Vorträge

AI coding assistants are used daily by most professional developers, yet organisations treat them as individual tools within stable team structures, leaving fragmentation and governance unaddressed. This study investigates whether practitioners perceive AI integration as an organisational design problem and whether they consider a three-layer model (System, Platform, and Product Teams) plausible.

We surveyed n = 99 professional developers with 28 Likert items normalised to a −3 to +3 scale. Practitioners endorse AI development environments as shared infrastructure, yet a gap between the model's abstract plausibility (M = +1.49) and its perceived applicability (M = +0.76) reveals an acknowledged maturity deficit.

Intensive AI users frame integration as an organisational challenge more strongly (d = 1.63, p < .001) and report pronounced shifts from producing code towards reviewing and steering AI outputs (r = .790, p < .001). Agile practices evolve selectively, with Continuous Integration and TDD gaining governance significance. Organisations therefore have a window of opportunity to respond structurally before fragmentation becomes entrenched.

Generative KI beschleunigt die Software-Produktion und einzelne UX-Tätigkeiten messbar. Zugleich gerät die Nutzerperspektive ökonomisch und methodisch unter Druck. Dieser Beitrag verdichtet die Evidenz der Jahre 2023 bis 2026 zu einer Landkarte für die Praxis. Welche KI-Einsätze in der UX-Arbeit sind empirisch tragfähig, welche nicht? Und woran erkennt man den Unterschied, bevor Schaden entstanden ist? Entlang von sieben Strängen (Demokratisierung von Research, tragfähige Werkzeug-Einsätze, synthetische Nutzer, Dark Patterns in generierten Oberflächen, komprimierte Iterationszyklen, Rollenwandel und Arbeitsmarkt) zeigt sich eine konsistente Trennlinie. KI ist dort tragfähig, wo der Maßstab für Richtigkeit vor dem Einsatz feststeht und geprüft werden kann. Sie scheitert dort, wo dieser Maßstab erst aus den Daten entstehen muss. Aus dieser Faustregel folgen konkrete Handlungsempfehlungen für Designer:innen und Researcher, UX-Manager:innen und Organisationen.

Large language models (LLMs) are marketed as "synthetic users": simulated research participants that promise user insights without recruiting humans. The empirical literature on this idea looks contradictory. Some studies report close alignment between simulated and human data, others report systematic failure.

We argue that this contradiction disappears once we change the question. Instead of asking whether synthetic users work, we ask which function of user research they can serve. Read this way, the evidence is consistent: LLM-simulated participants can approximate averages under favorable conditions, but they fail to reproduce variation, subgroup structure, cultural differences, and lived experience.

We show that this split is built into the technology rather than a sign of immaturity: it follows from the training pipeline of current LLMs, and persona prompting moves the model's average without restoring human diversity. We combine the evidence into an evidence map, a decision aid that places common UX research activities by the kind of knowledge they need and by how much depends on their results.

We close with a research agenda for HCI, including controlled comparisons on established UX questionnaires.