I don't have real-time internet access, so I can't browse for actual articles, which means I need to avoid fabricating fake citations since this is for government planning and accuracy matters a lot here. My plan is to list real, well-known open-access works on AI/automation in public administration that I'm reasonably confident actually exist, filtering for post-2020 publication dates, while clearly flagging that the user must independently verify each one. I'm going through candidates like Bullock, Young, and Wang's work on AI and bureaucratic discretion, but noting some fall in 2019-2020 which may not meet the "after 2020" cutoff, so I'm narrowing toward 2021+ publications like Madan & Ashok's work on AI adoption in government. Also considering de Sousa et al. (2019) on AI in the public sector as a literature review, and Selten, Robeer & Grimmelikhuijsen (2023) in PAR on street-level bureaucrats trusting AI recommendations—fairly confident both are real. Then trying to recall whether Bullock, Huang & Kim's work on machine intelligence and bureaucracy matches an actual title like "Artificial intelligence, discretion, and bureaucracy." I'm also checking Meijer, Lorenz & Wessels (2021) in PAR on algorithmization of bureaucratic organizations and predictive policing, and Giest & Klievink (2022) in Public Management Review on AI changing bureaucrats' roles—both seem like legitimate, verifiable citations. I'm also noting OECD's 2024 grey-literature report on governing with AI, and confirming Zuiderwijk, Chen & Salem (2021) in Government Information Quarterly as a strong, open-access systematic review fit for this topic. Continuing through the list, I recall Mikalef et al. on AI capabilities in European municipalities, Neumann, Guirguis & Steiner (2024) in Public Management Review on AI adoption comparisons, and Maragno et al. (2023) on chatbots in public entities — all seem like genuine, relevant citations with likely open access. I'm also checking van Noordt & Misuraca (2022) in Government Information Quarterly on landscaping AI use across the EU, which I'm fairly confident is real and open access, plus grey literature like JRC AI Watch reports and Deloitte's Eggers/Schatsky/Viechnicki work, while flagging lower confidence on some other citations like Nasim et al. Given the focus on AI/automation in government and its effects on administrative jobs, I'm settling on five strong picks, starting with Zuiderwijk, Chen & Salem (2021) in GIQ — a systematic review with high confidence that touches on workforce and capacity implications and is openly accessible. 2. Ahn & Chen (2022), GIQ, on government employees' perceptions and willingness to adopt AI — moderately strong. 3. Selten, Robeer & Grimmelikhuijsen (2023), PAR, covering street-level bureaucrats and AI recommendations, discretion, and trust — solid confidence. 4. Madan & Ashok (2023), GIQ, a systematic review of AI adoption and diffusion in public administration — good confidence. 5. Giest & Klievink (2022/2024) in Public Management Review, examining how AI is reshaping the bureaucrat's role across organizational contexts — reasonable confidence, though I should double-check the exact citation details. I'm also considering several alternates as substitutes: Meijer, Lorenz & Wessels (2021); van Noordt & Misuraca (2022); Neumann, Guirguis & Steiner (2024); Wirtz et al.; Tangi et al. (2023); and Bullock et al. on AI, discretion, and bureaucracy. Given I can't browse to verify these, I think the right move is to flag these as candidate entries pending verification, give the user a checklist to confirm them, and offer to either let them verify and use my table or have them supply five confirmed articles for me to populate — and since they want a one-page Word document, I'll need to address that I can't generate files directly. Since I can't attach a .docx, I'll provide paste-ready content along with precise formatting guidance — landscape orientation, narrow margins, small font size, and a compact table structure — so they can quickly assemble it in Word themselves. I also want to be upfront that citation specifics like volume, page numbers, DOI, and exact open-access status need independent verification, and that the summarized findings reflect my training knowledge rather than a fresh reading of the actual papers. Rather than cluttering the one-page table with a verification column, I'll add a short notes section below it flagging what needs double-checking. Drafting the first entry now — Zuiderwijk, Chen & Salem (2021) on AI in public governance, published in Government Information Quarterly. I'm being careful not to invent specific numbers like how many studies were reviewed since I'm not confident on that detail, so I'll describe the scope more generally