AI is changing technology work and creating demand for some new skills and tasks, but current evidence does not show that it is creating more tech jobs overall than it displaces. Exposure to AI, a task that can be automated, a job vacancy and net employment growth are different things. The distinction matters: AI may reshape what a developer or analyst does without eliminating the role, while some entry-level or routine work may face real pressure.
Is AI taking tech jobs or creating them?
It is doing both kinds of work: automating some tasks and enabling new ones. Whether that produces more jobs or fewer depends on how firms use AI, whether productivity gains lead to expanded output, and how many people are still needed for the work that remains.
As an Amazon Associate I earn from qualifying purchases.
The OECD’s 2026 synthesis describes the balance as unresolved: AI can automate existing tasks, create new tasks and occupations, and raise productivity. Those effects can push employment in different directions. More output might increase demand for workers, for example, but a firm may also use automation to produce the same output with fewer people.
That is why a rise in AI-related vacancies cannot by itself prove that AI is creating net jobs, just as an occupation’s exposure to AI does not prove that its workers will be replaced. The ILO’s 2025 review puts it plainly: “Generative AI is altering work processes, task composition, and organizational design, yet its effects on employment and the macroeconomy remain unresolved.”
What the headline numbers actually measure
Several widely cited figures describe adoption, exposure, skills demand or study results—not the net number of tech jobs gained or lost.
| Finding | What it measures | What it does not establish |
|---|---|---|
| Firm AI uptake rose from around 7% in 2021 to 20% in 2025 in OECD countries. | Adoption by firms, as summarized by the OECD in 2026. | How many jobs were created or eliminated as a result. |
| One in four workers globally is in an occupation with some degree of generative-AI exposure. | Occupational exposure estimated by the ILO in its 2025 update, using a refined method covering nearly 30,000 tasks. | That one in four jobs will disappear. The ILO says most jobs are more likely to be transformed than made redundant. |
| Mean automation score: 0.29 in the 2025 ILO update, compared with 0.30 in 2023; standard deviation: 0.14, compared with 0.30. | How the ILO’s estimated task-level automation potential is distributed across occupations in the two updates. | A forecast of actual layoffs or the number of jobs firms will automate. |
| AI-related skills accounted for nearly one-third of new IT skills in an IMF analysis of job postings across six countries. | The share of new IT skills identified as AI-related in vacancy data analyzed by IMF staff in 2026. | Net hiring, filled jobs or economy-wide employment growth. |
| Generative AI had the largest absolute increase in demand in 2024 among the four advanced economies in the IMF analysis. | A change in demand visible in postings within that sample. | A universal trend across all countries or proof of a net increase in tech employment. |
As the OECD’s 2026 skills synthesis puts it, “AI is transforming jobs, but not necessarily destroying them.”
Which technology tasks are most exposed?
Exposure is about whether AI could affect tasks within an occupation; it is not the same as full automation. A role that includes routine, repeatable work may be more exposed in those duties, while work requiring judgment, context, coordination or responsibility can remain dependent on people. Most jobs combine different kinds of tasks, so a single label such as “AI-proof” or “doomed” obscures more than it explains.
The ILO’s global estimate is deliberately framed at the occupation level and includes varying degrees of exposure. Its conclusion is that human input remains necessary in most work, making transformation more likely than wholesale redundancy. That does not rule out job losses in particular firms or tasks; it means exposure statistics alone cannot tell how many will occur.
Outcomes also vary by worker experience. The ILO’s 2025 evidence review finds indications that demand for novice workers may decline in some settings. If AI lets experienced staff complete routine assignments faster, employers might hire fewer beginners for that work, even while experienced workers take on oversight or more complex tasks. This is a potential pressure, not a settled prediction for every tech occupation.
Why productivity gains do not automatically mean more jobs
Studies reviewed by the ILO report productivity gains of 20% to 60% in controlled randomized trials and 15% to 30% in field experiments. These findings are context-dependent: they come from particular tasks, tools and study settings, and should not be treated as a forecast for every role or company.
Higher productivity can make a business more competitive or let it serve more customers. If that expands output, demand for workers could rise. But if a company can meet its needs with fewer labor hours, it may hire less or reorganize teams. The same technology can therefore increase output, reduce some tasks, create new responsibilities and change staffing at the same time.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errorsThe evidence has further limits. Many experiments focus on relatively simple tasks, and results depend on the models and workplaces studied. AI exposure measures do not show whether an employer has adopted a tool, how it has integrated it, or whether adoption substitutes for labor. A reported productivity gain is not a headcount result.
What vacancy data says about AI-related tech skills
IMF staff analysis of job postings in six countries found that AI-related skills made up nearly one-third of new IT skills in the analysis. Generative AI had the largest absolute increase in skill demand in 2024 among the four advanced economies in its sample. The postings also point to examples including software engineering, data engineering, Python, SQL and data pipelines.
These are useful signals about what employers are asking for in advertised roles. They are not a tally of new jobs filled: postings can change without the same change in hiring, and the analysis covers a specific set of countries and vacancies. Nor does a skills shortage mean that any particular short course will lead to a new position. The OECD notes that skills shortages can constrain firms’ adoption of AI, which may slow or shape its effects on work.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Why tech employment changes cannot all be blamed on AI
Technology hiring has also been affected by broader labor-market conditions. The OECD’s 2026 Employment Outlook describes weak hiring and cuts in high-tech and information-sector employment after pandemic-era overhiring. It also notes low hiring in the United States and challenges facing young entrants.
Those conditions matter when interpreting layoffs or fewer vacancies. A company reducing staff may be responding to earlier overexpansion, weaker demand or other business pressures; an AI connection should not be assumed without evidence. The available sources do not isolate AI as the cause of overall tech employment changes.
What skills are worth building?
The evidence supports preparing for work that combines technical fluency with skills that help people use, evaluate and apply technology. The OECD recommends foundational literacy and numeracy, ICT capabilities, AI literacy, critical thinking, creativity, collaboration and continued learning. The IMF posting analysis identifies technical examples such as software engineering, data engineering, Python, SQL and data pipelines.
- Build durable foundations: strengthen analytical and digital skills that transfer across tools and employers.
- Learn to work with AI: understand what AI tools can and cannot do, and check their output rather than treating it as authoritative.
- Pair technical ability with context: practice explaining trade-offs, collaborating with others and applying technical work to real problems.
- Keep adapting: tools and employer needs change, so ongoing learning is more realistic than expecting one credential to guarantee security.
These are evidence-informed areas to explore, not a guaranteed route to a job. Hiring depends on location, industry, firm adoption and economic conditions as well as skills.
So, will AI replace software developers?
The cited evidence does not support a reliable yes-or-no forecast for software developers as a whole. It supports a narrower conclusion: AI can affect tasks within technical roles, demand for AI-related skills is visible in some vacancy data, and some workers or duties may face substitution pressure. But exposure, productivity and job postings do not establish whether total developer employment will rise or fall.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
For an individual career decision, focus on the work involved rather than a job title alone. Consider how routine the tasks are, whether the role relies on judgment and collaboration, how much AI adoption is occurring in the relevant industry and region, and whether employers are actually hiring—not just listing new skills in vacancies.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
