Government Evaluation TIG Week: Responsible Innovation: How Government Agencies Navigate AI in Evaluation by Eliza Carroll and Will Minor

Date: Thursday, July 3, 2025

Hey there! We’re Eliza Carroll and Will Minor from Deloitte’s Evaluation and Research for Action Center of Excellence. In our work, we collaborate with government agencies to evaluate their impactful initiatives.

AI is revolutionizing evaluation by increasing efficiency across all steps of the process, from data aggregation to logic modeling and more. Government agencies at all levels—local, state, and federal—recognize AI’s potential to enhance evaluation. A recent article by Deloitte highlights how state and local governments are scaling their use of Generative AI. However, some enthusiasm is tempered by concerns over data privacy, security, and the need for transparency. This hesitancy reflects agencies’ unique responsibilities to safeguard sensitive information and maintain public trust.

Hot Tips

Below, we break down some of the biggest concerns we’ve heard from government agencies about implementing AI, along with potential mitigation strategies.

Concerns	Description	Solution
Validity of Findings	Agencies often question the accuracy and dependability of AI-generated insights. Sometimes, AI outputs appear credible but are actually incorrect or misleading—especially if the underlying data or algorithms are incomplete, biased, or flawed. This “false confidence” can undermine critical decisions.	Pilot test AI tools and compare outputs with results from established manual evaluation methods. This helps identify discrepancies and make improvements before broader implementation.
Data Security	Implementing AI often requires access to sensitive government data, raising concerns about data governance and access. There’s a risk of unauthorized access, whether through cyber threats or internal misuse, which could breach privacy laws or regulations.	Use closed AI models and strong data encryption to restrict public access, minimize data breach risks, and protect sensitive information.
Public Trust	Public trust in government depends on transparent, accountable decision-making. AI can introduce skepticism, especially when its recommendations are opaque or hard to interpret. Lack of explainability can erode stakeholder confidence and trust in agencies using AI.	Provide clear, accessible explanations and documentation about how AI models work, including limitations and decision logic. Transparently communicate when and how AI is used in evaluations, both internally and to the public, to build trust and accountability.
Environmental Impacts	The substantial computational resources required to train and operate advanced AI models can have notable environmental consequences. For example, the United Nations projects that global AI-related water consumption could reach 4.2 to 6.6 billion cubic meters by 2027—surpassing Denmark’s total annual water usage—highlighting the need for sustainable AI practices.	Prioritize AI deployment for high-impact tasks where the technology delivers substantial value or efficiency improvements, thereby justifying its environmental footprint and promoting responsible resource use.

Conclusion

As AI technology continues to evolve, so too will its applications in government evaluation. Ongoing innovation will bring new opportunities and challenges that require government agencies and those who support these agencies to remain adaptable and proactive in refining our approaches. By continuing to collaborate and problem solve together we can make this innovation responsible.

Have you supported the AI shift in government agencies? And, are there other concerns you have heard? Share in the chat to get a conversation going!

The American Evaluation Association is hosting Gov’t Eval TIG Week with our colleagues in the Government Evaluation Topical Interest Group. The contributions all this week to AEA365 come from our Gov’t Eval TIG members. Do you have questions, concerns, kudos, or content to extend this AEA365 contribution? Please add them in the comments section for this post on the AEA365 webpage so that we may enrich our community of practice. Would you like to submit an AEA365 Tip? Please send a note of interest to AEA365@eval.org. AEA365 is sponsored by the American Evaluation Association and provides a Tip-a-Day by and for evaluators. The views and opinions expressed on the AEA365 blog are solely those of the original authors and other contributors. These views and opinions do not necessarily represent those of the American Evaluation Association, and/or any/all contributors to this site.