Government can scale AI responsibly by standardizing the foundations—governance, data practices, infrastructure, skills, procurement and oversight—while designing each application around the work, risks and people of a particular public function. Scale should mean repeatable capabilities and accountable controls, not one model or decision rule imposed across every agency.
What does “vertical rigor at horizontal scale” mean?
“Vertical” and “horizontal” are useful ways to frame the challenge, not established government AI standards. Here, vertical rigor means grounding an AI system in a specific mission and workflow: its users, data, legal and operational context, consequences, and measures of success. Horizontal scale means making shared capabilities and guardrails available across agencies so each does not have to build them from scratch.
As an Amazon Associate I earn from qualifying purchases.
The balance matters because tasks differ. Organizing documents or improving a structured administrative workflow is not equivalent to supporting policymaking or accountability, where judgments may be contested and data, transparency, and governance needs are more complex. Reuse the foundations; tailor the system and controls to the mission.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →How quickly is government AI adoption growing?
OECD’s Digital Government Outlook 2026 reports that AI use in internal processes rose from 23 of 33 measured OECD countries in 2023 (70%) to 31 of 36 in 2025 (86%). For public services, it rose from 22 of 33 (67%) to 27 of 36 (75%). These are country-level survey shares, not percentages of agencies, deployments, or spending. The comparison has uneven coverage: 2025 data were unavailable for Germany and the United States, and several countries lacked 2023 data.
#1 Best Overall
Use was less widespread in areas involving public judgment and scrutiny: in 2025, 13 of 36 OECD countries (36%) reported AI use supporting policymaking, and 12 of 36 (33%) reported use to strengthen oversight and accountability. The gap reinforces the case for matching safeguards and ambition to the task rather than treating adoption as a single government-wide score.
In the United States, the Government Accountability Office’s July 2025 review found reported AI use cases in inventories from 11 selected agencies grew from 571 in 2023 to 1,110 in 2024; reported generative-AI cases rose from 32 to 282. These are agency-submitted inventory counts reviewed by GAO, not a census of federal use or proof that every listed case was operational. They should not be compared directly with OECD country survey rates.
Why can’t government simply deploy one system everywhere?
Different tasks carry different consequences
A tool that helps staff find information or route forms may be suitable for a repeatable workflow with human review. A system that shapes policy, allocates benefits, or informs oversight can affect rights, public trust, or accountability in ways that demand stronger evidence, transparency, and routes to challenge decisions. The same technical capability can therefore require different controls in different settings.
Free tools Windows power users keep installed
One-click scans. No signup required.
Data and systems are uneven
OECD identifies data availability, legacy systems, skills shortages, financial constraints, and demanding privacy, transparency, and representation requirements as adoption barriers. A shared platform cannot make unsuitable or inaccessible data fit for a mission, nor remove the work of integrating with older systems.
Policy and technology move at different speeds
GAO reports that agencies cited difficulty keeping pace with rapid technological change while complying with policy, alongside concerns about technical resources, budgets, and keeping use policies current. Scaling needs governance that can be updated as capabilities and context change, rather than a one-time approval.
What should be shared across agencies—and what should stay mission-specific?
| Shared foundations | Mission-specific decisions |
|---|---|
| Governance processes, risk-management methods, procurement practices, and common accountability expectations | Whether AI is appropriate for the public problem and workflow at all |
| Reusable infrastructure, data stewardship capabilities, technical skills, and investment capacity | Which data are fit for the task, and what privacy, representation, and quality constraints apply |
| Common approaches to transparency, performance evaluation, monitoring, and auditability | What errors mean for affected people, how consequential or contestable decisions are, and what human oversight is needed |
| Partnership mechanisms and opportunities to learn across government | How a system fits local processes, legacy technology, workforce needs, and routes for user engagement |
This is not a choice between centralization and agency autonomy. Shared capabilities can lower duplication, while mission owners remain responsible for whether an application fits its context and for the outcomes it produces.
How can agencies build AI adoption responsibly?
- Start with the public problem. Define the service or administrative need before selecting a model. OECD warns that automating a process without first examining it can preserve inefficiencies rather than solve them.
- Classify the task and its stakes. Identify who may be affected, how consequential errors could be, whether judgments are contestable, and what data and transparency requirements follow. Do not assume controls suitable for document handling are adequate for policy or oversight.
- Use shared foundations, with a named mission owner. Make governance, infrastructure, skills, procurement, and data capabilities reusable, but keep responsibility for context, performance, and public impact with the function deploying the system.
- Set proportionate guardrails before deployment. Define accountability, risk controls, transparency, and how performance will be evaluated and monitored. GAO’s AI Accountability Framework organizes practices around governance, data, performance, and monitoring; it is a 2021 principles-based reference, not a statement of current administration requirements.
- Engage the people who use or experience the system. Involve civil servants and affected communities during development and use, and provide meaningful ways to understand or raise concerns about system-supported processes.
- Review and adapt. Reassess data, performance, risks, and controls as the system, mission, and technology change. Reuse should make this review easier to repeat, not remove it.
What do current use cases show—and not show?
OECD’s 2025 report, Governing with Artificial Intelligence, analyzes 200 use cases across 11 government functions. Examples include chatbots that answer citizen questions or assist with forms, AI for disaster anticipation and response, and tax fraud detection. GAO’s selected federal examples include a Veterans Affairs medical-imaging automation effort and a Department of Health and Human Services effort to extract information from publications to identify possible poliovirus outbreaks.
Recommended Free Tools
These examples illustrate why mission context matters: a public-service chatbot, disaster-response tool, fraud detection system, and medical-imaging effort serve different purposes and face different risks. The reports describe examples and efforts; they do not establish measured benefits or successful production outcomes for each one.
What makes government-wide oversight difficult?
In a report published in September 2025, GAO identified 94 AI-related requirements with government-wide scope or implications, counting requirements as of July 2025, and 10 executive-branch groups with an AI oversight or advisory role. The figures describe the landscape identified at that time, which can change as policies and arrangements evolve. They illustrate why agencies need clear ownership and coordination alongside shared rules.
OECD’s framework groups the conditions for trustworthy adoption into three connected areas: enablers such as governance, data, infrastructure, skills, investment, procurement, and partnerships; guardrails including transparency, accountability, and risk management; and engagement with citizens and civil servants. It advises governments to account for their context and maturity rather than apply every measure at once. As the OECD chapter puts it, “Governments should take a systems approach and seek to anticipate future changes.”
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
