Paper Number
ECIS2026-1455
Paper Type
CRP
Abstract
Generative Artificial Intelligence (GenAI) is promoted as democratizing knowledge work by lowering entry barriers for non-experts. We examine this claim in process mining. Using a mixed-methods design, we compared a GPT-based assistant with a conventional toolkit in an experiment with 19 novices, complemented by interviews. Participants completed advanced tasks with both tools. GenAI appeared faster, but verification demands offset these gains, and accuracy did not improve. More importantly, GenAI introduced side effects: some participants tied responsibility to their confidence in AI outputs, others treated it as fixed, and many displayed blind trust in surface cues that raised confidence without enabling validation. These dynamics show how apparent empowerment can mask epistemic disempowerment. We contribute by (1) empirically testing GenAI’s democratization claims in a contested BPM domain, (2) theorizing responsibility duality as a socio-technical negotiation in human-AI work, and (3) adapting “transparency theater” to capture performative transparency inflating confidence without enabling validation.
Recommended Citation
Eigenmann, Martin; Matter, Philipp; Leikert-Boehm, Ninja; and Miscione, Gianluca, "At What Cost? Democratizing Process Mining With GenAI" (2026). ECIS 2026 Proceedings. 4.
https://aisel.aisnet.org/ecis2026/bpm/bpm/4
At What Cost? Democratizing Process Mining With GenAI
Generative Artificial Intelligence (GenAI) is promoted as democratizing knowledge work by lowering entry barriers for non-experts. We examine this claim in process mining. Using a mixed-methods design, we compared a GPT-based assistant with a conventional toolkit in an experiment with 19 novices, complemented by interviews. Participants completed advanced tasks with both tools. GenAI appeared faster, but verification demands offset these gains, and accuracy did not improve. More importantly, GenAI introduced side effects: some participants tied responsibility to their confidence in AI outputs, others treated it as fixed, and many displayed blind trust in surface cues that raised confidence without enabling validation. These dynamics show how apparent empowerment can mask epistemic disempowerment. We contribute by (1) empirically testing GenAI’s democratization claims in a contested BPM domain, (2) theorizing responsibility duality as a socio-technical negotiation in human-AI work, and (3) adapting “transparency theater” to capture performative transparency inflating confidence without enabling validation.
When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.