- Published on
The jobs did not disappear. The job description did.
- Authors
- Name
- AbnAsia.org
- @steven_n_t
Listen to this article · The jobs did not disappear. The job description did.
0:00 / 0:00
Watch this article · The jobs did not disappear. The job description did.
Two headlines landed in the same week and they look like they contradict each other.
Tech job postings rose 4% in October and are up 20% year on year, according to Dice. AI and machine learning postings grew 91% over the same period — more than four times the overall rate. CompTIA counted 290,892 new tech postings in October, up 22% on last year, with 366,322 active postings asking for AI skills.
At the same time, software firms are still smaller than they were a year ago, and the layoff announcements have not stopped. ServiceNow, Block and NICE all announced cuts in the first week of November.
Both are true, and the resolution is not "AI is taking the jobs". It is that the jobs are being rewritten faster than the people in them can be retitled.
The layoffs are a correction, not a cull
The 2026 numbers are gentler than the story suggests. Roughly 190,000 tech workers have been laid off this year, on track to finish well below the 430,000 of 2023. WARN notices fell to 16,600 in October, the lowest monthly total since 2022.
What is happening at the software firms is a post-pandemic correction meeting a change in what they build. Companies that hired for headcount in 2021 are now hiring for a specific capability, and the capability is not "writes code". Dice found employers naming Claude and GitHub Copilot in requirements — and, in the same postings, asking for technical debt, maintainability, design review and defect management. That is the profile of someone who reviews AI-assisted code, not someone who types it.
The squeeze lands hardest at the entry point. Indeed's data for the first quarter of 2026 found 69.3% of software development postings were senior and 4.5% were entry-level. NACE reports more than a third of entry-level jobs now require AI skills, nearly triple the share of a year earlier. The profession is not shrinking; its front door is.
What is actually growing
The clearest new title is the forward deployed engineer — FDE.
The job is what it sounds like: an engineer who sits with the customer, works out what the model can actually do in their environment, and builds the thing that ships. OpenAI's posting describes leading "complex end-to-end deployments of frontier models in production", owning discovery, scoping, system design and rollout, and feeding what the field learns back into the product and model roadmaps. Anthropic's version asks for eight or more years of technical, customer-facing work and 25–50% travel.
The demand is measurable. As of 3 November, OpenAI listed 75 open FDE roles with a median posted base of 145,000 to 205,000.
year on year, October 2026 (Dice)
year on year, same period
Why this role, and why now
A model is not a product. Getting one into a bank, a hospital group or a logistics operator means dealing with their data model, their security review, their approval process and the fact that the demo worked on a clean dataset.
That work used to be called solutions engineering, or consulting, or professional services, and it was priced accordingly. What changed is that the deliverable is now the company's core product, deployed inside someone else's building — and the person doing it has to be good enough to change the roadmap when the field says the model is wrong.
That is why the pay is closer to a senior engineer than to a sales engineer, and why the hiring is happening at the labs themselves rather than only at integrators.
Embed with the customer
Sit inside their process long enough to see where the model actually helps and where it quietly fails.
Scope the deployment
Turn a vague ambition into a system with an evaluation set, a data path and a rollback plan.
Build and ship it
Full-stack work in the customer’s environment, not a slide deck about what could be built.
Feed it back
Carry what the field learned into the product and research teams, which is the part most services roles never had.
the peak year
on track to finish well below 2023
What it means if you are the engineer
Three things follow from the numbers, and none of them is "retrain as a researcher".
Ship something that runs. The postings ask for production experience with models — evaluation, agents, deployment — not certificates. A working system you can describe is the entry ticket.
Customer contact is a skill, not a detour. The FDE premium exists because engineers who can sit with a customer and still write production code are scarce. That scarcity is the opportunity.
The generalist junior role is the one under pressure. Not because the work vanished, but because the first two years of a career used to be spent doing exactly the tasks models now do passably. The replacement for that apprenticeship has not been designed yet, and that is the real problem in the data.
The jobs are not going away. The description is being rewritten in public, and the people who read the new one first are the ones getting paid.
Source: Dice October 2026 tech jobs report (3 November 2026); CompTIA October 2026 analysis (7 November 2026); BLS payrolls release (6 November 2026); TrueUp and Revelio Labs layoff tracking; OpenAI and Anthropic job postings as listed 3–4 November 2026; Indeed Hiring Lab and NACE entry-level data.
Contact: [email protected]