Preparation Process for Artificial Intelligence Success in Federal Agencies

While federal agencies struggle to operationalize artificial intelligence, research shows that most pilot projects fail or are abandoned due to a lack of management, data preparation, and institutional change management rather than technical insufficiency.

Status of Pilot Projects

Research reveals that very few integrated AI pilots generate significant financial value, with organizations abandoning a substantial portion of their projects before reaching production. Most of the time, the model works, but the surrounding elements fail.

No one defines what is ready for production, establishes a baseline, and employees do not trust tools that are not integrated into their workflows. Management, meanwhile, becomes either an afterthought or a bottleneck.

Impact of the Data Environment

Artificial intelligence makes every condition in the data environment upon which it is built even more pronounced. A model reading contradictory records can reproduce that contradiction at machine speed.

When rewrites are made to systems, the scale of the impact area can be much larger than any error a single individual could make.

Business Units and Technology

While technical roadmaps define what is deployed, business unit owners experience continuous change in technology. Models are replaced or deprecated on a timeline measured in weeks.

It is of great importance that the institutional change management processes required for business units to assimilate this change are handled with care.

Regulations and Obligations

Within the executive branch, the question of whether agencies need an institutional AI strategy has been clarified. Relevant regulations require every agency to appoint a chief artificial intelligence officer.

Agencies are also required to establish an AI governance board, publish a strategy and compliance plan, and maintain a public use case inventory.