Issue 003: The AI numbers nobody's checking

Six months ago, most conversations about AI and headcount were speculative. Now the numbers are in, and they're messier than either side of the debate wants to admit. Some of the companies citing AI as the reason for cutting people can't actually prove AI improved anything. Meanwhile, well over half your workforce is already using AI tools you've never approved, whether you've noticed yet or not.

This week is about the mechanics underneath the headlines: how the organisations handling this well are actually structuring redeployment, transparency, and reskilling, not just announcing it. And it's about the questions people leaders should be asking before they repeat a productivity story that might not hold up.

Alisa Latty-Alleyne is back for a second week running too, and there's a reason for that.

Welcome to Issue 003 of The Work Life Reporter. This week you get:

Let’s get into it.

When Meta cut roughly 8,000 roles this year, it also opened close to 7,000 newly created AI-focused positions and ran them as an internal-first pipeline before recruiting externally. Employees whose roles disappeared got priority interview slots and a defined window to apply before anyone outside the company saw the postings.

The credibility of a move like this rests entirely on whether people actually get hired into it, not just invited to apply as a courtesy. What made this a real culture signal rather than a redundancy dressed up nicely is that Meta put both numbers next to each other in public: the cut and the created. That let employees do the maths themselves instead of taking leadership's word for it. If you're going to pair a layoff with a redeployment story, the numbers have to be checkable, or nobody believes the second half of the sentence.

When Atlassian announced a 10% cut to its global workforce for what its CEO called the "AI era," he didn't leave the criteria vague. Mike Cannon-Brookes said plainly that employees with transferable skills were the ones this round of change was built around protecting.

Naming the criterion out loud is the actual move here, not the criterion itself. Skills mattering in a restructuring isn't news. What's rarer is a leader saying so publicly instead of letting it stay an unspoken logic that only surfaces in someone's exit conversation, if it surfaces at all. That single act turns an anxiety-inducing black box, how do they actually decide, into something a person can reason about, even if they don't like the answer. It only holds up if managers can then explain the specific skills call to each individual afterwards, not just repeat the company line back at them.

Accenture is cutting around 11,000 roles as part of a restructuring its CEO Julie Sweet tied directly to how work is changing inside the firm. Her line to staff was blunt: those who cannot be reskilled will be exited.

Quoted alone, that sentence sounds harsh. What doesn't make the headline is the structure that has to sit behind it for the sentence to mean anything other than a threat. A runway with a real length. Reskilling tracks tied to specific open roles inside the business, not generic e-learning modules nobody finishes. A calendared checkpoint where the call actually gets made. The debate worth having internally isn't whether the line was too direct, it's whether the runway behind it was long enough and resourced well enough to be a genuine shot, rather than cover for a decision that had already been made.

IKEA has spent years being held up as a case study in handling technological disruption responsibly, built on a reputation for retraining people through earlier waves of digital and store-format change. New layoffs in 2026 forced a re-examination of that story, and whether the original promise still holds.

The lesson here is an uncomfortable one for any organisation that's earned a good reputation once. That reputation doesn't carry forward automatically. Each new wave of change reopens the question of whether the promise still holds, and the employees living through the new round don't get to benchmark it against the old case study. They experience it fresh. If leadership doesn't proactively address whether this time is the same or different, out loud, before people have to guess, the credibility built last time doesn't transfer.

Recent survey data puts the scale of the problem plainly: 59% of employees use AI tools their employer never approved, and 75% of those admit sharing sensitive information, employee data, customer details, internal documents, with those tools. Some organisations have responded not with a blanket ban but with a fast swap: replace the unsanctioned tool with an approved, secure equivalent before drawing the line.

The instinct in most organisations is to treat this purely as a security problem, and it partly is, but the culture read underneath it is different. Shadow AI use isn't rebellion. It's a signal that whatever tool people were officially given wasn't good enough for the job they actually needed to do, so they found one that was. Punishing the workaround without fixing the reason for it just pushes the behaviour further underground. The organisations getting ahead of this aren't running crackdowns first. They're offering a no-blame disclosure window to find out what people are really using, then shipping a sanctioned alternative before they ban anything.

Gallup's latest workplace data found something that should worry anyone running an AI rollout right now: employees whose managers actively support AI use are close to nine times more likely to say the technology has actually changed how they work day to day. The tool, the training budget, the executive mandate, all of it matters far less than whether someone's direct manager is visibly using it and backing it.

