AI, software development, and work
The impact of AI on software development and work
The presentation places different findings about AI adoption, productivity, learning, labor markets, media trust, and resource use side by side. It provides a starting point for a differentiated discussion.
What it covers
AI changes tasks in software development, but its effects depend heavily on the task, experience, tooling, team process, and measurement method.
Technical decisions should distinguish controlled experiments, field studies, surveys, correlations, and forecasts. They answer different questions and cannot be condensed into one overall claim about impact.
Main ideas
- High adoption of AI tools does not by itself establish the quality or long-term value of their output.
- Productivity measurements depend on the task, time frame, experience level, and quality criteria.
- Effects on learning and professional judgment deserve attention alongside short-term speed.
- Labor-market forecasts describe possible developments. They do not promise an outcome for a particular role, company, or time frame.
- Trust, energy demand, water use, and hardware costs belong in an evaluation of AI use even though they sit outside the source code.
Limits and current status
- The slides reflect the research and talk as of September 2, 2026. Metrics, products, and forecasts may have changed since then.
- The referenced studies use different designs. Results from experiments, field studies, surveys, correlations, and forecasts are not directly comparable.
- The sources slide lists titles but is not a complete bibliography. Consult each original source and its current edition before making a decision.
- This context page does not reproduce third-party images or press copy from the slides, and it draws no blanket claim of benefit or harm from the sources.
Original presentation
2.LemgoTechTalk - KI.pdf
14 pages • 14.6 MB