Record Revenue. Record Layoffs.
So What’s Actually Happening to Jobs in the Age of AI?
This week in AI and jobs news, we saw record revenues, job cuts, competing narratives, and enough contradictory headlines to make anyone wonder whether AI is creating jobs, destroying jobs, or somehow doing both at the same time.
On one side, companies are posting strong financial results and AI leaders are arguing that artificial intelligence is a powerful engine for growth. On the other, we're seeing layoffs, shrinking graduate programs, and growing concern about what happens to workers caught in the middle of this transition.
So who's right?
The answer is that almost everyone is telling a version of the truth. They're just describing different parts of the same story.
The Week's Scoreboard
Let's start with the headlines.
Cisco announced record revenue of US$15.8 billion, up 12 per cent year-on-year. On the same day, it revealed plans to cut around 4,000 jobs. The explanation? Increased investment in AI across the company.
General Motors announced the loss of 500–600 technology roles as part of what it described as a workforce re-evaluation.
Reports emerged of further layoffs at LinkedIn. Yes, the professional network built around careers and employment is reducing headcount.
Meanwhile in Australia, one of the country's largest law firms, MinterEllison, reduced its graduate intake by almost a third. The reason was refreshingly direct: AI is increasingly handling the routine work that traditionally served as training ground for junior lawyers.
Not every firm is moving in the same direction. Competitor Gilbert + Tobin increased graduate recruitment. But the contrast itself tells a story. Organisations are making different bets on how quickly AI will reshape professional work.
The broader labour market data is equally complicated.
In the United States, roughly 300,000 layoffs were recorded during the first four months of 2026. That sounds alarming until you learn it is approximately 50 per cent lower than the same period last year.
Economists were quick to point out that "better than last year" is not the same thing as "healthy."
The emerging pattern isn't mass unemployment. It's something quieter.
Jobs disappear and simply aren't replaced.
Vacancies close.
Teams shrink through attrition.
Positions remain unfilled.
The headcount reduction often happens without a dramatic announcement.
Jensen Huang Says AI Creates Jobs
This week, NVIDIA CEO Jensen Huang appeared on a podcast and made the case that AI is fundamentally a job creator.
His argument is straightforward.
AI increases productivity.
Productive companies grow faster.
Growing companies hire more people.
According to Huang, AI has contributed to the creation of roughly 500,000 jobs over the past two years.
He also made an important distinction that often gets lost in public debate: automating part of a task is not the same thing as eliminating an entire job.
Most jobs consist of dozens of tasks. AI may remove some of them while making workers more effective at others.
Huang's broader warning was that the biggest threat may not be AI itself, but fear of AI. Workers who avoid learning new tools risk falling behind those who embrace them.
There is truth in that.
But it is not the whole story.
Amy Webb's Warning
Futurist Amy Webb offered a different perspective at the Milken Conference.
She's not arguing that AI is inherently dangerous or that mass unemployment is inevitable.
Her concern is simpler.
We've been discussing automation and labour disruption since the 1950s. For seventy years, we've known that technological progress creates winners and losers.
Yet we have done remarkably little to prepare workers, education systems, or policy settings for what happens during those transitions.
The issue isn't whether AI creates economic gains.
It will.
The issue is whether the people who lose opportunities are the same people who benefit from the new ones.
History suggests that doesn't happen automatically.
The gains and losses are rarely distributed evenly.
The Research Almost Nobody Is Talking About
The most interesting piece of AI jobs research released this week received surprisingly little attention.
The paper is called What Jobs Can AI Learn? and it examines something that most AI exposure studies miss entirely.
Traditional AI risk assessments focus on what AI can do today.
This study focuses on what AI can learn to do through reinforcement learning — the training approach behind many of the recent advances in AI capability.
That distinction matters.
Most previous studies have suggested that writers, analysts, programmers, and other knowledge workers face the highest exposure.
This new framework identifies a different category of risk.
Jobs such as:
Railroad conductors
Power plant operators
Aircraft cargo supervisors
These occupations often score relatively low on traditional AI exposure measures.
Yet they score surprisingly high on reinforcement learning feasibility.
Why?
Because their work contains characteristics that reinforcement learning thrives on:
Clear objectives
Measurable outcomes
Strong feedback loops
Simulatable environments
These are precisely the conditions under which AI systems improve rapidly.
