ADP’s Maria Black and the 3.5% Growth Trap for AI-Obsessed CEOs
If you use AI to eliminate every junior job, don’t pretend you’re running a lean company. You’re selling off your future managers for a one-quarter margin bump.
If you use AI to eliminate every junior job, don’t pretend you’re running a lean company. You’re selling off your future managers for a one-quarter margin bump.
The clever cost cut that leaves you with nobody capable
The corporate world has found a new way to make an old mistake sound intelligent: call it AI transformation, cut the lower rungs out of the business, then marvel when nobody inside the company knows how the work actually gets done.
That is the real management story sitting underneath this week’s AI chatter. Not whether a chatbot can draft a slide deck. Not whether your board has learned to say “agentic workflow” without wincing. The question is whether leaders are quietly destroying the development pipeline that produces their next operators, sales leaders, product heads and executives.
Fortune reported on September 8 that US employment is projected to grow just 3.5% from 2025 to 2035, down sharply from 10.9% in the prior decade. At the same time, executives are talking about skills shortages, poor engagement, AI uncertainty and cost pressure. That combination makes junior talent more valuable, not less. ([fortune.com](https://fortune.com/2026/09/08/jobs-report-hiring-ceos-ai-employee-engagement-leadership-pipeline/?utm_source=openai))
ADP chief executive Maria Black put the central issue plainly: AI should act as a teammate that increases the value of judgement and leadership skills. The catch is obvious. If companies automate or eliminate the jobs where people learn judgement, how exactly do they expect to manufacture experienced leaders five years from now?
They don’t. Most are hoping somebody else trains them.
That is not a strategy. It is freeloading with a PowerPoint deck.
Voya’s Heather Lavallee sees the problem before most boards do
Heather Lavallee, chief executive of Voya Financial, has been asking the uncomfortable question that ought to be on every board agenda: if automation takes too much of the entry-level work, where do future experts come from?
That is not nostalgia for graduate programs, mailrooms or making young staff suffer through pointless admin. Plenty of low-value work deserves to disappear. Nobody should preserve a rubbish process just because someone once had to do it manually.
But early-career work has always done two jobs at once. It produces output today, and it creates the pattern recognition that makes somebody useful tomorrow. A junior analyst learns by cleaning a bad spreadsheet, asking why a customer churned, sitting in a sales call that goes sideways, or watching a competent manager make a difficult call. You cannot outsource all of that to software and expect wisdom to appear by magic at age 32.
That matters even more in a slower-growth labour market. When overall employment growth is forecast to be only 3.5% over a decade, businesses cannot casually assume a deep external pool of trained replacements will be waiting whenever they need them. ([fortune.com](https://fortune.com/2026/09/08/jobs-report-hiring-ceos-ai-employee-engagement-leadership-pipeline/?utm_source=openai))
The firms that keep investing in capability will have options. The firms that treat people purely as a cost line will have a nice quarter, followed by a desperate hunt for senior talent at inflated prices.
I have seen this movie in business. The bloke who says, “We’ll just hire someone later,” is usually the same bloke ringing recruiters six months later, shocked that good people cost money.
AI should remove drudgery, not remove the apprenticeship
The sensible version of AI adoption is brutally practical: use it to remove repetitive work, shorten the time from question to answer, improve preparation, and let capable people take on more meaningful responsibility earlier.
The stupid version is using AI as a euphemism for “we no longer want to train anyone.”
Axios’s Jim VandeHei made a useful management point this week: leaders need to understand the technology properly, communicate repeatedly about how it will be used, and enable adoption across the organisation rather than confining it to the tech team. He also argues that companies need clear AI governance before opening the gates. ([axios.com](https://www.axios.com/2026/09/08/ai-ceo-tips-hiring-communication-org-chart))
Fair enough. But I would add a fourth requirement: every AI deployment needs an apprenticeship test.
Before you automate a junior role, ask four questions:
1. What skill did this role previously teach? 2. Where will a new employee now learn that skill? 3. Who owns the teaching, specifically? 4. What measurable output will prove the replacement training works?
