Introduction
Artificial intelligence is widely promoted as a way to modernize public services, but new research suggests most government AI initiatives never get off the ground. According to a study by Adobe, the vast majority of proposed public-sector AI use cases remain stuck in planning or pilot phases. The issue is rarely the technology itself. Instead, organizations often lack clear ownership, defined success metrics, and a process for moving a successful pilot into everyday operations.
What Happened
Adobe surveyed 5,633 executives across 12 industries and 11 markets, including 502 public-sector leaders. The findings show that 90 out of every 100 proposed AI use cases in government never reach production. Only 7 percent of public-sector respondents said their organization has a formal AI owner with a defined title and budget authority. Meanwhile, 92 percent described the relationship between business functions and IT as siloed, transactional, or consultative, and just 4 percent said their organizations comprehensively measure AI's impact.
Adobe labeled the common bottleneck the "Frozen Middle." These organizations have moved past initial experimentation but have not reached expansion or transformation. A tool might work well in a single department, yet fail to spread because no repeatable process exists to approve and fund the next step.
Why This Matters
The findings extend beyond government agencies. Small businesses that build AI tools for public clients, or adopt AI internally, can face the same delays when leadership, IT, security, compliance, and frontline teams operate separately. A pilot may prove that a system works technically, but scaling it requires decisions about budget, risk, data handling, training, and measurable results.
For companies selling to government, the research also highlights what agencies increasingly expect from vendors. Proposals that explain how performance will be monitored, how risks will be managed, and who will make decisions when circumstances change may stand out in competitive procurement processes.
Key Takeaways
- Most public-sector AI pilots never scale beyond initial testing because no single person owns the initiative after launch.
- Formal AI ownership with budget authority is rare in government but strongly associated with higher-performing organizations.
- Siloed relationships between business units and IT slow down or prevent expansion.
- Involving security, privacy, and accessibility teams early can avoid expensive rework later.
- Small businesses can apply lightweight governance by naming an accountable project owner, defining intended outcomes, and setting evidence-based criteria for continuing or ending a project.
Conclusion
AI adoption in the public sector may depend less on selecting the perfect tool and more on deciding who owns the work, how results will be proven, and what happens after the pilot ends. Small businesses do not need a large AI department to benefit from these lessons. Assigning one accountable person, documenting the intended outcome and baseline, involving technical and oversight partners during design, and establishing clear criteria for expansion can help turn promising pilots into production-ready solutions.
Adobe's full findings are available in its State of AI Readiness in the Public Sector report. The central message is straightforward: successful AI implementation is as much about organizational clarity as it is about technology.

Discussion
Join the conversation
Thoughtful reactions, questions, and follow-up ideas help shape the next story.