instead. For findings, I'm noting the benefits side (efficiency, reduced administrative burden, better service quality and decision support) against the risks (accountability gaps, bias, privacy, data quality issues, over-reliance), plus the point that successful implementation depends more on data infrastructure, skills, and legal frameworks than on the technology itself, and that the evidence base is fragmented with little empirical work on workforce impacts. For implications, I'm drawing out that this points toward task-level augmentation rather than job replacement, with recommendations around investing in data governance and skills, building accountability mechanisms, being skeptical of vendor claims, and documenting a baseline to address the evidence gap. Now starting the second row for Ahn & Chen's study. g and redeployment commitments, and measure employee readiness before deployment. Next, I'm recalling Selten, Robeer & Grimmelikhuijsen (2023) in Public Administration Review, which found street-level bureaucrats trust AI recommendations more when they confirm their own professional judgment. I believe the setting involved a survey experiment with frontline public professionals, possibly Dutch police officers, though I want to flag that detail for verification. Key findings show professionals accept algorithmic advice mainly through confirmation bias, discounting it when contradicted, and that transparency alone doesn't fix this — meaning automation risks entrenching existing practice rather than improving decisions. This implies agencies need decision-support protocols, training on overriding AI, audits of override patterns, and clear accountability resting with staff rather than the system. Now I'm turning to Madan & Ashok (2023) on AI adoption. I should budget for sustained scaling rather than one-off pilots, with workforce plans emphasizing reskilling into data stewardship and exception handling. For Giest & Klievink's work on Dutch public organizations, I'm noting that AI reshapes bureaucratic roles rather than eliminating them—discretion shifts upstream to system design while frontline discretion narrows, with effects varying by organizational context and carrying risks of de-skilling and diffused accountability between IT and administrative staff. This suggests job descriptions need rewriting to reflect these new responsibilities. I'm also keeping a backup list of alternative sources in case any primary citation fails verification: Meijer, Lorenz & Wessels (2021) in PAR, van Noordt & Misuraca (2022) in GIQ, Neumann, Guirguis & Steiner (2024) in PMR, and Tangi et al. (2021) on digital government transformation factors in GI, while excluding Wirtz, Weyerer & Sturm since it's from 2020. I'm also considering Maragno et al. (2023) on chatbots in public entities since it's highly relevant to administrative support functions, plus some Nordic/UK work on AI in local government, Bright et al. (2024) on generative AI in the public sector, and various Alan Turing Institute reports as potential backups. I should also flag OECD's "Governing with Artificial Intelligence" (2024) as useful grey literature, and especially the ILO working paper by Gmyrek, Berg & Bescond (2023) on generative AI and jobs — it's open access and finds clerical/admin work has the highest automation exposure, making it extremely relevant for administrative support positions. Since the user asked for exactly five articles, I'll present five in the main table and note these as strongly recommended alternates, flagging that the ILO paper is a working paper rather than peer-reviewed. Worth also noting that most public-sector AI studies don't actually quantify admin job displacement — the real evidence for that comes from occupational exposure literature like ILO's work and Eloundou et al.'s "GPTs are GPTs." For the output, I'll give the table content in markdown here for readability, but since markdown won't paste as a real table into Word, I should also provide clear step-by-step instructions for inserting a table and manually filling it in. I should also mention the HTML-to-Word trick as an optional method, since it would actually produce a formatted table the user could open directly — worth offering as a bonus rather than the default. I want to keep the response practical: note what I can and can't verify, give the table content, and the Word setup instructions without overloading the response. I'll structure the answer with the table, formatting steps, and a verification checklist noting which details per source need double-checking, plus suggested backup sources given the evidence gaps. I'll keep each cell concise and avoid stating specific numbers I'm not confident about, then offer to refine further once the user verifies