If your answer is “they can ask the AI,” you have not built a training system. You have handed a learner an answer machine and hoped competence comes bundled with the subscription.
It doesn’t.
The overlooked opportunity: give juniors bigger work sooner
Here is the contrarian bit. AI does not have to flatten the leadership ladder. Used well, it can make the ladder climb faster.
A sharp 24-year-old who can use AI to research a market, prepare a customer brief, inspect a process, draft options and test a basic prototype should not be left doing the same junior work we assigned graduates ten years ago. That person should be given a tighter feedback loop and more real responsibility.
This is where too many established companies will lose to smaller, hungrier businesses. Start-ups will give young operators genuine ownership because they have no choice. Big companies will often use AI to centralise more control, add another approval layer and call it productivity.
Sam Altman made the bullish version of this argument in an Axios interview published September 7. He said AI has reset the economics of starting a business, citing the example of software sold to 50 customers for roughly $500 a month—an opportunity that would once have required a far larger engineering budget. Axios also reported that more than 3 million new US businesses were formed in the first half of 2026, the highest January-to-June total in Census Bureau data going back to 2005. ([axios.com](https://www.axios.com/2026/09/07/axios-interview-altman-touts-revenge-of-ideas-guy))
Now, before everyone starts applauding the “idea guy,” remember this: an idea has never been the hard part. Customers, distribution, judgement, taste, resilience and execution are still hard. AI has lowered the cost of producing a first draft. It has not lowered the cost of being right.
But it does mean ambitious junior people can create evidence faster. A good manager should exploit that. Give them a small customer problem, a revenue target, a decision boundary and access to an experienced operator. Let AI accelerate the grunt work. Make the human do the thinking.
That is how you turn AI from a redundancy machine into an operator factory.
BlackRock and Meta are pointing in a better direction
There are signs that some large organisations understand this is a training problem, not merely a headcount problem. Fortune reported that BlackRock is investing $100 million in skilled-trades training programs and has partnered with Ford, Carhartt and Alphabet through the Alliance for America’s Skilled Trades. Meta has partnered with CBRE and others on a five-week program that guarantees participants a job upon completion. ([fortune.com](https://fortune.com/2026/09/08/jobs-report-hiring-ceos-ai-employee-engagement-leadership-pipeline/?utm_source=openai))
Good. More of that.
The exact programs are not the point. The point is that serious leaders are accepting responsibility for building the workforce they will need. They are not waiting for universities, governments or somebody else’s P&L to do it for them.
This matters because the jobs AI disrupts and the jobs business needs will not line up neatly. A company can automate reporting while desperately needing people who can interpret a customer, manage a supplier, negotiate a deal, lead a team through change or make a call with incomplete information. Those are not side skills. They are the business.
The real competitive advantage is not “we bought AI licences.” Everyone can buy licences. The advantage is building a company where people learn faster, decisions get better and good operators are produced internally at a reliable rate.
What this means for you
If you run a business, do this tomorrow.
First, list every junior role you plan to automate, shrink or stop hiring for over the next 12 months. Be honest. Then write down the capabilities each role used to develop.
Second, assign an owner for every capability. Not HR in the abstract. A named executive. If junior analysts will no longer learn commercial judgement through manual reporting, perhaps the head of finance now owns a monthly decision-review session where they see how numbers become action.
Third, make AI fluency part of the job, not a perk for the curious. Give teams approved tools, clear guardrails and practical use cases. VandeHei is right that leaders need to communicate constantly and establish governance before companywide use becomes a free-for-all. ([axios.com](https://www.axios.com/2026/09/08/ai-ceo-tips-hiring-communication-org-chart))
Fourth, promote responsibility earlier. Do not make talented young people spend three years polishing decks that software can produce in 30 seconds. Put them closer to customers, revenue, operating decisions and experienced managers who can tell them when they are wrong.
And finally, stop congratulating yourself for cutting a cost if you have also cut the company’s ability to grow its own leaders.
AI will absolutely change the org chart. But the businesses that win will not be the ones with the fewest humans. They will be the ones that use the machines to make their humans more capable, more accountable and harder to replace.