Most AI job-threat articles are written by people with something to lose. Consulting firms that sell transformation projects. Law firms that want to keep billing hours. Software companies that don't want to spook their customers. This one isn't.
Here's how to read it: the risk levels below are tied to the AGI timeline, not just current AI. Today's tools are impressive but narrow. What changes everything is when models cross the threshold of general reasoning — when they can handle ambiguity, novel situations, and multi-step judgment the way a smart human does. The median expert forecast for that is 2031. It was 2060 five years ago.
So the question isn't "can ChatGPT do my job today?" It's "what does my job look like in 2028, 2031, 2034?" Here's the breakdown.
| Profession | Risk level | What goes first | What survives longest | Key inflection |
|---|---|---|---|---|
| Software Engineer | High | Junior dev work, boilerplate, bug fixing, code review | Systems design, architecture, high-stakes debugging | ~2027–28 for junior roles |
| Marketer / Content Creator | High | Copy, SEO content, ad creative, email sequences | Brand strategy, audience trust, creative direction | Already underway |
| Lawyer | Medium | Contract review, discovery, legal research, due diligence | Courtroom advocacy, client relationships, judgment calls | ~2028–30 for associate-level work |
| Accountant / Financial Analyst | High | Bookkeeping, tax prep, financial reporting, variance analysis | CFO-level strategy, audit sign-off, complex advisory | ~2026–27 for routine work |
| Doctor / Physician | Medium | Radiology reads, diagnostic pattern-matching, notes | Surgical intervention, patient relationship, complex diagnosis | ~2029–32 for diagnostic roles |
| Graphic / UX Designer | High | Spec work, asset production, basic UI patterns | Creative direction, user research, novel problem framing | Already underway |
| Journalist / Writer | Medium | Commodity reporting, press release coverage, data articles | Original reporting, investigative work, unique voice | ~2026–28 for commodity content |
| Data Analyst | High | Report generation, dashboards, SQL queries, EDA | Strategic framing, stakeholder translation, novel analysis | ~2026–27 for routine work |
| HR / Recruiter | High | Resume screening, JD writing, initial outreach, scheduling | Sensitive conversations, culture judgment, final decisions | ~2026–28 |
| Teacher / Trainer | Low–Med | Curriculum writing, grading, explainer content | Mentorship, motivation, in-person presence, accountability | ~2030+ for core roles |
| Therapist / Counselor | Low | Psychoeducation, between-session support, symptom tracking | Therapeutic relationship, embodied presence, nuanced judgment | Companion tools supplement; replacement unlikely before 2035+ |
| Plumber / Electrician | Low | Diagnostics, scheduling, quoting | Physical execution — almost all of it | Physical AI/robotics: 2033+ at meaningful scale |
Each post below ties the AI risk directly to where the AGI timeline sits and what the benchmark data actually shows. No job boards, no "upskill in Python" platitudes.
SWE-bench data, what junior vs senior risk actually looks like, and the 2027–28 inflection point.
This one is already happening. The question is which part of marketing survives longest.
The honest take — without law firm PR. Associate-level work, 2028–30. Here's what the data says.
Intuit says no. They also sell accounting software. Here's the unbiased breakdown.
AI already outperforms radiologists on specific imaging tasks. The honest breakdown of what's actually at risk in medicine.
Midjourney and Figma AI are already reshaping spec work. What survives — and what doesn't.
Commodity reporting is already AI-assisted. Original reporting and unique voice are a different story.
Natural-language BI tools and AI coding assistants are compressing junior analyst demand. Now.
Resume screening and JD writing are nearly automated. The bias problem, and what survives.
Khan Academy's Khanmigo and Duolingo are operational AI tutors. Corporate L&D is most exposed. K–12 is different.
The therapeutic alliance is the most resilient mechanism in all of professional services. Here's why.
Plumbers, electricians, HVAC — dexterous robotics in unstructured environments. The 2033+ question.
howcloseisagi.com reaches builders, founders, and professionals tracking the AGI timeline. One sponsor slot per post — no programmatic ads, no clutter.
Get in touch →The honest answer is: it depends on where you sit in the distribution within your profession — not whether your profession is "at risk." The question isn't "will AI replace lawyers?" It's will AI replace the work that the bottom 40% of lawyers spend most of their time on? The answer to that question is almost universally yes, and it's a matter of when, not if.
What that means practically:
The people best positioned aren't the ones ignoring this. They're the ones building with it now — understanding what it can do, where it breaks, and how to sit above the automation layer rather than in it.