Your job isn't being automated — but parts of it are. Here's how to think about which parts.

There's a framing problem in most "AI and jobs" discourse: we talk about job titles when we should be talking about task composition.

A new piece from Patterson Consulting makes this concrete using what I think is the most useful version of the industrial revolution analogy I've seen: the mechanical loom.

The power loom didn't eliminate weavers. It automated established-pattern work (thread production at scale) and shifted weavers toward design, quality control, and coordination — work that requires judgment you can't fully specify in advance. Employment in the textile industry grew after mechanization.

The same frame applies to knowledge work and AI. The key distinction isn't easy vs. hard:

• AI-tractable work: tasks where "the pattern is established before you begin" — variance reports, status summaries, first-pass code reviews, boilerplate generation
• Judgment-intensive work: tasks that produce the pattern — architectural decisions, debugging novel failure modes, understanding what the customer actually needs vs. what they said they needed

This maps cleanly to Simon's programmed/unprogrammed decision framework from 1960, and to Polanyi's Paradox: skilled engineers know more than they can articulate, and that tacit knowledge is a structural barrier to full automation.

Worth noting: engineering job openings are currently at multi-year highs (67K+ globally), with demand accelerating through early 2026 even as AI coding tools are everywhere. The loom didn't kill the weaver. It changed what weaving means.

Read it here: https://pattersonconsultingtn.com/content/hitchhikers_guide_kw/mechanical_loom_mental_synthesis.html