Here's the problem. Fewer than one in three employees say their manager is doing that. And manager engagement itself has been sliding for three years, dropping faster than employee engagement generally, which means the layer you most need showing up for AI adoption is the layer that's checked out the hardest.

Most AI rollout plans budget for tools, training, and comms. Almost none of them budget for manager coaching specifically about AI, as its own line item. If your adoption numbers are flat despite the investment, stop looking at the tool. Go find out whether managers have actually talked to their teams about AI at all, and whether they believe in it themselves.

A lot of restructuring announcements this year have leaned on the same unspoken logic: AI is making the business more productive, so fewer people are needed to do the same work. It's worth checking whether that's actually true before a headcount decision gets built on it.

A widely cited MIT study found that 95% of organisations investing in AI saw no measurable impact on profit. Separate labour research puts the figure at 89% seeing no effect on productivity at all. And in Gallup's own numbers, only around one in eight employees at AI-adopting companies strongly agree that AI has genuinely changed how work gets done.

None of that means AI isn't useful. It means the productivity story used to justify some of this year's biggest workforce decisions is doing more work than the data can support. Ask finance or ops for the actual measured evidence first, not the projection, not the vendor case study.

Global survey data this year shows 40% of workers now fear AI will make their job obsolete, up from 28% just two years ago. But the more interesting number sits underneath that one: 63% expect AI to make the workplace feel less human, and that loss of humanity, not redundancy, is the concern people rank highest when you actually ask them properly.

There's also a perception gap worth sitting with. Employees are roughly twice as likely as HR leaders to name job loss as their number one AI fear. HR leaders, generally more confident and less anxious about the technology than the people they support, are underestimating both the scale and the shape of the concern.

The fix isn't a different reassurance script. It's asking the actual question, properly, before writing the message that's supposed to land.

Rather than reheat last week's spotlight while you wait for the live conversation, we went back to Alisa with five new questions. All still on AI, none of them repeats.

If you missed last week's issue: Alisa is the founder of HebrewsEleven, an ethical AI advisory consultancy, and H11 Labs, its social impact arm focused on AI equity for underrepresented communities. She's currently completing an Executive MBA in AI and Digital Transformation at Warwick Business School, where she was awarded the Inspiring Female Leaders Scholarship.

When Alisa sees a leadership team draft a strict AI-use ban as their first move, she pushes back before it goes out. People go around a policy when the sanctioned option can't do the job they actually need done, and a ban without a working alternative just moves the risk somewhere leadership can't see it anymore. Her advice to every board considering a crackdown: find out what people are actually using, and why, first.

In her board advisory work, Alisa is seeing boards spend hours on governance frameworks and almost no time asking whether the people managing teams day to day are actually equipped for, or believe in, what's being asked of them. Her reframe: readiness isn't an organisational abstraction. It lives entirely in whether a manager can and will have the AI conversation with their own team.

In her advisory conversations, Alisa is increasingly hearing leadership teams reach for an AI productivity narrative to explain headcount decisions that were, on closer inspection, made for other reasons entirely: budget, board pressure, a competitor's move. Her position is blunt. If you can't show your working on the productivity claim, don't use it as your reason. Use the real one, however uncomfortable.

Since we spoke to Alisa last week, H11 Labs has opened its first community cohort, working directly with young people and faith-led organisations on practical AI skills, not policy documents. Her point for people leaders: most organisations have the same uneven-adoption problem internally that Alisa is addressing externally. If you're not running the equivalent inside your own walls, a defined programme, not a webinar, your inequity risk is just as real. You've just not measured it yet.

Ahead of Tuesday's live conversation, Alisa's one request of anyone joining is simple. Bring an actual, specific situation from your own organisation, not a general question about AI. She finds the best hour of conversation happens when someone says "here's what actually happened at my company," not "what do you think about AI in general." If you want a genuine answer applied to your own situation, that's the way in.

Thank you, Alisa, again!

Connect with Alisa on LinkedIn or find out more at hebrewseleven.com

And that's a wrap for Issue 003. Something land? Hit reply and tell us which section you're taking back to your team.

Follow The Work Life Reporter on LinkedIn for what we share between issues, and join us live on Tuesday.

PS. Know someone who leads people and would find this useful? Forward it on. Ten seconds to share, a lot longer to write.

PPS. Got a story, a leader, or a topic you'd like us to dig into? Reply with your ideas.