Developing Public Sector Artificial Intelligence Maturity Model

There is a lot of academic literature on developing maturity models, but almost exclusively for business entities. Moreover, even when these models are created, they remain at a very high conceptual level with seemingly no applicability in the real world. In Information Systems research, there is a whole scientific approach called Design Science Research (DSR) to developing IS artifacts that address a real-life problem. This kind of research intends to create unique or vastly improved artifacts while at the same time contributing to knowledge-creation which is the whole purpose of science. Unfortunately, although the DSR requires that such artifacts need to be tested, and improved based on the feedback, tested again, and eventually validated to be useful in real life, these studies rarely get beyond developing the AIMMs. Hence, in academic literature, you will not find a case of AIMM that would be tested in a real-life setting, and either proved to be useful and moved into the 'production' phase, or sent back to the drawing board. That being said, some consultancies have developed AIMM-s and use them most likely successfully in their daily work. This is a very instrumental use of such assessment tools with no need to adhere to scientific

Probably the reason for such a shortcoming is that testing in real life takes a lot of effort, requires access to many organisations who need to be willing to experiment and share the feedback, and an ever-present need to publish articles in reputable academic journals does not leave enough time to engage in such a long-term endeavour.

My master's studies at the Tallinn University of Technology eGovernance Technologies and Services programme have allowed me to study e-governance from many angles. One that I really like is the interplay between organisational context and technology. AI capability of public organisations depends to a large extent on organisational factors, such as people's skills, leadership interest, or business challenges, as much as on technological capabilities, such as the level of digital service provision, data governance, etc. In addition, there is always an environmental context, which in public administration depends on political expectations, a central policy direction, peer pressure, and a legal framework. Developing a public sector AIMM is where all these factors need to be dialed in to create a viable model, which makes it fascinating. Estonia, where new technologies are relatively quickly embraced and which consequently has some 130 AI use cases in the public sector from around 60 organisations, is an ideal test bed for developing such a model.

For the master's thesis, together with my supervisor Richard Michael Dreyling III, I intend to develop an AIMM, which covers all the grounds (called dimensions in the model) relevant to becoming a successful public sector AI innovator and allows one to place an organisation onto one of five maturity levels on each of these dimensions, demonstrating a progressively higher AI-related capability. In addition, I plan to collect feedback on the model prototype from experts in the field before deploying it to at least one public organisation to simulate a self-assessment. Most likely, it will reveal different weaknesses compared to experts, considering that the concepts in the field of AI are not yet well understood and shared. My purpose is to identify such concepts to some extent in advance, and also through the self-assessment exercise, to create a guideline document as an additional artifact of my work. The self-assessment environment is created in a publicly available online survey tool, and the respondent profiles are agreed upon in advance with the participating public organisation. The benefit for the public organisation at this stage is that they learn something about their own AI capabilities, and perhaps identify also some of the areas for improvement that subsequently can be addressed through a plan of action. Communication within the organisation to jointly analyse the results and agree on the improvement projects is normally part of a self-assessment exercise, but due to the limited timeframe will not be fully pursued at this test phase.

More to follow after the results can be reported.

The Methodological Guidelines are available below

AIMM Guidelines_Final.docx

Download DOCX • 269KB

.