Fate of handloom weavers shows what AI is going to do to middle class jobs: David Richards

It is also exactly what the handloom weavers said two centuries ago. In the late 18th century, handloom weaving was a good job. A skilled weaver could earn a respectable, sometimes enviable income. Many worked from home, owned their looms, controlled their hours and trained apprentices.

In parts of Yorkshire and Lancashire, weaving supported a solid, independent working and lower middle class. It was not drudgery. It was skilled, semi-autonomous labour with status.

Then came mechanisation.

When the first power looms appeared, they were inferior. Output was inconsistent. Threads snapped. The finish lacked the touch of a trained artisan. A master weaver could easily demonstrate why the machine was not as good.

But the machine did not need to be better. It needed to be cheaper and capable of improvement.

From around 1790 to the 1820s, as mechanisation scaled, the economics turned brutal.

Estimates vary by region, but real earnings for handloom weavers fell by as much as half, and in some areas by more than two-thirds over a generation. Men who had once supported families respectably found themselves working longer hours for subsistence pay. By the 1830s, handloom weaving had shifted from a decent trade to a poverty occupation.

The Luddites, who began smashing frames and looms in 1811 under the banner of the mythical “General Ludd”, were not anti-progress caricatures. They were skilled workers watching a good profession being hollowed out. The state responded with troops. Frame breaking was made a capital offence. Seventeen were executed in York in 1813. The protest was crushed because the economics were overwhelming.

The key point is this. The job did not vanish overnight. It was squeezed. Wages collapsed first. Status followed. The craft was reorganised around machinery, with far fewer people required.

That is the pattern now emerging in parts of the professional middle class.

Software development is simply the most visible example. The frontier work remains deeply human. Architecture, system design, product judgement and taste will still matter enormously. But much day-to-day development is structured and repeatable. Standard integrations, boilerplate features, regression testing, documentation and deployment scripts.

These are precisely the domains where AI improves fastest.

And this is not confined to engineers.

Paralegals reviewing contracts line by line, junior lawyers drafting standard agreements, compliance analysts mapping regulation to checklists, financial analysts building repeatable spreadsheet models, recruiters screening CVs, market researchers compiling public data and tier-one support teams handling predictable queries. These are good jobs today. They are respected. They are often well paid. They look secure.

So did handloom weaving in 1790.

AI does not need to eliminate these roles to change them. It only needs to enable one person to do the work of five or ten. That is how a profession shrinks without a dramatic headline.

The entry layer thins. The ladder narrows. Wages come under pressure.

At Yorkshire AI Labs, this is not theoretical. It shapes how we build. Three years ago, when planning a new software startup, we would assume multiple squads.

Core development was a squad of six engineers. UI another six. QA another six. Release and DevOps another six. It was entirely normal for an ambitious early-stage company to expect to hire 20 or more engineers before it had meaningful traction.

Today we model on one highly capable, AI native engineer.

One.

Each of those former functions was effectively a six-person squad. Now much of that work can be orchestrated, generated and validated by a single individual working with AI systems.

A company still planning for 20 is building for a cost base that no longer reflects reality. A CTO in one of our portfolio companies told me last week he needed to replicate a bank’s environment to build and test an integration. Historically, that would have required weeks of work.

Stubbing APIs, simulating services, building harnesses. Using AI, he generated a functioning mock banking environment in minutes.

That is the loom moment. The handloom weavers were right about the flaws in the first machines. They were catastrophically wrong about the direction of travel.

AI does not need to be perfect to change the labour market. It only needs to change the economics. The loom was not ready either. It did not need to be.

David Richards MBE is managing partner of Yorkshire AI Labs