things. ## Before the table: one important flag I don't have live internet access in this conversation, so I cannot run the searches, open the articles, or confirm their current open-access status and citation details. I also can't generate and attach a .docx file. What I've done instead is build the complete table using five real, post-2020, peer-reviewed articles that I'm reasonably confident exist in the open literature and that fit your brief. **Every citation must be verified before this goes into a strategic planning package**, and the summarised findings should be checked against the actual text — my summaries come from training knowledge, not from reading the current versions. I've included a verification checklist and backup sources at the end, plus Word formatting steps to get it onto one page. If you'd prefer, send me the five articles you locate and I'll populate the table from your sources instead. --- ## Research Summary Table — AI & Automation in Government: Implications for Administrative Support (DRAFT — citations pending verification) | Study Information | Key Findings | Implications for the Department | |---|---|---| | **Implications of the use of AI in public governance: A systematic literature review and a research agenda**
Zuiderwijk, Chen & Salem (2021)
*Government Information Quarterly*, 38(3) — open access
**Setting:** Systematic review of peer-reviewed literature; multi-country, all levels of government
**Goal:** Map benefits, risks and implications of AI in public governance; set research agenda | • Benefits concentrate in efficiency, reduced administrative burden, service quality, decision support
• Risks: opacity, weak accountability, bias, privacy, poor data quality, over-reliance on outputs
• Success depends on data infrastructure, skills and legal frameworks — not the technology
• Evidence base fragmented; little empirical research on workforce/job effects | • Frame the 5-year goal as *task-level* automation, not position elimination
• Sequence investment: data governance and staff capability before tool deployment
• Build accountability and human-review mechanisms into every deployment
• Treat vendor efficiency claims as unverified; establish our own baseline metrics | | **Digital transformation toward AI-augmented public administration: The perception of government employees and the willingness to adopt AI in government**
Ahn & Chen (2022)
*Government Information Quarterly*, 39(2)
**Setting:** Survey of government employees *(verify jurisdiction/sample)*
**Goal:** Identify factors shaping employee perceptions of and willingness to adopt AI | • Willingness to adopt driven by perceived usefulness, organisational/managerial support and training
• Job-security anxiety and low AI literacy suppress adoption
• Trust in the organisation matters as much as trust in the tool
• Employee perceptions vary by role and seniority | • Adoption risk is cultural, not technical — budget for change management
• Communicate redeployment/no-forced-layoff intentions early and explicitly
• Involve Administrative Support staff in design and tool selection, not just rollout
• Survey divisional readiness and AI literacy as a planning baseline | | **"Just like I thought": Street-level bureaucrats trust AI recommendations if they confirm their professional judgment**
Selten, Robeer & Grimmelikhuijsen (2023)
*Public Administration Review*, 83(2) — open access
**Setting:** Experimental study with frontline public professionals *(verify sample/country)*
**Goal:** Test when and why frontline staff accept algorithmic recommendations | • Staff accept AI advice mainly when it confirms their own judgment; discount contradicting advice
• Explanations/transparency features alone did not produce appropriate reliance
• Automation can entrench existing practice rather than improve decision quality
• Confirmation bias undermines the assumed consistency gains | • Do not assume decision-support tools will improve consistency or fairness
• Require protocols on when staff must document an override; audit override patterns
• Train on *critical* use, not just tool operation
• Keep final accountability with the employee and make that explicit in procedures | | **AI adoption and diffusion in public administration: A systematic literature review and future research agenda**
Madan & Ashok (2023)
*Government Information Quarterly*, 40(1)
**Setting:** Systematic review across countries and government levels
**Goal:** Synthesise adoption/diffusion determinants and identify research gaps | • Determinants cluster into technological, organisational and environmental factors
• Organisational capacity and leadership commitment are the strongest predictors
• Public sector lags private sector; pilots frequently fail to scale
