Gartner projects that artificial intelligence will create 10.6 million jobs and eliminate 8.9 million during 2025–2030, leaving a net gain of 1.7 million. Published on November 10, this is a forecast rather than a report of employment already achieved. Kaldata brought the subject into fresh Bulgarian coverage on November 15. Our analysis examines the original announcement and separates its aggregate scenario from promises about an individual worker's future.

Gross flows and a net balance answer different questions. A positive overall figure can coexist with redundancies at a particular business. New roles may require different skills, arise somewhere else or arrive later. Subtracting two global numbers therefore does not measure how readily someone affected by displacement will find another job. That distinction is SIExpo's interpretation of the indicator, not an additional finding attributed to Gartner.

Gartner also considers transformation within existing work rather than only the disappearance of entire positions. A separate reference point is the International Labour Organization's May 2025 study of generative-AI occupational exposure. It combines task information, expert input and model-based assessment. Under that framework, roughly one in four workers is in an occupation with some exposure, while task transformation is considered more likely than complete automation. This is not a second measurement of Gartner's projected 1.7 million gain, and it does not independently validate that number.

For an employer, those frameworks support different discussions. A forecast of positions can inform broad workforce planning. An analysis of tasks can help identify what a person will actually do after a tool is introduced. Consider a proposed system that drafts a document: generating the draft, checking its accuracy, discussing it with a customer and accepting responsibility for the final decision are separate activities. This is an evaluation example, not a description of a deployment measured by either source.

A useful internal assessment would begin with a map of the work: which actions recur, where errors are costly and who can detect them. A bounded trial should then measure quality and time, including the effort spent correcting generated output. Only after that can a business reasonably discuss whether the tool releases capacity, changes a role or creates an additional oversight burden. These are SIExpo's suggested assessment criteria; we have not tested them inside a company for this article.

Workers can similarly ask which tasks will remain their responsibility and how the result will be judged, rather than relying on a single prediction about whether a whole profession will disappear. Training deserves evaluation beyond the number of people attending a course. A practical review should ask whether participants can use the tool on relevant work, recognise its limitations and handle data safely. Access to practice and a clear allocation of responsibility matter alongside familiarity with a new interface.

Neither source offers a personal employment guarantee for Bulgaria. The figures are useful planning inputs when their time horizon, method and uncertainty remain visible. A positive net scenario warrants attention, but the meaningful local question is how actual work changes and what support people receive during that transition. Presenting the projection with those boundaries preserves its value without turning a global estimate into a certainty about anyone's next job.

Editorial verification: 2026-11-15T17:59:15.509180+00:00.