Integrated Research Seminar: Public Sector Innovation Lab

AI-GENERATED PROBLEMS FOR PUBLIC SERVICES

Group members: Griwan Raj Khakurel, Mahrukh Khan, Mansoor Aman, Muhammad Huzaifa, Muhammad Umar

1. The Case and Problem

Generative AI and autonomous agents make it possible to interact with public authorities at unprecedented scale and very

low cost. AI systems could automatically generate and submit large numbers of information requests, requests for

clarification, complaints, or applications. While many such interactions may be legitimate, the volume and speed of

machine-generated demand could create significant administrative burden and potentially disrupt normal service delivery.

Our case therefore examines how public services can prepare for new types of demand and misuse created by AI agents.

2. Focus and Research Question

We will initially examine this issue across the Estonian public sector rather than assuming that one organisation owns the

problem. Our central question is: How might Estonian public authorities remain open and accessible while handling

legitimate machine-generated interactions at AI-enabled scale? This will guide exploration without presupposing a

particular solution. We will investigate where AI-enabled demand could enter existing service processes, how

organisations currently distinguish legitimate requests from misuse, which forms of automated interaction create

administrative pressure, and what legal, procedural, organisational, or technical safeguards could address that pressure. If

evidence identifies a particular service or process as a clearer and more feasible case, we will narrow the study

accordingly.

3. Stakeholders and Fieldwork

We will seek a range of perspectives, including process owners, technical and data specialists, frontline officials, service

users, suppliers, and oversight actors. Initial contacts include Otto Mättas at the State Information System Authority (RIA),

Anni Lehari, the Ministry of Finance, the Ministry of Justice and Digital Affairs, and the Government Office. Potential

service-level cases include the Estonian Unemployment Insurance Fund, Social Insurance Board, and City of Tallinn.

Contact has already been made with Otto Mättas and Anni Lehari, and the contact list will expand as interviews clarify

which organisations and processes are most directly affected.

4. Data Collection

We will combine semi-structured interviews with desk research. The fieldwork will include a minimum of eight

interviews, approximately 45–60 minutes each, with stakeholders representing process ownership, technical/data expertise,

frontline or user experience, and policy or oversight. Interviews will emphasise concrete examples, previous incidents,

attempted responses, failures, and practices currently taken for granted. Desk research will cover relevant Estonian policy

documents, legislation, administrative rules, existing digital-service arrangements, AI initiatives, and publicly available

information on automated or high-volume interactions.

5. Analysis and Expected Outcome

We will map the selected service or process, including its current workflow, decision points, data foundations, legal

mandate, organisational capacity, users and access, and oversight mechanisms. We will compare stakeholder accounts with

documentary evidence and clearly distinguish reported facts, documented evidence, and our own inferences. A key

analytical tool will be an assumptions register containing approximately 10–15 substantive technical, data, legal,

organisational, behavioural, financial, or political assumptions. For the most important assumptions, we will assess

whether evidence supports, questions, or contradicts them, what would break if they proved false, and how they could be

tested or de-risked. This will help us distinguish immediate workload or capacity concerns from deeper structural

constraints and identify a specific, researchable problem that can later support a design brief and actionable pathway for

change.

6. Immediate Next Steps

We will complete the first round of stakeholder outreach, conduct and refine the initial interviews, begin desk research,

and use early evidence to identify the organisation and process most suitable for deeper case study. The interview log and

assumptions register will be maintained from the first interview onward so that findings remain traceable throughout the

project.

Integrated Research Seminar: Public Sector Innovation Lab

AI-GENERATED PROBLEMS FOR PUBLIC SERVICES

Group members: Griwan Raj Khakurel, Mahrukh Khan, Mansoor Aman, Muhammad Huzaifa, Muhammad Umar

1. The Case and Problem

Generative AI and autonomous agents make it possible to interact with public authorities at unprecedented scale and very

low cost. AI systems could automatically generate and submit large numbers of information requests, requests for

clarification, complaints, or applications. While many such interactions may be legitimate, the volume and speed of

machine-generated demand could create significant administrative burden and potentially disrupt normal service delivery.

Our case therefore examines how public services can prepare for new types of demand and misuse created by AI agents.

2. Focus and Research Question

We will initially examine this issue across the Estonian public sector rather than assuming that one organisation owns the

problem. Our central question is: How might Estonian public authorities remain open and accessible while handling

legitimate machine-generated interactions at AI-enabled scale? This will guide exploration without presupposing a

particular solution. We will investigate where AI-enabled demand could enter existing service processes, how

organisations currently distinguish legitimate requests from misuse, which forms of automated interaction create

administrative pressure, and what legal, procedural, organisational, or technical safeguards could address that pressure. If

evidence identifies a particular service or process as a clearer and more feasible case, we will narrow the study

accordingly.

3. Stakeholders and Fieldwork

We will seek a range of perspectives, including process owners, technical and data specialists, frontline officials, service

users, suppliers, and oversight actors. Initial contacts include Otto Mättas at the State Information System Authority (RIA),

Anni Lehari, the Ministry of Finance, the Ministry of Justice and Digital Affairs, and the Government Office. Potential

service-level cases include the Estonian Unemployment Insurance Fund, Social Insurance Board, and City of Tallinn.

Contact has already been made with Otto Mättas and Anni Lehari, and the contact list will expand as interviews clarify

which organisations and processes are most directly affected.

4. Data Collection

We will combine semi-structured interviews with desk research. The fieldwork will include a minimum of eight

interviews, approximately 45–60 minutes each, with stakeholders representing process ownership, technical/data expertise,

frontline or user experience, and policy or oversight. Interviews will emphasise concrete examples, previous incidents,

attempted responses, failures, and practices currently taken for granted. Desk research will cover relevant Estonian policy

documents, legislation, administrative rules, existing digital-service arrangements, AI initiatives, and publicly available

information on automated or high-volume interactions.

5. Analysis and Expected Outcome

We will map the selected service or process, including its current workflow, decision points, data foundations, legal

mandate, organisational capacity, users and access, and oversight mechanisms. We will compare stakeholder accounts with

documentary evidence and clearly distinguish reported facts, documented evidence, and our own inferences. A key

analytical tool will be an assumptions register containing approximately 10–15 substantive technical, data, legal,

organisational, behavioural, financial, or political assumptions. For the most important assumptions, we will assess

whether evidence supports, questions, or contradicts them, what would break if they proved false, and how they could be

tested or de-risked. This will help us distinguish immediate workload or capacity concerns from deeper structural

constraints and identify a specific, researchable problem that can later support a design brief and actionable pathway for

change.

6. Immediate Next Steps

We will complete the first round of stakeholder outreach, conduct and refine the initial interviews, begin desk research,

and use early evidence to identify the organisation and process most suitable for deeper case study. The interview log and

assumptions register will be maintained from the first interview onward so that findings remain traceable throughout the

project.