’t ready; here’s the cost if it doesn’t and we were. Stress-test it together.

Structured uncertainty beats confident guessing. Every time.

And doing nothing is not the neutral option — it just has invisible invoices: payroll running on tasks your existing licences could already do, the 60–120% market premium you’ll pay in panic for AI talent, the external consultants doing what internal reskilling should have built. You pay these every month. They just never appear in one cell.

“What can a model tell us that our experience can’t?”

Your experience is real — and it prices the past. The honest answer is that AI doesn’t change work at the level where experience operates. Leaders have good intuition about jobs. But AI changes tasks inside jobs: in the same role, one activity gets substituted, another accelerated, a third untouched. Two roles that look identical from the org chart can point in opposite directions at task level.

That structure is invisible to intuition and invisible to headcount planning — you need the work decomposed, scored, and traceable. Not because the model is smarter than your people. Because it looks at a resolution where “experience” has literally never had data. (And every score stays overridable by a named human expert — the model proposes, your people decide. That’s a design principle, not a disclaimer.)

“We already spend plenty on reskilling.”

You spend plenty on training. That’s not the same thing, and I’ll say the uncomfortable part out loud: most of the money currently labelled “AI reskilling” — the prompt engineering courses, the blanket AI e-learnings rolled out to everyone — is useful, but it will not make a single person capable of filling a new role.

Don’t take my word for it; take the market’s. If those courses worked, executives would not near-unanimously name skill shortage as the number one barrier to scaling AI. Both things are true at once: record training volumes, and a workforce leaders don’t consider ready. That paradox has a simple explanation — we’re measuring inputs (completions, hours, certificates) and calling it readiness.

Real reskilling starts from the other end: which roles change, into what, which skills bridge the gap, who can realistically make the transition, in what time, at what cost — and whether the target roles can actually absorb them. That last one matters: “we’ll just move everyone into the new roles” is where most transformation plans quietly lie to themselves. Target roles have absorption capacity — growth plus natural turnover — and a plan that ignores it isn’t a plan.

“This isn’t an ROI question — it’s about how people feel.”

I agree more than you’d expect. And that’s exactly why the current approach fails people.

What do employees actually feel right now? Increasingly, a quiet fear of becoming obsolete — of there being no place for them in the new setup. And what do we offer against that fear? A motivational townhall and the sentence “AI won’t take your job.” Which, let’s be honest, nobody in the room believes — including the person saying it.

You know what actually moves engagement? A credible, personal answer: here is how your role changes, here is a realistic path for you, and here is the reskilling behind it — funded and scheduled. Structured plans are not the opposite of caring about people. They are what caring looks like when it’s serious. Hope is not an EX strategy.

“AI is just hype.” / “Not relevant for us yet.”

If it’s hype, explain shadow AI to me. In most organisations, a large share of the workforce already uses AI every single day — privately, on their own accounts, with zero control, zero measurement, and your data in the prompt. The transformation is not waiting for your steering committee. It’s running. The only question is whether it runs governed or ungoverned.

“Not yet relevant” usually means “not yet visible.” Those are different things.

“We can’t deal with this now — pay transparency is eating all our capacity.”

This one I love, because it’s my favourite door. Pay transparency is not a competing priority — it’s the perfect occasion to think all of this through, because it’s the same homework.

I’ve written this many times: the EU directive is not merely a compliance exercise. Transparency is a principle — an ethical statement that goes far beyond the right to equal pay. And in the AI era, I’d argue the whole diversity conversation needs to move from PR programmes to structural questions: How fair and bias-free are the AI tools making decisions about people? Does AI impact women — or older employees — disproportionately in your role structure, and how do you preserve diversity of age, gender and talent through the transition? And can you push the unjustified pay gap toward zero — and prove it?

Equal value, defensible data, documented decisions: whether the trigger is a directive or a transformation, the foundations are identical. Do the homework once.

What responsible actually looks like

Strip away the objections and what remains is, I think, HR’s three biggest responsibilities in the AI era — each with a KPI that is not theatre:

Responsible about the business. Training money goes where the business case is, not where the loudest request is. The KPI is not e-learning completion; it’s AI readiness measured against the skill set of the future operating model — role by role.

Responsible about people. Not slogans — plans. The KPI: how many people sitting in the most-affected roles have a realistic, funded career path with reskilling behind it? (And its dark twin: how many people are you laying off while simultaneously hiring the same capacity from the market at a premium? That number is the cost of not planning.) Plus the oldest fairness KPI of all: equal pay for work of equal value — actually measured.

Responsible about the technology. No rubber-stamping. If a human “oversees” your CV screening tool but has no time, no training and no real authority to override it, that’s not oversight — that’s liability with extra steps. The KPI: bias tested, decisions explainable, oversight real.

Why I stopped just writing about it

For years, this was theory with my name on it. So we built it: gAIn for strategic workforce planning with named assumptions and scenario logic; FairPath for AI governance that turns principles into controls; EAS and PMS on the reward side, so equal pay for equal work is a calculation, not a slogan. Not because tools replace judgement — because tools end the argument about the numbers, so the argument can finally be about the decisions.

Scenario-based workforce planning with numbers you can defend.

Explore gAIn →Book a call →