Meanwhile, occupations such as CEOs, musicians, and microbiologists often show the opposite pattern.
AI can assist these professionals, but their work depends heavily on judgment, ambiguity, creativity, trust, and contextual decision-making.
Those qualities remain much harder to automate.
The Mid-Career Squeeze
One of the most fascinating findings in the research is that exposure is not evenly distributed across seniority levels.
Entry-level workers are not necessarily the most vulnerable.
Executives are not necessarily the most vulnerable.
The pressure appears to concentrate in the middle.
Mid-career professionals often perform structured, repeatable processes that organisations value but can increasingly automate.
They are experienced enough to command higher salaries but not always senior enough to be protected by strategic decision-making responsibilities.
The research describes this through reinforcement learning feasibility.
Put more simply, if your primary value comes from executing a well-defined process, you're entering the highest-risk zone.
And there are already early signs appearing in labour market data.
Since 2024, occupations with high reinforcement learning exposure have begun showing relative declines in job postings.
The trend is still emerging.
But it is moving in a consistent direction.
An Ex-Microsoft Executive Tried to Score Careers by AI Resistance
Another piece of analysis making the rounds comes from a former Microsoft executive who built a framework assessing 35 careers according to their resistance to AI disruption.
His motivation was personal.
His sixteen-year-old daughter wanted career advice, so instead of relying on opinions, he attempted to analyse the data.
His findings align surprisingly closely with academic research.
The most AI-resistant careers included:
Mental health therapists (98%)
Firefighters (96%)
Paramedics (96%)
Surgeons (96%)
What these professions share is not technical complexity.
It is human complexity.
They require judgment under uncertainty, trust, ethical decision-making, emotional intelligence, and adaptation to novel environments.
Finance and law showed a much sharper divide.
Senior partners and investment leaders remain relatively protected because their value comes from judgment and relationships.
Junior professionals performing process-driven work appear far more exposed.
His comparison of AI literacy is worth remembering.
Learning AI today is similar to learning typing in the 1970s.
Eventually everyone needed the skill.
At first, however, it created a genuine advantage.
Currently, workers with AI fluency are reportedly earning significantly more than those without it.
That premium may not last forever.
But it exists today.
Why Different Countries Are Seeing Different Outcomes
A recent Morgan Stanley survey suggested that the United Kingdom is currently one of the labour markets experiencing the greatest AI impact.
Not because AI is uniquely disruptive to British workers.
Because British organisations are adopting AI aggressively and prioritising productivity gains.
The productivity improvements are real.
So are the workforce reductions.
Germany, by contrast, has seen more positive employment outcomes.
Different regulatory environments.
Different labour market structures.
Different corporate cultures.
Same technology.
This is a useful reminder that technology itself does not determine outcomes.
Policy choices, organisational decisions, and labour market institutions matter too.
It's also notable that both OpenAI and Anthropic continue expanding their presence in London, citing the depth of local AI talent.
That creates an interesting question for the next 6–12 months.
Will the concentration of AI investment offset the labour disruption that increased automation may create?
The answer remains unclear.
The Part Nobody Wants to Say Clearly
After following this week's news, I keep returning to the same conclusion.
Jensen Huang is right.
AI creates jobs.
Amy Webb is right.
Those benefits won't be distributed automatically or fairly.
The White House is right.
Mass unemployment has not appeared in the data.
The labour market researchers are right.
Many jobs are disappearing through non-backfill rather than headline-grabbing layoffs.
Everyone is describing a different part of the same elephant.
What few people are saying directly is that the group facing the greatest pressure right now appears to be mid-career, mid-wage professionals.
People experienced enough to be expensive.
People valuable enough to matter.
But not yet senior enough to be irreplaceable.
The real dividing line isn't whether your role is "white collar" or "blue collar."
It's whether your value comes primarily from executing a process or exercising judgment.
If your value comes from following a well-defined process, AI is increasingly moving into your territory.
If your value comes from navigating ambiguity, building trust, understanding context, making difficult decisions, and applying judgment where no obvious answer exists, your position remains considerably stronger.
That doesn't make you immune.
But it does place you in a more defensible position.
And that's the practical question every professional should be asking themselves right now:
What part of my work creates value?
Is it the process I execute?
Or the judgment I bring when the process runs out?
Because the answer to that question may tell you far more about your future than any AI headline this week.
