Work

Automaation hyödyntäminen kolmannen sektorin tilivelvollisuuden tukena? Generatiivisen tekoälyprosessin kehittäminen governance-first-periaatteella ja tämän arviointi kansalaisjärjestöjen apurahatyön apuna

Year: 2026
Type: dissertation
Abstract: Non-governmental organisations (NGOs) are responsible for a substantial share of sustainability work. Despite this, an ever-growing share of their scarce capacity is going towards catering to administrative demands imposed on them by funders. This thesis takes a look at how generative artificial intelligence (AI) can be used as part of an NGO’s grant application and reporting processes in such a way that efficiency gains do not come at the cost of auditability, accountability or trust. The study follows a Design Science Research approach with the case organisation Mieli ry (Mental Health Finland). It proceeded in three phases: 1. four ex-ante interviews mapped the organisation’s current grant workflow and its pain points, 2. a governance-first, technology-agnostic operating model with a Retrieval- Augmented Generation (RAG) reference implementation was designed and constructed, and 3. the artifact was evaluated in workshops with three of the four participants interviewed in the ex-ante interviews. The findings show that the administrative burden is one primarily concerning evidence, coordination, as well as interpretation. Another finding shows that the acceptance of AI support depends on visible claim-level provenance, transparent data and model boundaries, as well as clear human responsibility. The third finding shows that wrongly contextualised, or otherwise misleading, output is viewed as more dangerous than omissions. The thesis contributes with a theoretical framing of governance-first AI adoption as governed evidence work, supported by five governance principles for AI-assisted grant and reporting work, as well as a demonstration of how a design artifact can serve the role of a research instrument in third-sector settings, including those that are accountability- heavy. The thesis concludes by finding that automation can support third-sector accountability, but only under the following conditions: as a bounded assistant in a workflow that retains responsibility with a human, and only as long as the ability of the organisation to account for what it says is preserved.
Author: Kull, Rasmus
Institution: Aalto University
Language: English
Open Access status: green
OpenAlex ID: W7204015870
Cites: 63
Cited by:
Related to: 10