• Main brakes: skills shortages, legacy data/systems, procurement and legal constraints | • Build the plan around foundations first; resist pilot proliferation
• Fund sustainment, maintenance and data stewardship, not just implementation
• Identify executive sponsor and governance body before scaling
• Create a reskilling pipeline: admin staff into data quality, exception handling, QA roles | | **More than a digital system: how AI is changing the role of bureaucrats in different organizational contexts**
Giest & Klievink (2022/2024)
*Public Management Review* *(verify year/volume)*
**Setting:** Qualitative comparative case study, public organisations (Netherlands)
**Goal:** Examine how AI reshapes bureaucratic roles across organisational contexts | • AI reshapes rather than removes administrative roles; new tasks emerge (data handling, system monitoring, explaining outputs to citizens)
• Discretion shifts "upstream" to system design and data choices; frontline discretion narrows
• Effects are highly context-dependent (task type, structure, maturity)
• Risks of de-skilling and of accountability diffusion between IT and administrative staff | • Expect role redesign and redeployment as the dominant workforce impact
• Update job descriptions, classifications and competency profiles proactively
• Create new role families (data steward, AI liaison, quality assurance)
• Clarify accountability boundaries between IT and Service Delivery
• Monitor for de-skilling and loss of institutional knowledge | --- ## Getting it onto one page in Word 1. **Layout → Orientation → Landscape** 2. **Layout → Margins → Custom:** 0.5" all sides 3. **Insert → Table:** 3 columns × 6 rows 4. Paste each cell's content individually (pasting the table above won't create a Word table) 5. Select all → font **8 pt** (Calibri or Aptos); **Layout → Paragraph spacing: 0 pt before/after; line spacing: single** 6. Column widths: ~25% / 37.5% / 37.5% 7. Header row: bold, shaded, **Table Layout → Repeat Header Rows** 8. Bullets: use the smallest bullet style with a 0.1" indent to save space If you'd like, I can output this as HTML that you can save with a `.doc` extension and open directly in Word as a pre-formatted table — that avoids the cell-by-cell paste. --- ## Verification checklist (do this before circulating) For each of the five entries, confirm: - [ ] Article exists with the author/title/year as stated (search the exact title in Google Scholar) - [ ] **Publication date is after 2020** — note that journal "online first" and issue years often differ; Giest & Klievink in particular may be listed as 2022 online / 2024 in issue - [ ] Full text is freely accessible **without** institutional login or paywall (check for a PDF link, the journal's open-access badge, or an institutional repository copy such as Leiden, Utrecht or Birmingham repositories) - [ ] Volume, issue, article number and DOI are accurate - [ ] The setting/sample details I've marked *"verify"* are correct — I'm least confident on the Ahn & Chen jurisdiction and the Selten et al. participant group - [ ] My summarised findings match the article's abstract and conclusions **Backup articles** if any of the five fail verification (all post-2020, likely open access — verify the same way): - Meijer, Lorenz & Wessels (2021), "Algorithmization of bureaucratic organizations," *Public Administration Review* - van Noordt & Misuraca (2022), "Artificial intelligence for the public sector," *Government Information Quarterly* - Neumann, Guirguis & Steiner (2024), "Exploring artificial intelligence adoption in public organizations," *Public Management Review* - Maragno, Tangi, Gastaldi & Benedetti (2023), on introducing chatbots in public entities, *Public Management Review* — especially relevant to citizen-facing administrative functions --- ## One gap worth raising in the strategic planning discussion The public-administration literature above tells you a great deal about *adoption conditions, role change and governance risk*, but very little about **how exposed administrative support occupations actually are to automation**. That question is answered in a different literature — occupational exposure studies, which consistently find clerical and administrative work among the most exposed categories (e.g. the ILO's 2023 generative AI and jobs analysis, and Eloundou et al. on occupational exposure to language models). Those sources are open access but are working papers or general-economy studies rather than peer-reviewed public-sector research, so they may not meet your "academic article" criterion. I'd suggest either adding one as a sixth row, or noting the gap in your covering memo — otherwise the summary risks understating the scale question your executive will almost certainly ask.

kkbhb,