AI innovation at the boundaries: Justifying a generative AI decision support tool

The growing uptake of generative AI by public administrations raises important questions about the continuity of AI innovations and the ‘thingness’ of AI, despite recurrent failures of these emerging technologies to fulfil their promises. We combine the frameworks of boundary work and regimes of justification and methodologically draw on nearly 1.5 years of ethnographic fieldwork following the innovation process of a generative AI decision support tool in a Finnish public organization. We capture how the team driving this AI innovation engages in boundary-crossing practices, enacting diverse conceptions of moral and common good to ensure the continuity of the tool's development process. We identified nine powerful justification frames related to the tool, its innovation process and its underlying ideology that sustained the generative AI tool's innovation. Three frames in particular – centring on efficiency, well-being and equality – formed a powerful trio capable of making the tool compelling across organizational boundaries within the context of the Nordic welfare state. The frames fostered a protective structure around the tool, simultaneously legitimizing it as it was developed. We show how enacting these frames to safeguard the tool and its imagined value from criticism – specifically a lack of accuracy, precision and consistency – made it an irresistible thing to influential organizational actors.