AI and work: which tasks really change (and why England's 1351 labor law didn't work)

"Will AI take our jobs?" It's the question I hear most often, at companies and at dinner parties. The honest answer is duller and more useful than yes or no: what will change most are tasks, the individual pieces of work inside each job. And anyone who tries to stop that change by decree, history teaches us, usually loses.

1348: the law that tried to stop time

In 1348 the Black Death reached England, as it did much of Europe, and killed an enormous share of the population. Estimates vary widely from place to place, but the effect on work was immediate: there weren't enough hands in the fields and workshops, and survivors asked for higher wages or moved to wherever paid better.

The Crown's response was an ordinance of 1349, issued in a year when, as the UK National Archives note, Parliament had been unable to meet precisely because of the epidemic. In 1351 came the Statute of Labourers, which supplemented it. Its aims, again according to the National Archives: to combat rising wages by capping what an employer could pay, to control the movement of laborers, to mandate employment for all able-bodied workers and to prevent the poaching of servants with better offers.

In essence, the law tried to put the organization of work back the way it was before the plague. Historians still debate how strictly it was enforced, but many agree that in the long run it couldn't stop wages from rising or labor relations from changing. The world had changed, and no statute could put it back.

We're in a reversed but similar situation today. It's not hands that are missing: a tool has arrived that can do in seconds some tasks that used to take hours. And the temptation is the same: pretend nothing changes, or cry catastrophe. Both reactions prevent us from doing the right thing, which is redesigning tasks with judgment.

What the data says: exposure doesn't mean replacement

The most solid study available was published in May 2025 by the International Labour Organization together with Poland's NASK institute: Generative AI and Jobs: A Refined Global Index of Occupational Exposure. The key points:

• roughly one worker in four worldwide is in an occupation with some degree of exposure to generative AI;
• only 3.3% of global employment falls into the highest exposure category, where most tasks could be done by AI;
• clerical and administrative occupations are the most exposed;
• women are more exposed than men: 4.7% of women work in the most exposed category versus 2.4% of men, and in high-income countries the gap widens to 9.6% versus 3.5%;
• the authors' conclusion: the most likely effect is transformation of jobs, not replacement, because few jobs consist only of fully automatable tasks.

In Italy, one more figure completes the picture. According to Istat, in 2025 16.4% of companies with at least 10 employees used AI, and among those that didn't, 58.6% cited lack of skills. For now, the problem isn't that AI replaces people: it's that people aren't equipped to use it.

Which tasks change, area by area

An indicative map for a small business, thinking in terms of individual tasks rather than job titles.

Notice the pattern: preparation and copying tasks change, while judgment, relationship and responsibility tasks stay central. That's no accident: it's exactly the line between what a language model does well and what it can't do. For sector details, see AI for logistics and warehouses and AI document management.

The real risk isn't the one you think

That said, it would be naive to claim everything is fine. The risks are real, but they're different from the headlines.

• Workload that grows instead of shrinking. If AI drafts in half the time, the temptation is to ask for twice as many drafts. Freed-up time has to be allocated on purpose, not left to chance.
• The apprenticeship that disappears. Many professionals learned their trade by doing exactly the repetitive tasks AI now does for them. If juniors no longer do them, you need another way to help them grow.
• Skills that wither. People who only check what the machine produces, without ever doing it themselves, gradually check less well.
• Opaque control. Systems that assign tasks or evaluate performance without workers knowing. Here the law has already stepped in.

What the rules require

Two references to know if you employ people in Italy or the EU:

• the EU AI Act (Regulation (EU) 2024/1689), Article 4, has required since February 2, 2025 that organizations using AI systems ensure a sufficient level of AI literacy among their staff;
• Italy's AI law (Law 132/2025, in force since October 10, 2025), Article 11, states that AI use at work must respect workers' dignity and rights, and that employers must inform workers about AI systems in the cases set out by the rules on transparent working conditions.

The message is organizational before it's legal: people must be informed and trained. Do it early and you get cooperation; do it because you must and you get suspicion.

How to lead the change in a small business

Here's where 1351 England teaches us something: you don't manage change by trying to freeze it. You manage it by redesigning work together with the people who do it.

1. Break roles down into tasks. For each role, list a typical week's tasks and how long they take. It's the same mapping exercise we propose in our guide to business process automation.
2. Classify each task. Repetitive and checkable (an AI candidate), judgment-based (stays human), mixed (AI prepares, a person decides).
3. Decide where freed-up time goes. More time with customers? More quality control? Training? Write it down, or it'll be filled with more of the same work.
4. Train before you introduce. Not a generic course: exercises on each person's real tasks. More in AI training for employees.
5. Set clear usage rules. What can be done with AI, with which data, who checks what. A one-page policy beats a rulebook nobody reads: see ChatGPT at work: rules for employees.

And when more autonomous tools arrive, such as AI agents, the principle stays the same: the machine prepares, the person decides and is accountable.

Talk it through with people, not just vendors

Penzai's Human Tech approach comes from exactly this: understand people's work first, then choose the technology. We also explain it in why we start from people. If you run an SMB and want to bring in AI without losing your team's trust, book a free intro call.