The Myth vs. the Data
The industry line on AI and jobs has settled into a slogan: “You won’t lose your job to AI. You’ll lose it to someone who uses AI.” Satya Nadella said it, and it’s become the default reassurance whenever job-loss fears come up. It’s a fair correction to the crudest version of the fear: AI isn’t a one-for-one replacement machine, quietly swapping out roles overnight. But taken on its own, it undersells what’s actually showing up in the data.
Goldman Sachs estimates AI is eliminating roughly 16,000 net US jobs a month in 2026, about 192,000 over the past year, with 25,000 positions eliminated through direct substitution, partly offset by 9,000 created through augmentation. Stanford’s 2026 AI Index found the effect isn’t evenly spread by age: employment for software developers aged 22–25 has fallen nearly 20% since 2024, and workers 22–30 broadly face displacement at close to three times the rate of workers 40–55.
Who’s Actually at Risk
Goldman Sachs’ own research names which occupations fall on which side of that line:
| Higher Substitution Risk | Higher Augmentation Potential |
| Telephone operators, insurance claims clerks, bill collectors, billing clerks, telemarketers, payroll clerks, legal assistants, procurement clerks, proofreaders, word processors | Industrial engineers, operations managers, lawyers, physicians & surgeons, construction managers, education administrators |
The pattern holds across the list: the more a role consists of following a fixed script – transcribing, routing, processing – repeating, the higher its substitution risk. The roles with augmentation potential instead lean on judgment, accountability, and unstructured problems AI can assist with but not fully take over.
The World Economic Forum’s Future of Jobs Report 2025 found 40% of employers already plan to reduce headcount specifically where AI can automate tasks within five years. That’s not a vague future risk; it’s a near-term hiring plan at four in ten employers, concentrated on exactly the substitution-risk column above.
Substitution vs. Multiplication: The Real Distinction
This is where Bloor Research’s framing earns its keep. An AI system built on what Bloor Research calls an “imitation-first” design; built to pass for a person, judged on how convincingly it substitutes for one, makes displacement the logical outcome by construction. That’s the AI behind the exposure scores above: task-shaped, script-following, built to replace the task rather than extend the person doing it.
Bloor Research’s alternative design principle, OAI℠: origin-first, starts from the opposite premise: the human is the origin of the system, not a variable fitted around it afterwards. Read against that distinction, Nadella’s line: “you’ll lose your job to someone who uses AI” stops being a comforting slogan and becomes a design specification. The job at risk is the one still being done the old way, with no structure of its own behind it. The job that’s protected is the one now operating through something built from that person’s own expertise: captured, structured, and deployed continuously – what Bloor calls Digital Me℠.
What Actually Protects You
The defensive move isn’t avoiding AI, that ship has sailed. AI fluency requirements in job postings grew roughly sevenfold between 2023 and 2025, and the number of roles explicitly requiring it rose from about 1 million to 7 million in the same window. Being AI-literate is now table stakes, not protection.
Real protection looks different from tool fluency. It’s the distinction this series keeps returning to: a rented chatbot session, however fluently used, produces nothing that outlasts the session; no asset, nothing that compounds. What actually sits on the safe side of the exposure scores above is a structured, owned version of a person’s own expertise, deployed continuously rather than borrowed moment to moment. That’s not a slogan-level answer to “will AI take my job”; it’s the specific, buildable difference between being the task AI was designed to replace and being the person AI was built to extend.
Frequently Asked Questions (FAQs)
It’s measurable and current. Goldman Sachs estimates roughly 16,000 net US jobs are being eliminated by AI each month in 2026, and Stanford’s 2026 AI Index found nearly 20% lower employment among software developers aged 22–25 since 2024.
Roles built around fixed, repetitive, rules-based tasks: telephone operators, data entry clerks, insurance claims clerks, and customer service representatives all carry Goldman Sachs exposure scores above 75 out of 100.
Only partly. AI fluency is now close to a baseline expectation, not a differentiator; postings requiring it grew roughly sevenfold in two years. What protects a role is owning a structured, deployable version of your own expertise, not just knowing how to prompt a shared tool.
That an AI system built origin-first, starting from a specific person’s own knowledge, extends what that person can do rather than standing in for them. Bloor calls the design principle OAI℠, and its output for an individual, Digital Me℠.
Sources
- Goldman Sachs, via Yahoo Finance – “AI is cutting 16,000 U.S. jobs a month” – the monthly net job-loss estimate.
- Goldman Sachs – “The Jobs AI Is Likely to Boost – and Those It May Disrupt” – occupation exposure scores.
- Stanford HAI – 2026 AI Index Report, Economy chapter – entry-level and age-cohort employment data.
- McKinsey – “Agents, robots, and us” – the 57%-of-hours automation ceiling and AI-fluency hiring data.
- World Economic Forum – Future of Jobs Report 2025 – employer headcount-reduction plans.
Print Article for this blog Download here